Bibliography - Confex · Bibliography Charles C, ... paper attempts to provide greater conceptual clarity about shared treatment ... mation to be collected as relevant examples

Bibliography Charles C, Gafni A, Whelen T. Shared decision-making in the medical encounter: what does it mean? (Or it takes at least two to tango). Soc Sci Med 1997;44:681–692 .......................................2 Charles C, Gafni A, Whelen T. Decision-making in the physician–patient encounter: revisiting the shared treatment decision-making model. Soc Sci Med 1999;49:651–661. ............................14 Joosten EA, DeFuentes-Merillas L, de Weert GH, Sensky T, van der Staak CP, de Jong CA. Systematic review of the effects of shared decision-making on patient satisfaction, treatment adherence, and health status. Psychother Psychosom 2008;77:219–226. . ...................................25 Wilson SR, Strub P, Buist AS, et al. Shared treatment decision making improves adherence and outcomes in poorly controlled asthma Am J Respir Crit Care Med 2010;181:566-77. . ..............33 Wilson SR, Strub P, Buist AS, et al. Shared treatment decision making improves adherence and outcomes in poorly controlled asthma Am J Respir Crit Care Med 2010;181:566-77. ONLINE SUPPLEMENT. .............................................................................................................................45 Tapp H, Hebert L, Dulin MF. Comparative effectiveness of asthma interventions within a practice based research network. BMC Health Services Research 2011, 11:188………………..72

1

Pergamon

S0277-9536(96)00221-3

Soc. Sci. Med. Vol. 44, No. 5, pp. 681-692, 1997 Copyright © 1997 Elsevier Science Ltd

Printed in Great Britain. All rights reserved 0277-9536/97 $17.00 + 0.00

S H A R E D D E C I S I O N - M A K I N G IN THE M E D I C A L ENCOUNTER: WHAT DOES IT MEAN? (OR IT TAKES AT

LEAST TWO TO TANGO)

CATHY CHARLES, ''2"3. A M I R A M G A F N V '3 and TIM W H E L A N 3'4'5

'Department of Clinical Epidemiology and Biostatistics, McMaster University, Hamilton, Ontario, Canada, "Centre for Health Economics and Policy Analysis, McMaster University, Hamilton, Ontario, Canada, 3Supportive Cancer Care Research Unit, McMaster University, Hamilton, Ontario, Canada,

'OCTRF--Hamilton Regional Cancer Centre, Hamilton, Ontario, Canada and 5Department of Medicine, McMaster University, Hamilton, Ontario, Canada

Abstract--Shared decision-making is increasingly advocated as an ideal model of treatment decision- making in the medical encounter. To date, the concept has been rather poorly and loosely defined. This paper attempts to provide greater conceptual clarity about shared treatment decision-making, identify some key characteristics of this model, and d scuss measurement issues. The particular decision-making context that we focus on is potentially life threatening illnesses, where there are important decisions to be made at key points in the disease process, and several treatment options exist with different possible outcomes and substantial uncertainty. We suggest as key characteristics of shared decision-making (1) that at least two participants--physician and patient be involved; (2) that both parties share information; (3) that both parties take steps to build a consensus about the preferred treatment; and (4) that an agreement is reached on the treatment to implement. Some challenges to measuring shared decision- making are discussed as well as potential benefits of a shared decision-making model for both physicians and patients. Copyright © 1997 Elsevier Science Ltd

Key words- -shared treatment decision-making, physician/patient communication

INTRODUCTION

Shared decision-making is increasingly advocated as an ideal model of treatment decision-making in the medical encounter (Veatch, 1972; Brody, 1980; Quill, 1983; Brock and Wartman, 1990; Gray ei' al.,

1990; Emanuel and Emanuel, 1992; Levine et al.,

1992; Deber, 1994). Yet, it is by no means clear what shared decision-making really means or the criteria by which to judge what falls within or outside the boundaries of this model. In order to be able to evaluate the merits and limitations of a shared decision-making model it is first necessary to be clear about what the model is. To date, the concept of shared decision-making has been rather poorly and loosely defined. This leads to cons:tier- able confusion because the same conceptual label can be used to subsume different underlying philos- ophies and principles of physician-patient interaction. Our goal in this paper is to provide greater conceptual clarity in thinking about shared decision-making, to identify some basic characteristics of this model and to discuss measurement issues.

Improved conceptualization of shared deci,,;ion- making would have several benefits. It would make more explicit what advocacy of shared deci,,don-

*Author for correspondence.

making means, allow for easier recognition when it does occur, and perhaps facilitate its practice by physicians and patients who have a preference for this model of joint decision-making. In so far as shared decision-making has been linked with positive patient outcomes (e.g. satisfaction and improvements in functional status), clarification of this model is clinically relevant (Egbert et al. , 1964; Schulman, 1979; Greenfield et al., 1985, 1988; Brody et al. , 1989; Wennberg, 1990; Mahler and Kulik, 1990; Lerman et al. , 1990). Greater conceptual clarity could also guide research in this area by providing clearer direction on the types of information to be collected as relevant examples of shared and non-shared decision-making.

The increased and fairly new interest in shared decision-making derives from a number of different factors. For example, informed consent, now ethi- cally and legally ensconced as a patient right, seems to imply at least a minimum of shared decision- making in the form of patient consent to treatment prior to any intervention (Sutherland e t al., 1989). Moreover, the principle of "informed choice", ie. disclosure of treatment alternatives rather than merely informed consent has been endorsed at several government levels in Canada and the United States (Evans, 1987; Greene, 1992; Ontario Ministry of Health, 1994; Nayfield et al. , 1994).

681

2

682 Cathy Charles et al.

Interest in shared decision-making also has its origins in the consumer rights movement. Here, concern with patient participation in treatment decision-making has moved well beyond informed consent to include broader principles of patient autonomy, control, and patient challenge to physician authority (Haug and Lavin, 1983; Ende et al. ,

1989; Charles and DeMaio, 1993; Llewellyn- Thomas, 1992, 1995). Shared decision-making is seen as a mechanism to decrease the informational and power asymmetry between doctors and patients by increasing patients' information, sense of autonomy and/or control over treatment decisions that affect their well-being (Eddy, 1990; Ryan, 1992; Emanuel and Emanuel, 1992).

A final factor is the changing nature of medical practice. During the last 20 to 30 years, there has been a dramatic shift away from acute care to chronic care and caregivers often manage illnesses or combinations of illnesses rather than cure disease. For such patients, sickness is not just a temporary status; rather, long-term and chronic illness may become a permanent part of their identity and status. In such cases, the physician patient relationship is potentially a long-term one. The plethora of new drugs available requires that physicians work closely with such patients to develop the optimum pharmacological solution, a process that takes time, continuous monitoring and adjustment of medication types and levels. This process is likely to work best if both patients and physicians have a role in managing the illness and medication regimens.

For certain diseases, such as cancer, which are both potentially life-threatening and widely prevalent, there are key treatment decision points which may occur only once and arise early on in the course of the disease which have major consequences for the patient. Women with early stage breast cancer, for example, may be faced with the decision to have a lumpectomy versus mastectomy, and, following surgery, whether to have adjuvant chemotherapy and/or radiation. These are decisions which cannot be delayed without potentially serious implications for the health of the patient. Here shared decision-making becomes particularly important to address in the medical encounter, first, because several treatment options exist with different possible outcomes, and substantial uncertainty. Second, there is often no clear-cut right or wrong answer. Third, treatments will vary in their impact on the patient's physical and psychological well- being (Pierce, 1993). It is this latter type of treatment decision-making context which is our focus in this paper and we use early stage breast cancer as an example of a potentially life threatening disease.

We recognize that there are different types of treatment decision-making contexts (e.g. emergency treatment, long-term monitoring of medications in the treatment of chronic disorders such as hyperten-

sion, palliative care) and that different models of treatment decision-making may be more or less appropriate or feasible in specific contexts. To discuss issues of shared decision-making in all these would be an enormous task; hence, we choose the specific decision-making context as described above as our focus for this paper.

This focus is at the micro as opposed to the macro level of analysis where, clinically, there are several treatment options available and the choice of the best treatment for a particular patient requires value judgements on the part of the patient and physician. Our paper does not address macro level economic constraints, i.e. where policy makers have decided that for certain medical conditions, there will be a limited number of treatment options available through public or third party insurers. Also, we limit the discussion of decision-making at the micro level to competent patients.

MODELS OF TREATMENT DECISION-MAKING

Shared decision-making is only one among several treatment decision-making models discussed in the literature (Veatch, 1972; Thomasma, 1983; Emanuel and Emanuel, 1992; Levine et al., 1992; Roter and Hall, 1992; Mooney and Ryan, 1993; Deber, 1994). Prominent among these are the paternalistic model, the informed decision-making model and the professional-as-agent model. Analysis of the prototype depictions of each of these models reveals some overlap in specific characteristics of each. In addition, many of these characteristics, such as information sharing, involve gradations rather than absolutes. In this paper, we want to clarify the central elements of the latter 3 models as they are portrayed in the literature, and argue that each falls short of depicting a model of shared treatment decision-making. We then identify what we see as necessary components of a model of shared decision-making, in particular, the exchange of both information and treatment preferences by both physician and patient and agreement by both parties on the treatment to implement.

On first glance, it seems easiest to differentiate the paternalistic model from shared decision-making because the former explicitly assumes a passive role for the patient in the treatment decision-making process. An early formulation of this model was Parsons' conceptualization of the sick role. Parsons argued that the sick role carried with it certain rights and obligations for patients (Parsons, 1951). Persons granted (by physicians) the (temporary) sick role status, for example, were excused from other role-related activities such as those of family and work, but they also had an obligation to try to get well, to seek expert help, and to comply with the medical regimen (Parsons, 1951). This model clearly placed the patient in a passive, dependent role vis-a-vis the physician as expert.

3

Shared treatment decision-making: what does it mean? 683

In more recent depictions of the paternalistic model, the physician is seen as dominating the medical encounter and using his skills to diagnose and recommend tests and treatments for the patient. In the extreme case, "the physician authoritatively informs the patient when the intervention will be in- itiated" (Emanuel and Emanuel, 1992). In the less extreme, the physician will give the patient sele,zted information and will encourage the patient to ~'on- sent to what the physician considers best (Ema~auel and Emanuel, 1992). The role of physician depicted in this model is guardian of the patient's best irLter- est. The physician does what he thinks is best for the patient, without eliciting the latter's preferences. Patient involvement (if there is any) is limited to providing consent to the treatment advocated by the physician (Emanuel and Emanuel, 1992).

Most would agree that this is not a model of shared decision-making in any sense. Efforts to for- mulate alternative treatment decision-making models have arisen, in part, in reaction to the perceived prevalence of the paternalistic approach (Levine et al. , 1992; Deber, 1994) which is viewed by many as inappropriate for many current treatment decision-making contexts. The extent to which this approach is currently practised by physicians is an empirical question; in emergency situations, for example, it may still be widely accepted and might, in practice, be the only feasible model for the task.

Both the informed and the physician-as-agent decision-making models derive from a recognition of informational asymmetry between patient and physician (Levine et al., 1992). As Hurley et al. (1992) note: "The crux of the information prolclem is that while the health care provider possesses better knowledge regarding the expected effective~aess of health care in improving health status, the individual knows best how improvements in health status affect his or her well-being" (p. 4). Techr,ical knowledge resides in one party to the interaction-- the physician, while preferences reside in the other-- the patient. Yet both types of information need to be combined if effective care that lead:~ to health status improvements valued by patients is to be provided (Hurley et al., 1992; Levine et al., 1992). In the informed model this is accomplished by increasing the patient's knowledge of the possible risks of alternative therapeutic options and their clinical effectiveness, so that patients can r~ake decisions that reflect both their preferences and the best scientific knowledge available (Hurley et al., 1992).

The informed decision-making model incorporates the idea of information sharing (primarily E~om physician to patient); but we would argue that information sharing does not necessarily lead to a sharing of the treatment decision-making process. In fact, since the informed patient has overcome the problem of information deficit, presumably she is now enabled to make the treatment decision on her

own. Theoretically, in this model, an informed patient no longer needs to share the treatment decision-making process because she now possesses both components (information and preferences) viewed as essential to the task (Levine et al. , 1992). In this model, treatment decision-making control is clearly seen to be vested in the patient (Eddy, 1990). The physician's role is limited to that of information exchange, communicating the needed technical or scientific knowledge to the patient (Williams, 1988; Mooney and Ryan, 1993). As noted by Emanuel and Emanuel (1992), the physician ~is pro- scribed from giving a treatment recommendation for fear of imposing his or her will on the patient and thereby competing for the decision-making control that has been given to the patient" (p. 2225). In other words, the physician's treatment preferences for the patient do not enter into the decision-making process.

If the paternalistic model leaves the patient outside the decision-making process, the informed model leaves the physician outside by limiting the role of the physician to one of information transfer. In the extreme case, information transfer can be done without the presence of any health care worker, for example, by the patient viewing an interactive video. We argue that unless both patient and physician share treatment preferences, a shared treatment decision-making process did not occur, no matter how much information may have been exchanged by either party.

The informed model is premised on the assumption that information is an enabling strategy, "empowering" the patient to become a more autonomous decision maker. Research has shown, however, that while patients typically express high preferences for information about their illness and its treatment (Cassileth et al., 1980; Strull et al., 1984; Beisecker, 1988; Blanchard et al., 1988; Ende et al., 1989; Lerman et al. , 1990; Beisecker and Beisecker, 1990; Waterworth and Luker. 1990; Silverstein et al., 1991; Biley, 1992; Deber, 1994), their preferences for participation in treatment decision-making are much more diversely distributed (Vertinsky et al., 1974; Strull et al., 1984; Pendleton and House, 1984; Ende et al. , 1989; Beisecker and Beisecker, 1990; Silverstein et al. , 1991; Degner and Sloan, 1992; Ryan, 1992; Hack et al., 1994). In other words, patients want information about their medical condition and treatment options without necessarily being responsible for making treatment decisions (Ende et al., 1989; Beisecker and Beisecker, 1990; Ryan, 1992).

An informed patient may prefer to make the decision herself (or be required to do so by the physician), to share the decision-making process, or to delegate this responsibility to the physician. In Scenario 1 in the Appendix, for example, the physician clearly wants the patient to make the treatment decision. The patient, while claiming to be

4


well informed, nonetheless prefers that the physician decide. Many patients faced with a serious illness, substantial uncertainty as to the outcome, and a time pressure to make a treatment decision among several competing alternatives, feel extreme psychological and/or physiological vulnerability, which may make it difficult for them to participate in treatment decision-making no matter how well informed they may feel (Gray et al., 1990; Ryan, 1992). Also, as this scenario points out. before patients can decide whether or not to share in decision-making, they must be offered the choice of participation by their physician (a point also made in a recent article by Ong et al.). Other patients may wish to participate but lack a systematic way of structuring the decision-making process. In this case, efforts to promote shared decision-making may well require interventions that not only provide patients with information but also with a way of thinking about treatment decision-making that helps them focus on key issues and evaluate relevant options.

One such intervention is the treatment decision aid, ranging in type from high cost interactive videos to low cost decision boards. Several decision boards have recently been developed and tested with women with breast cancer (Levine et al., 1992; Whelan et al., 1995). They provide the patient with detailed information about her treatment choices, outcomes, the probability of these outcomes and quality of life associated with each outcome. Treatment decision aids are a form of educational intervention. But they are also aids to treatment decision-making in that they provide a way of structuring the decision-making process, and breaking it down into a number of specific and sequential steps. As Llewellyn-Thomas (1995) notes: "A distinguish- ing feature of a decision aid is the inclusion of exer- cises designed to promote clarification of the patient 's values regarding what is at stake and what it is that he or she is trying to achieve as a result of treatment" (p. 104). Decision aids are intended to provide information and to promote "self help'" in the treatment decision-making process which enables the patient to more actively participate in this process, if this is her preference.

It seems fair to say that physicians such as medical oncologists who develop, test and use treatment decision-making aids are those who are already motivated to share treatment decision-making, since a primary goal of the aids is to help elicit patients'

*In the agency model, the physician is usually depicted as acting as the patient's agent, but this is not always the case. Williams, for example, defines a perfect agency relationship as one in which the physician gives the patient all the information the patient needs and the patient then makes the decision (Williams, 1988). Hence, the agency and informed models are also con- fused in the literature.

treatment preferences. Physicians adopting the paternalistic approach are unlikely to use such aids precisely because they can help overcome traditional professional dominance over interactions in the medical encounter (Freidson, 1970; West, 1984; Hill Beu£ 1989; Waitzkin, 1991). Whelan, for example, recently reported (1995) that 97% of women with breast cancer who were assigned to a group in which the physician used a treatment decision board with information about the risks and benefits of breast irradiation following lumpectomy felt that they were offered a treatment choice compared with 70% of women in the no decision board group. To return to our main point, however, an informed patient who perceives that she has choices may still prefer not to make the treatment decision.

The professional-as-agent model is the flip side of the informed patient model of treatment decision- making. Its goal is also to resolve the informational asymmetry between physician and patient. But here, "the professional-as-agent assumes responsibility for directing the health care utilization of the patient- . . . a s an agent trying to choose what the patient would have chosen, had she been as well-informed as the professional" (Evans, 1984, p. 75). In other words, in this model, the physician makes the treatment decision, either assuming that he knows, or having elicited the patient's preferences for future health states, lifestyle choices etc. Both components (information and preferences) then reside in the physician, rather than the patient, and the former becomes the sole decision-maker.* Legally physicians cannot implement a treatment without at least eliciting patient consent. In practice, there may still be some physicians who assume that they know the patient's treatment preference, and act on this without first explicitly testing this assumption. Such cases are not likely to attain public visibility unless the patient subsequently launches a legal suit, claiming violation of patient rights to informed consent.

While there have been critiques of the physician- as-agent model (Evans, 1984; Mooney and Ryan, 1993), our interest is not to take sides in this debate. Rather we argue that this model is also not necessarily one of shared decision-making. In the purest form of this model, the physician makes the treatment decision as if he had the same preferences as the patient; decision-making is again portrayed as a one-sided process. By definition, in this model the physician's treatment preferences do not count (are excluded). The only treatment preferences that matter are those of the patient.

In summary, several models of treatment decision-making have been developed, partially in reaction to the paternalistic model. A closer examin- ation reveals that none of these explicitly describes a process in which both physicians and patients necessarily share in decision-making, no matter how much information they share. The notion that information sharing and treatment decision-making are

5


two separate goals in the medical encounter is beginning to be recognized in the literature on doctor-patient communication (Ong et al., 1995). In the next section we argue that for shared treatment decision-making to occur, there needs to be a twoway exchange not only of information but also of treatment preferences.

CHARACTERISTICS OF SHARED DECISION-MAKING

In order to specify what falls or does not fall within the boundaries of shared decision-making, it is necessary to try to identify some of its key characteristics. Characteristics are signals of identification. The characteristics we identify can be thought of as minimum or necessary criteria for classifying a physician-patient decision-making interaction as shared decision-making, i.e. necessary but not always sufficient. There may well be other characteristics that are important that we have not included. It is also important to recognize that many of these characteristics are continuous rather than dichotomous variables in recognition of the fact that shared decision-making involves gradations rather than absolutes.

Shared decision-making involves at least two parlici- pants the physician and patient

The first characteristic of shared treatment decision-making is that it involves at least two par- t icipants--a clinician, who in many cases will be a physician, and a patient. This seems self-evident; as noted above, if only one person makes the decision, the process is not shared.

Frequently treatment decision-making involves more than one patient and one physician (or other health care professional) in a single or sequential medical encounter. For example, research on physician and elderly patient medical encounters increasingly focuses on triad rather than dyad relationsl~ips in recognitions that many elderly patients bring a relative or friend to the physician's office (Rosow, 1981; Coe and Prendergast, 1985; Adelman et al., 1987; Haug and Ory, 1987; Beisecker, 1988; Haug, 1994). Both conceptual and research interest iaas focused on the types of coalitions that can occur in such encounters. What this literature highlights is that family members or friends can play a variety of different roles within (or outside) the medical encounter relating to the patient's illness, treatment selection and management.

Roles which relate specifically to the treatment decision-making process can include, for example: (1) information gatherer, recorder or interpreter; (2) coach, e.g. prompting the patient to ask the physician certain questions; (3) advisor, e.g. advising the patient which treatment option to select; (4) nego- tiator, e.g. advocating on the patient's behalf regarding the timing or place of treatment or the patient priority in receiving treatment; (5) caretaker,

e.g. supporting and or reinforcing the patient's treatment decision. The involvement of family members in treatment decision-making may be particularly important with serious illnesses because of the stress engendered by the diagnosis, the uncertain outcome, and the potentially major impact of the illness trajectory and treatment management on other family members. These issues have received little attention in the shared treatment decision- making literature which focuses almost exclusively on the dyad relationship of physician and patient in the medical encounter.

When patients bring a relative or friend to a physician visit, the range and complexity of the interactional dynamics is automatically increased. In addition, the introduction of a third person enables the formation of coalitions. Coalitions (e.g. between the physician versus the patient and family member) may occur over such things as how much and what type of information is given by the physician, what is the best treatment and how and when to implement it. Studies of triad interactions between elderly patients, a family member and the physician have revealed the formation of numerous (and different) coalitions over numerous (and different) issues in a single medical encounter. However, the frequency and pattern of coalition formation is likely to vary depending on the treatment decision- making context. The triad relationship of elderly patient, physician and family member discussing treatment options for the long-term management of a chronic illness is different from the situation of a post-surgery woman with breast cancer who needs to make an important treatment decision under tight time constraints. Here, we hypothesize that fewer coalitions will be likely to occur because all parties will be motivated to reach consensus quickly on the treatment to implement so as to afford the best chance of recovery.

Also not well recognized in the literature on shared treatment decision-making is the fact that several physicians often participate in this process with a single patient. A breast cancer patient, for example, may have a family physician, surgeon, radiation oncologist, and medical oncologist, all of whom may have specific treatment preferences for the patient. For example, the surgeon may suggest a post-surgery treatment to the patient which she then has in mind when subsequently visiting the medical or radiation oncologist. If the latter presents a different treatment option, the result may be increased uncertainty and confusion for the patient. One study of patients in a cardiology unit reported that, not infrequently, five or more physicians were "making treatment decisions" about the patient's care (Lidz et al., 1985, p. 248).

These examples suggest that limiting the conceptualization of shared decision-making to a physician-patient dyad may not, in many cases, reflect the current realities of clinical practice, where other

6


participants may well be involved. It misses altogether the important role that the patient 's friends or family may play and the case of incompe- tent or seriously ill patients who require third parties to act on their behalf. There is almost no discussion in the shared decision-making literature on the implications of these situations for conceptualization and measurement of shared decision-making. Hence, we emphasize that a shared decision- making process requires at least two participants (i.e. it takes two to tango) but may often involve more than two.

Both parties (physicians and patients) take steps to participate in the process o f treatment decision-making

Much attention has been focused on exploring patient preferences for participation in treatment decision-making. Over the years, several scales have been developed, with preferences for participation often conceptualized as a single cont inuum of the amount of participation the patient prefers (e.g. from none, to shared participation, to complete patient control or autonomy (Strull et al., 1984; Beisecker, 1988; Brody et al., 1989; Ende et al., 1989; Beisecker and Beisecker, 1990)). The terms control and autonomy are usually not defined; nor is the term participation. To some, (Strull et al., 1984) participation seems to incorporate the idea of sharing on an equal basis, but the specifics of what this really means in terms of input by both physician and patient is left unclear. Presumably, the complete patient control end of the cont inuum cor- responds to the pure type informed model where the patient is the sole decision-maker. An implicit assumption behind much of this research seems to be that if patients express preferences, they will act in accordance with these preferences. Preferences become a proxy for behavioural intent, predictive of future behaviour.

Empirical research suggests that the link between patient preferences for participation in treatment decision-making and actual participation is not that strong. Patient preferences for information do not necessarily translate into information seeking behaviour; nor do patients who express preferences for some form of shared decision-making necessarily act on these in the medical encounter (Haug and Lavin, 1983; Beisecker, 1988; Beisecker and Beisecker, 1990; Ryan, 1992). Patient preferences

*Alternatively, it is possible that some patients preferring an autonomous decision-making role are constrained to share the process with their physician. In our culture, physicians have the authority to act as gate- keepers to health care. They control medical knowledge, technology, access to treatment and even norms of behaviour in the medical encounter (Ryan, 1992). At a minimum, the patient needs the physician's consent to her treatment choice in order to obtain most services.

for shared decision-making do not seem sufficient to make it happen in reality*.

If the majority of patients say that they have a preference for information about their illness and potential treatment options, while a much smaller number express preferences to participate in treatment decision-making, this raises some interesting questions. First, why do patients want information if it is not to be used by them instrumentally to help them make a treatment decision? We discuss this issue in the next section. Second, why do patients say that they do not want to participate in treatment decision-making and what does this preference really mean? We think that the latter is an under researched area and one that deserves much more attention because there are a variety of different possible explanations which have different implications for determining when a shared treatment decision-making process is appropriate.

First, a stated preference (e.g. a patient's stated preference not to participate in treatment decision- making) may reflect an underlying and salient per- sonality characteristic. In this case, introducing a decision aid to help the patient structure the treatment decision-making process is unlikely to have the desired effect, because the patient simply is not motivated to take an active role in the process and prefers that the physician decide.

For other individuals, a stated preference not to participate in treatment decision-making may reflect a situationally specific response. For example, a woman faced with the diagnosis of early stage breast cancer may not want to take an active role in treatment decision-making because of perceived information or skill deficits. Here, an intervention such as a decision aid can provide information and help the patient structure the decision-making process. The intervention could, in fact, help to promote a more active treatment role preference by the patient.

In still other instances patients may express a pre- terence for a passive role in treatment decision-making because they have learned through previous interactions that a more active stance is not well received by providers. For shared decision-making to occur, there needs to be complementary role expectations and behaviour between physician and patient around this issue. No matter how much the patient wants to participate, if the physician is not willing, then shared decision-making will not occur. Similarly, if the physician is willing but the patient is not, then the process will not be shared. It is in this sense that we emphasize that "it takes at least two to tango".

Finally, preferences for a passive role in treatment decision-making may reflect, in some instances, a cohort effect. For example, research suggests that elderly patients often prefer a more passive role in treatment decision-making than younger patients (Coe and Prendergast, 1985; Haug, 1994). This is

7

Shared treatment decision-making: what does it mean?

likely to change over time as the generation of baby boomers with higher education levels and highel expectations encounter the health care system.

In summary, the patient preferences literature provides only a partial answer to the question of what role patients want to play in treatment decision-making because it fails to consider that preferences may be situationally determined, and hence, subject to change. Moreover, by limiting the conceptualization and measurement of prefere~aces to only the patient side of the equation, it misses the important interactional dynamic with the plays- ician which is crucial to making shared decision- making happen in reality*. By and large, physicians set the norms of interaction in the medical encounter. If the physician is not motivated to share decision-making, the patient cannot force this to happen. Her only option is to seek out another physician whose expectations about how the decision-making process should occur is similar to hers.

If preferences are not enough, then what steps, do physicians and patients need to take in order to share in the treatment decision-making process? We suggest that for the physician it means first, es~:ab- lishing a conducive atmosphere so that the patient feels that her views about various treatment options are valued and needed (Brody, 1980). Second, it means eliciting patient preferences so that treatment options discussed are compatible with the patient's lifestyle and values. Third, it means transferring technical information to the patient on treatment options, risks and their probable benefits in as unbiased, clear and simple a way as is possible. Fourth, physician participation would also include helping the patient conceptualize the weighing process of risks versus benefits, and asking patients questions in order to ensure that the causal assumptions (information) underlying their treatment preferences are based on fact and not misconception.

Finally, shared decision-making would also involve the physician in sharing his treatment :recommendation with the patient, and/or affirming the patient's treatment preference. The physician would need to be careful, however, not to impose his values about the best treatment onto the patient.

There may always be a danger in a shared decision-making process that the physician's values will influence the patient, even if this is not his intent. Many patients have been socialized to think that the physician knows best, that they lack the expertise to make the treatment decision, that a trusting relationship with the physician means trust-

*This research does, however, demonstrate heterogeneity in patient preferences for participation in treatment decision-making. This in turn suggests that any atte~aapt to define a single normative model of treatment decision-making which both physicians and patients "ought" to follow might not fit with empirical reality.

687

ing his judgement about the most appropriate treatment, or that agreeing with the physician will result in better or more personal care.

For the patient, shared decision-making means that she must be willing to engage in the decision- making process, that is, to take responsibility for disclosing preferences, asking questions, weighing and evaluating treatment alternatives and formulating a treatment preference. This is a problem sol- ving task that goes beyond information transfer. Scenario 2 in the Appendix depicts such a shared decision-making process. Both patient and physician discuss and evaluate treatment options and together they build a consensus. We suggest that the test of a shared decision-making process is if both parties adopt the complementary roles outlined above, and if both parties are satisfied with their level of involvement.

Information sharing is a prerequisite to shared decision-making

In the typical case, information sharing is a prerequisite for shared decision-making. At minimum, the physician needs to lay out treatment alternatives and their potential consequences for the patient in order to obtain informed consent. Without such information, there might be nothing for the patient to evaluate and deliberate about. As noted earlier, patients may also bring information obtained through other means to the encounter. Both patients and physicians bring both information and values; it is not simply a question of physicians bringing knowledge and patients bringing values.

As a practical problem, it is often not clear what type and amount of information patients want from physicians, or why they want it. Patients may want more information than physicians think is needed instrumentally to help distinguish the pros and cons of various treatment options. Similarly, the types of information physicians desire from patients is not always clear. Is it the patient's full illness narrative or only those elements of the patient's story that the physician thinks relevant to suggesting treatment options?

The physician's primary professional obligation is to apply expertise in the treatment of patients. Time constraints, the potentially rapid course of many diseases, and financial incentives all operate to encourage physicians to complete the decision-making process as quickly as possible. In this situation it is likely that patient information is sought by the physician primarily for its instrumental use in identifying treatment options that are compatible with patient values.

The value of information about treatment alternatives, and their risks and benefits for the patient, is less well understood (Haug and Lavin, 1981; Beisecker and Beisecker, 1990; Ryan, 1992). As noted earlier, studies indicate that many patients have high preferences for information about their

8


illness and its management but do not engage in information seeking behaviours or in treatment decision-making. The value of information, from the patient's perspective, does not appear to lie (or lie solely) in its potential use as an aid to decision- making. More important may be the psychological reassurance or reduced uncertainty which information is thought to provide at a time of great stress and vulnerability (Mooney and Ryan, 1993). More research is needed in order to more completely understand the value of information from the patient's perspective.

Research evidence also suggests that when physicians infer patient preferences for information and for participation in treatment decision-making, they often fail to get them right (Strull et al., 1984; Waitzkin, 1984; Ryan, 1992). If the medical interaction is to reflect patient role preferences, these need to be made explicit by the patient either on her own initiative or elicited by the physician as part of the information exchange.

making (paternalistic and informed) but it remains a necessary prerequisite for shared decision-making.

However, mutual acceptance does not always indicate a shared decision-making process. A paternalistic process of decision-making, for example, can result in a shared decision. How do we know when mutual agreement on the decision to implement is reached? An explicit and verbal acknowledgement provides one potential indication and can be ascer- tained through observation. Sometimes, the agreement may be implicit, and signified by a behavioural intent by the patient to return for treatment or by the physician to book a future treatment appointment. A verbal agreement at one point in time may not hold over time. In Scenario 4 in the Appendix, it appears that an agreement has been reached at the end of the medical encounter but, subsequently, the patient changed her mind and did not return for treatment.

A treatment decision is made and both par t ies agree to the decision

Shared decision-making is usually depicted, either implicitly or explicitly as a type of decision-making process. But shared treatment decision-making can also refer to an outcome, i.e. a shared or agreed upon decision. The shared process relates to the roles that patients and physicians play and involves complementary role expectations and behaviour. Agreement between physician and patient about the treatment decision is one possible outcome of this process; others include no decision or disagreement as to the preferred treatment. In Scenario 3, for example, both the physician and the patient reveal their preferences about treatment but they cannot reach an agreement.

The test of a shared decision (as distinct from the decision-making process) is if both parties agree on the treatment option. This does not mean that both parties are necessarily convinced that this is the best possible treatment for this patient, but rather that both endorse it as the treatment to implement. The physician may feel, for example, that the patient would really be better off with another treatment but agrees to endorse the patient's choice as part of a negotiated agreement in which the patient's views count. Through mutual acceptance, both parties share responsibility for the final decision.

This is an important characteristic and helps to distinguish shared decision-making from other models of decision-making. In the extreme case of the paternalistic and informed models, decision- making and ultimate responsibility for the decision are clearly vested with the physician or the patient respectively, and whether the opposite party accepts the decision is not relevant. Mutual acceptance may or may not occur with other models of decision-

SOME MEASUREMENT |SSUES

If there is confusion about what shared treatment decision-making means conceptually, then there is bound to be difficulty in measuring it empirically because of disagreement over what one is looking for as defining characteristics. We have described our ideas about this above. Even so, major methodological issues remain. For example some of the defining characteristics of a shared decision-making model overlap with those of other models and some characteristics may be present in varying degrees. Both these conditions make it difficult to establish clear boundaries or thresholds of what is inside versus outside the shared treatment decision-making box. This is particularly true for the measurement of a patient's level of participation in treatment decision-making. For example, if a patient simply agrees to a treatment decision suggested by the physician, does this mean she participated in the treatment process? We do not think so, but the issue still remains, to what extent does the patient need to reveal and discuss treatment preferences in order for her interaction to be defined as one of shared decision-making? Also, a given medical encounter may start as one model of interaction, but evolve into something else as the encounter unfolds. This may make it difficult to precisely clas- sify any given medical encounter as falling within one model or another.

Empirically measuring if and how patients deliberate over treatment choices and the process they use to arrive at a decision is no easy task. Observation techniques are frequently used to measure physician-patient communication more generally, using different interaction analysis systems. Many of these have been recently summarized in an article by Ong et al. (1995). These systems are

9


designed to identify and quantify salient features of physician-patient communication. Observation methods alone, however, have some limitations when it comes to measuring decision-making processes that occur "in the patient's head" and may not be made explicit either through verbal or non- verbal behaviour. From what we observe, we would need to make attributions about the patient's imen- tions and internal decision-making processes. For example, in Scenario 1, is the patient's questior to the physician a form of information gathering so that she can better evaluate treatment options or is she trying to get the physician to make the decision for her? Does her ultimate agreement with the physician's recommendation reflect a thoughtful con- sideration of the alternatives, deference to physician authority, or reluctance to engage in the decision- making process? The same observation could y:eld several different stories or interpretations.

Similarly, structured questionnaires that have used a Likert type scale to measure patient preferences for or actual participation in treatment decision-making in previous studies also have limitations. They simplify and structure the measurement process to such a degree that little information is gleaned about the dynamics of shared decision- making or the interactional process involved. Moreover, as noted earlier, it is often not clear how to interpret different preferences.

We think that an important area for future research is a more in-depth exploration of why patients hold different treatment preferences, and what meanings patients ascribe to these preferences. Qualitative methods such as illness narratives or semi-structured interviews seem well suited to lhis task for several reasons. First, they allow for an in- depth exploration and, hopefully, increased understanding of a complex process: how patients think about decision-making in general and treatment decision-making in particular. Second, they allow more freedom for the patient to structure the discourse of the interview so that the information cap- tured reflects patients' perspectives rather than researchers' preconceived measurement categor:es. Third, with qualitative interviews it is also possible to pay much more attention to social, cultural and illness contexts which may influence patients' views about appropriate roles in treatment decision-making. Such interviews may also be used to generate information about what patients perceive are barriers to shared decision-making and the kinds of interventions they suggest (if any) that would help them to adopt a more active treatment decision- making role.

In making the above recommendation we do not mean to suggest that qualitative methods are 1:he only appropriate ones for studying issues related to shared decision-making. For many issues, a combination of qualitative and quantitative methods r~ay well be appropriate. The particular methods

adopted should be those most useful in generating the best information about the specific question of interest.

CONCLUSION

Because shared decision-making is increasingly advocated as an ideal treatment decision-making process, one might be tempted to try to identify explicit behaviours prescribing how to engage in this process. The ability to define specific behaviours holds appeal because it would provide a form of instructional guide, but we think it is also proble- matic. There is no single route to shared decision- making; some of the characteristics we describe can be met by a variety of different behaviours. Also, the physician-patient relationship is a personal, and often trusting relationship. To develop a standardized checklist of concrete and invariant steps for shared decision-making would not fit this decision- making context where the preferences of both physicians and patients can vary widely and change over time. While a checklist approach may resonate with many physicians in terms of a clinical framework of decision-making (Sackett et al., 1985), it may not resonate with patients' models of decision-making or constructions of their illness experience. Finally, shared decision-making is in some sense a matter of perception. Similar objective states of shared decision-making (if these could be defined) might well be both perceived and valued differently by different patients.

Perhaps more important than prescribing specific behaviours would be to agree on certain fundamen- tal principles around treatment decision-making for example, that patients' preferences for participation be elicited and acknowledged, that patients be given choices as to how the decision-making process will proceed, and that their choices be respected and adhered to in the physician's behaviour.

Herein lies one of the benefits of a shared decision-making approach. It offers a potential middle choice of treatment decision-making models between two polar extremes. On the one hand, the paternalistic model is characterized by physician dominance of the decision-making process. On the other hand, the informed decision-making model seems to limit the role of the physician to one of transferring information, enhancing the patient's ability to engage in autonomous decision-making with ultimate control but also responsibility for the treatment choice. Empirical research demonstrates that many patients, for whatever reasons, prefer not to assume full decision-making control. But many may also not like the idea of no say at all. Shared decision-making offers an intermediate alternative for both physician and patient. For the patient it offers some say without total responsibility, and for the physician, an opportunity to go beyond a role

10


of transferring information to also participate in, but not dominate, the decision-making process.

Acknowledgements We would like to thank the Health Polinomics Research Workshop and participants in the Continuing Education sessions organized by the Department of Clinical Epidemiology and Biostatistics at McMaster University for helpful comments on earlier drafts of this paper. We would also like to thank two anonymous reviewers for their very helpful and construc- tive comments. To Dawna Jaworsky who typed several drafts of this paper we express our appreciation. The usual disclaimer holds.

REFERENCES

Adelman, R. D., Greene, M. G. and Charon, R. (1987) The physician--elderly patient companion triad in the medical encounter: The development of a conceptual framework and research agenda. The Gerontologist 27, 729.

Beisecker, A. E. (19883 Aging and the desire for information and input in medical decisions: Patient consumerism in medical encounters. The Gerontologist 28, 330.

Beisecker, A. E. and Beisecker, T. D. (1990) Patient information-seeking behaviors when communicating with doctors. Medical Care 28, 19.

Biley, F. C. (1992) Some determinants that effect patient participation in decision-making about nursing care. Journal of Advanced Nursing 17, 414.

Blanchard, C. G., Labreque, M. S. and Ruckdeschel, J. C.et al. (1988) Information and decision making preferences of hospitalized adult cancer patients. Social Science & Medicine 27, 1139.

Brock, D. W. and Wartman, S. A. (1990) When competent patients make irrational choices. New England Journal ~[ Medicine 322, 1595.

Brody, D. S., Miller, S. M. and Lerman, C. E.et al. (19893 Patient perception of involvement in medical care: relationship to illness attitudes and outcomes. Journal q/ General Internal Medicine 4, 506.

Brody, D. S. (1980) The patient's role in clinical decision making. Annals o['Internal Medicine 93, 718.

Cassileth, B., Zupkis, R. V. and Sutton-Smith, K.et al. (1980) Information and participation preferences among cancer patients. Annals of Internal Medicine 92, 832.

Charles, C. and DeMaio, S. (1993) Lay participation in health care decision-making: a conceptual framework. Journal of'Health Politics. Policy and Law lg, 881.

Coe, R. M. and Prendergast, C. G. (1985) The formation of coalitions: Interaction strategies in triads, Sociology of Health and Illness 7, 236.

Deber, R. (1994) Physicians in health care management: 7. The patient-physician partnership: Changing roles and the desire for information~ Canadian Medical Association 151, 171.

Degner, L. F. and Sloan, J. A. (1992) Decision making during serious illness: What role do patients really want to Play?. Journal o/" Clinical Epidemiology 45, 941.

Eddy, D. M. (1990) Anatomy of a decision. Journal of the American Medical Association 263, 441.

Egbert, L. D., Battit, G. E. and Welch, C. E.et al. (19643 Reduction of postoperative pain by encouragement and instruction of patients. New England Journal of Medicine 270, 825.

Emanuel, E. J. and Emanuel, L. L. (19923 Four models of the physician-patient relationship. Journal of the American Medical Association 267, 2221.

Ende, J., Kazis, L. and Ash, A.et al. (1989) Measuring patients' desire for autonomy: decision making and in-

formation seeking preferences among medical patients. Journal of General Internal Medicine 4, 23.

Evans J. R. (1987) Toward a shared direction for health in Ontario. Report of the Ontario Health Review Panel. Government of Ontario, 0-7729-2701-4, Toronto.

Evans R. G. (1984) Strained Mercy: The Economics of Canadian Health Care. Butterworths, Toronto.

Freidson E. (1970) Professional Dominance. Atherton Press, New York.

Gray. R. E., Doan, B. D. and Church, K. (1990) Empowerment and persons with cancer: Politics in cancer medicine. Journal of Palliative Care 6, 33.

Greene B. (19923 Breast Cancer: Unanswered Questions. Report o/" the Standing Committee on Health and WelJare, Social Affairs, Seniors and the Status of Women.

Greenfield, S., Kaplan, S. and Ware, J. E. (1985) Expanding patient involvement in care: effects on patient outcomes. Annals of Internal Medicine 102, 520.

Greenfield, S., Kaplan, S. H. and Ware, J. E.et al. (1988) Patients' participation in medical care: effects on blood sugar control and quality of life in diabetes. Journal of General Internal Medicine 3, 448.

Hack, T. F.. Degner, L. F. and Dyck, D. G. (19943 Relationship between preferences for decisional control and illness information among women with breast cancer: A quantitative and qualitative analysis. Social Science & Medicine 39, 279.

Haug, M. R. and Lavin, B. (1981) Practitioner or patient: Whose in charge? Journal of Health and Social Behavior 22, 212.

Haug, M. and Lavin, B. (19833 Consumerism in Medicine. Sage, Beverly Hills.

Haug, M. R. and Ory, M. G. (1987) Issues in elderly patient-provider interactions. Research on Aging 9, 3.

Haug, M. R. (19943 Elderly patients, caregivers and physicians: Theory and research on health care triads. Journal of Health and Social Behavior 35, 1.

Hill Beuf, A. (1989) Biting o[:[' the bracelet." A study of chil- ~#en in hospitals. University of Pennsylvania Press, Pennsylvania.

Hurley, J., Birch, S. and Eyles, J. (1992) Information, effi- ciency and decentralization within health care systems. CHEPA Working Paper 92-21. McMaster University, Hamilton.

Lcrman, C. E., Brody, D. S. and Caputo, G. C.et al. (19903 Patients" perceived involvement in care scale: Relationship to attitudes about illness and medical care. Journal ~?/' General Internal Medicine 5, 29.

Levine, M. N., Gafni, A. and Markham, B.et al. (19923 A bedside decision instrument to elicit a patient's preference concerning adjuvant chemotherapy for breast cancer. Annals of Internal Medicine 117, 53.

Lidz, C. W., Meisel, A. and Munetz, M. (1985) Chronic disease: The sick role and informed consent. Culture, Medicine and Po'chiatry 9, 241.

Llewellyn-Thomas, H. A. (19923 Studying patients' preferences in health care decision making. Canadian Medical Association Journal 147, 859.

Llewellyn-Thomas, H. A. (19953 Patients" health-care decision making: A framework for descriptive and experimental investigations. Medical Decision Making 15, 101.

Mahler, H. 1. and Kulik, J. (1990) Preferences for health care involvement, perceived control and surgical recovery: A prospective study. Social Science & Medicine 31, 743.

Mooney, G. and Ryan, M. (1993) Agency in health care: Getting beyond first principles. Journal of Health Economics 12, 125.

Nayfield, S. G., Bongiovanni, G. C., Alciatti, M. H., Fischer, R. A. and Bergner, L. (1994) Statutory require-

11

Shared treatment decLsion-making: what does it mean?

ments for disclosure of breast cancer treatment alternatives. Journal of the National Cancer Institute 86, 1202.

Ong, L. M. L., de Haes, J. C. J. M., Hoos, A. M. and Lammes, F. B. (1995) Doctor-patient communica~:ion: A review of the literature. Social Science & Medicine 40, 903.

Ontario Ministry of Health (1994) Consent to treatment: A guide to the Act. Queen's Printer for Canada, Ontario.

Parsons, T. (1951) The Social System. Free Press, Glellcoe Illinois.

Pendleton, L. and House, W. C. (1984) Preferences for treatment approaches in medical care. Medical (?are 22(7), 644.

Pierce, P. F. (1993) Deciding on breast cancer treatment: A description of decision behavior. Nursing Researck 42, 22.

Quill, T. E. (1983) Partnership in patient care: A contractual approach. Annals oflnternal Medicine 98, 228.

Rosow, I. (1981) Coalitions in Geriatric Medicine. In Elderly Patients and Their Doctors, ed. M. H.aug. Springer, New York.

Roter, D. L. and Hall, J.A. (1992) Doctors talking with patients/Patients talking with doctors: Improving communication in medical visits. Auborn House, Westport, CT.

Ryan, M. (1992) The Economic Theory of Agency in Health Care: Lessons from Non-Economists for Economists. Health Economics Research Unit Discussion Paper, 03/92. University of Aberdeen, Scotland.

Sackett, D. L., Haynes, R. B. and Tugwell, P. (1985) Clinical Epidemiology: A Basic Science for Clitlical Medicine. Little, Brown and Co, Toronto.

Schulman, B. A. (1979) Active patient orientation and outcomes in hypertensive treatment: Application of a socio- organizational perspective. Medical Care 17, 267.

Silverstein, M. D., Stocking, C. B. and Antel, J. P.eJ al. (1991) Amyotrophic lateral sclerosis and life sustaining therapy: Patients' desires for information, participation in decision making and life sustaining therapy. Mayo Clinic Proc 66, 906.

Strull, W. M., Lo, B. and Charles, G. (1984) Do patients want to participate in medical decision making?. Journal of the American Medical Association 252, 2990.

Sutherland, H. J., Llewellyn-Thomas, H. A. and Lockwood, G. A.et al. (1989) Cancer patients: Their desire for information and participation in treatment decisions. Journal of the Royal Society of Medicine 82, 260.

Thomasma, D. C. (1983) Beyond medical paternalism and patient autonomy: A model of physician conscience for the physician patient relationship. Annals of Inte,nal Medicine 98, 243.

Veatch, R. M. (1972) Models for ethical medicine in a revolutionary age. Hastings Center Report 2, 5.

Vertinsky, I. B., Thompson, W. A. and Uyeno, D. (1974) Measuring consumer desire for participation in clinical decision making. Health Services Research, Summer, 121.

Waitzkin, H. (1984) Doctor-patient communication: Clinical implications of social scientific research. Journal of the American Medical Association 252, 2441.

Waitzkin, H. (1991) The Politics of Medical Encounters: How Patients and Doctors Deal with Social Problems. Yale University Press, New Haven.

Waterworth, S. and Luker, K. (1990) Reluctant collabor- ators: Do patients want to be involved in decisions concerning care?. Journal of Advanced Nursing 15, 971.

Wennberg, J. E. (1990) Outcomes research, cost contain- ment and the fear of health care rationing. New Engiand Journal of Medicine 323, 1202.

691

West, C. (1984) Routine Complications: Troubles with Talk Between Doctors and Patients. Indiana University Press, Bloomington.

Whelan, T. J., Levine, M. N., Gafni, A., Lukka, H., Mohide, E. A., Patel, M. and Streiner, D. L. (1995) Breast irradiation post lumpectomy: Development and evaluation of a decision instrument. Journal of Clinical Oncology 13, 847.

Williams, A. (1988) Priority setting in public and private health care, a guide through the ideological jungle. Journal of Health Economics 7, 173.

APPENDIX

A patient with newly-diagnosed early breast cancer comes to see her oncologist. The physician reaffirms the diagnosis and discusses with the patient her concerns. The physician outlines the various treatment options, the pros and cons of each, and asks the patient about her preferences for treatment, given her lifestyle and values. The patient responds that she feels knowledgeable about the disease and the pros and cons of various treatment options but can't decide; she would prefer that the physician make the decision. The physician responds that the patient is in the best position to judge because this decision involves placing a value on various treatment outcomes and weighing the benefits versus the risks of each option. The physician tells the patient to think it over and come back in another week with the decision. The patient comes back but still has not made a decision. Finally, the patient says: what would you do, if it was your wife? The physician says: if it was my wife, I would choose option A, but my preferences may well be different from yours. The patient replies: I will go with option A. (Scenario 1)

A patient with newly diagnosed early breast cancer comes to see her oncologist. The physician reaffirms the diagnosis and discusses with the patient her concerns. The physician outlines the various treatment options, the pros and cons of each, and asks the patient about her preference for treatment, given her lifestyle and values. The patient responds that she feels knowledgeable about the disease and the pros and cons of various treatment options. After discussing these issues at length, the patient says that she prefers option A. The physician reminds her: do you understand that with option A, as I mentioned earlier, there are a number of side effects, e.g. you will feel sick for several months. The patient responds that she under- stands and still prefers option A, for reasons which she re- iterates. The physician agrees and says: 1 think that is a good choice for you. (Scenario 2)

A patient with newly diagnosed early breast cancer comes to see her oncologist. The physician reaffirms the diagnosis and discusses with the patient her concerns. The physician outlines the various treatment options, the pros and cons of each and asks the patient about her preferences for treatment, given her lifestyle and values. The patient responds that she feels knowledgeable about the disease and the pros and cons of various treatment options. She indicates that she wants to try a new experimental treatment (A) that she has read about in a magazine. The physician responds that he does not think that this is an appropriate treatment option because its effectiveness has not been proven. The patient says that she still prefers treatment A. The physician again tries to dissuade the patient from this decision and says he cannot provide this treatment. The patient says then I will have to go else- where, and she leaves the physician's office. (Scenario 3)

A patient with newly diagnosed early breast cancer comes to see her oncologist. The physician reaffirms the diagnosis and discusses with the patient her concerns. The physician

12


outlines the various treatment options, the pros and cons of each, and asks the patient which one she prefers, given her lifestyle and values. The patient responds that she feels knowledgeable about the disease and the pros and cons of various treatment options but she wants the physician to decide. The physician responds that the patient is in the best position to judge because this decision involves pla-

cing a value on various treatment outcomes and weighing the benefits versus the risks of each. The patient says, in that case, I will go with treatment A. The physician sets up the follow-up treatment visit. The patient does not return for the treatment. The physician later discovers that the patient has instead gone to another physician. (Scenario 4)

13

Decision-making in the physician±patient encounter:revisiting the shared treatment decision-making model

Cathy Charlesa,b,c,d,*, Amiram Gafnia,b,d, Tim Whelana,d,e,f

aDepartment of Clinical Epidemiology and Biostatistics, McMaster University, 1200 Main Street West, Hamilton, Ont., Canada

L8N 3Z5bCentre for Health Economics and Policy Analysis, McMaster University, Room 3h5, Health Sciences Centre, 1200 Main Street

West, Hamilton, Ont., Canada L8N 3Z5cDepartment of Sociology, McMaster University, 1200 Main Street West, Hamilton, Ont., Canada L8N 3Z5

dSupportive Cancer Care Research Unit, McMaster University, 1200 Main Street West, Hamilton, Ont., CanadaeOCTRF, Hamilton Regional Cancer Centre, Hamilton, Ont., Canada

fDepartment of Medicine, McMaster University, 1200 Main Street West, Hamilton, Ont., Canada L8N 3Z5

Abstract

In this paper we revisit and add elements to our earlier conceptual framework on shared treatment decision-making within the context of di�erent decision-making approaches in the medical encounter (Charles, C., Gafni, A.,

Whelan, T., 1997. Shared decision-making in the medical encounter: what does it mean? (or, it takes at least two totango). Social Science & Medicine 44, 681±692.). This revised framework (1) explicitly identi®es di�erent analyticsteps in the treatment decision-making process; (2) provides a dynamic view of treatment decision-making by

recognizing that the approach adopted at the outset of a medical encounter may change as the interaction evolves;(3) identi®es decision-making approaches which lie between the three predominant models (paternalistic, shared andinformed) and (4) has practical applications for clinical practice, research and medical education. Rather thanadvocating a particular approach, we emphasize the importance of ¯exibility in the way that physicians structure the

decision-making process so that individual di�erences in patient preferences can be respected. # 1999 ElsevierScience Ltd. All rights reserved.

Keywords: Treatment decision-making; Doctor±patient relationship; Communication; Breast cancer

Introduction

Although shared treatment decision-making is a con-

cept that has gained widespread appeal to both phys-icians and patients in recent years, there is stillconfusion about what the concept means. To help clar-

ify this issue, we published a paper which tried to de-®ne shared treatment decision-making and its keycharacteristics and to show how this interactionalmodel di�ers from other commonly cited approaches

to treatment decision-making such as the paternalistic

Social Science & Medicine 49 (1999) 651±661

0277-9536/99/$ - see front matter # 1999 Elsevier Science Ltd. All rights reserved.

PII: S0277-9536(99 )00145-8

* Corresponding author. Tel.: +1-905-525-9140, ext. 22907;

fax: +1-905-546-5211.

E-mail address: [email protected] (C. Charles)

14

and the informed models1 (Charles et al., 1997a). Thepaternalistic model is by now well known and articu-

lated (Emanuel and Emanuel, 1992; Levine et al.,1992; Beisecker and Beisecker, 1993; Deber, 1994;Coulter, 1997). Hence, we concentrated on exploring

the di�erences between the informed and the sharedmodels because these two labels have often been usedinterchangeably to describe quite di�erent types of in-

teraction between physician and patient in treatmentdecision-making.The context for our discussion was a life threatening

disease where several treatment options were availablewith di�erent possible outcomes (bene®ts and risks orside e�ects), outcomes could vary in their impact onthe patient's physical and psychological well-being and

outcomes in the individual case were uncertain. In thiscontext, we argued that a shared treatment decision-making model could be identi®ed as such by reference

to four necessary characteristics (Charles et al., 1997a)as follows:

1. At a minimum, both the physician and patient areinvolved in the treatment decision-making process.

2. Both the physician and patient share information

with each other.3. Both the physician and the patient take steps to par-

ticipate in the decision-making process by expressingtreatment preferences.

4. A treatment decision is made and both the physician

and patient agree on the treatment to implement.

In this paper we revisit and add elements to our con-ceptual framework based on further analytic thinkingand our current research on the meaning of shared de-

cision-making to women with early stage breast cancerand to physicians who specialize in this area (Charleset al., 1998). Our revised framework (1) identi®es

di�erent analytic stages in the treatment decision-mak-ing process; (2) provides a dynamic view of treatmentdecision-making by recognizing that the approach

adopted at the outset of any given physician±patientencounter may change during the course of thatencounter; (3) identi®es di�erent approaches that lie inbetween the three predominant treatment decision-

making models and (4) has practical applications forclinical practice, research and medical education.Before exploring these issues, we brie¯y review factors

that have led to the development of new treatment de-cision-making models as alternatives to the traditionalpaternalistic approach.

The rise and fall of paternalism

Prior to the 1980s, the most prevalent approach totreatment decision-making in North America waspaternalistic with physicians assuming the dominant

role. Underlying this deference to professional auth-ority were a number of assumptions. First, that formost illnesses, a single best treatment existed and that

physicians generally would be well versed in the mostcurrent and valid clinical thinking. Second, physicianswould not only know the best treatments available,

they would consistently apply this information whenselecting treatments for their own patients. Third,because of their expertise and experience, physicianswere in the best position to evaluate tradeo�s between

di�erent treatments and to make the treatment de-cision. Fourth, because of their professional concernfor the welfare of their patients, physicians had a legiti-

mate investment in each treatment decision. This legiti-mation of physician control was further buttressed byprofessional codes of ethics which bound physicians to

act in the best interests of their patients (Lomas andContandriopoulous, 1994; Charles et al., 1997b). All ofthese assumptions led both physicians and patients to

expect a dominant role for physicians in treatment de-cision-making. Status di�erences between physiciansand patients in terms of education, income and genderalso contributed to power di�erentials in the medical

encounter.During the 1980s and beyond, the credibility of the

above assumptions began to be questioned. For an

increasing number of illnesses, for example, there wasno one best treatment and a more murky and complex

1 There is a fourth model of treatment decision-making de-

rived from health economics called the physician-as-agent for

the patient. This model stresses the informational asymmetry

between physician and patient. Physicians, in the usual case,

are seen as possessing more technical expertise about various

treatments and their bene®ts and risks than patients, while

patients are seen as possessing superior knowledge about how

these di�erent treatments resonate with their lifestyle, health

beliefs and desired health states. This model assumes that

both knowledge and preferences are needed to make the best

decision, yet a dilemma arises because each of these com-

ponents resides in a di�erent person. To resolve this dilemma,

the model speci®es that the patient communicate her prefer-

ences to the physician. The latter, who then becomes the sole

decision-maker, `àssumes responsibility for directing the

health care utilization of the patient, as an agent trying to

choose what the patient would have chosen, had she been as

well-informed as the professional.'' (Evans, 1984, p. 75). An

alternative way to bring values and preferences together in

one person is for the physician to communicate the technical

knowledge about treatment bene®ts to the patient. She will

then possess the relevant knowledge, preferences and values

to enable her to make the decision. This is called the informed

model of decision-making. In another paper (Gafni et al.,

1998) we have compared these two models and argued that

the informed model appears to be superior to the physician-

as-agent model in terms of feasibility of implementation.

C. Charles et al. / Social Science & Medicine 49 (1999) 651±661652

15

decisional context evolved where di�erent treatments

had di�erent types of tradeo�s between bene®ts and

risks. Since the patient rather than the physician would

have to live with the consequences of these tradeo�s,

the assumption that physicians were in the best pos-

ition to evaluate and weigh these was increasingly chal-

lenged (Eddy, 1990; Levine et al., 1992; Lomas and

Lavis, 1996). At the same time, research into the qual-

ity of medical care began to focus on the e�ectiveness

and appropriateness of a wide range of services deliv-

ered by physicians (Roos, 1984; Chassin et al., 1986,

1987b; Roos et al., 1988; Berwick, 1989; Lomas, 1990;

Wennberg, 1990). The research on small area vari-

ations, for example, found consistent evidence that

physician procedures for the same disease often varied

considerably across small geographic areas and that

these variations did not seem to be related to di�er-

ences in the health status of the respective populations

(Roos, 1984; Chassin et al., 1986, 1987a; Wennberg et

al., 1987; Roos et al., 1988; Leape et al., 1993; Iscoe et

al., 1994). Variations in treatment patterns were also

found for diseases for which clinical guidelines had

been developed on best practices (Lomas et al., 1989).

Patient preferences may have accounted for some of

this variation, but the data also suggested that either

some physicians were unaware of recommended best

practices for the treatment of particular diseases or

that they were aware, but were not implementing the

recommended guidelines.

Concern with rising health care costs in both

Canada and the United States was another health pol-

icy issue focusing attention on the medical profession

(Katz et al., 1998). The joining together of cost and

quality concerns resulted in recommendations to make

physicians more explicitly accountable to patients, the

public, and in the case of the United States, to third

party payers. In addition, the twin principles of caveat

emptor (let the buyer beware) and consumer sover-

eignty gained popularity (Haug and Lavin, 1981, 1983;

Charles and DeMaio, 1993), as manifested in new

legislation precluding treatment implementation with-

out informed consent and in legislation safeguarding

the rights of patients to be informed about all available

treatment options (Nay®eld et al., 1994; Ontario

Ministry of Health, 1994). These two principles were

also evident in the emergent interest among both

patients and physicians in developing and advocating

new approaches to treatment decision-making which

would incorporate a larger role for patients in the de-

cision-making process (Brody, 1980; Quill, 1983;

Thomasma, 1983; Eddy, 1990; Hughes and Larson,

1991; Emanuel and Emanuel, 1992; Ryan, 1992;

Deber, 1994; Llewelyn-Thomas, 1995; Cahill, 1996;

Quill and Brody, 1996; Charles et al., 1997a; Coulter,

1997; Gafni et al., 1998).Table

1

Modelsoftreatm

entdecisionmakinga

Models

Analyticalstages

Paternalistic

(inbetweenapproaches)

Shared

(inbetweenapproaches)

Inform

ed

Inform

ationexchange

Flow

Oneway(largely)

Twoway

Oneway(largely)

Direction

Physician4

patient

PhysicianF

patient

Physician4

patient

Type

Medical

Medicalandpersonal

Medical

Amountb

Minim

um

legallyrequired

Allrelevant

fordecision-m

aking

Allrelevant

fordecision-m

aking

Deliberation

Physicianalone

orwithother

physicians

Physicianand

patient(pluspotentialothers)

Patient(plus

potentialothers)

Decidingon

treatm

entto

implement

Physicians

Physicianandpatient

Patient

aIllustrationforanencounterfocusingonthecase

ofa(treating)physician±patientdyad.Formore

complexcasesseetext.

bMinim

um

required.

C. Charles et al. / Social Science & Medicine 49 (1999) 651±661 653

16

Models of treatment decision-making

Both the informed and the shared models of treat-ment decision-making were developed largely in reac-tion to the paternalistic model and to compensate for

alleged ¯aws in the latter approach. These threemodels are the most prominent and widely discussed inthe treatment decision-making literature. Key charac-

teristics of each model and how they di�er from oneanother are summarized in Table 1. In Table 1 treat-ment decision-making is subdivided into three analyti-

cally distinct stages, even though, in reality, these mayoccur together or in an iterative process. The steps are:information exchange, deliberation about treatmentoptions and deciding on the treatment to implement.

The latter is the outcome of the deliberation process.

Information exchange

Information exchange refers to the type and amountof information exchanged between physician and

patient and whether information ¯ow is one or twoway. Types of information that the physician mightcommunicate to the patient include: the natural history

of the disease, the bene®ts and risks (side e�ects) ofvarious treatment alternatives, a description of thetreatment procedure(s) to be used and communityresources and information that the patient could access

about her disease. These are primarily technical typesof knowledge which most patients will not have.Information that the patient might reveal to the phys-

ician include: aspects of the patient's health history,her lifestyle, her social context (e.g. work and familyresponsibilities and relationships), her beliefs and fears

about her disease and her knowledge of various treat-ment options obtained from lay networks and/or otherinformation sources. Except for the latter, these areprimarily types of self-knowledge that the patient

brings to the encounter and that the physician typicallyhas no way of knowing except through direct com-munication with the patient in this or in prior consul-

tations. In addition, at the outset of the encounter,either the physician, the patient, or both may exchangepreferences regarding their own and each other's role

in the decision-making process. The goal of thisexchange is to make explicit how each expects the de-cision-making process to proceed.

The ¯ow of information exchange may be one wayor two way. In the paternalistic model, the exchange islargely one way and the direction is from physician topatient. At a minimum, the physician must provide the

patient with legally required information on treatmentoptions and obtain informed consent to the treatmentrecommended. Beyond this, the patient as depicted in

this modal is a passive recipient of whatever amountand type of information the physician chooses to

reveal. In some cases, the physician may ask thepatient about speci®c issues such as pain tolerance or

allergies that could a�ect the latter's reaction to thetreatment selected by the physician. In general, thismodel assumes that the physician knows best and will

make the best treatment decision for the patient. In ad-dition, information exchange from patient to physicianis not seen as a major prerequisite to completing this

task.In a shared decision-making model, the information

exchange is two way (Charles et al., 1997a). At a mini-

mum, the physician must inform the patient of all in-formation that is relevant to making the decision, i.e.information about available treatment options, thebene®ts and risks of each and potential e�ects on the

patient's psychological and social well being. Thepatient needs to provide information to the physicianon issues raised above, e.g. her values, preferences, life-

style, beliefs and knowledge about her illness and itstreatment. The ®rst type of information exchangeensures that all relevant treatment options are on the

table; the second ensures that both the physician andpatient evaluate these within the context of thepatient's speci®c situation and needs rather than as a

standard menu of options whose impact and outcomesare assumed to be similar for clinically similar patients.In the informed model, information exchange is one

way, from physician to patient. This exchange is the

very crux of the model, de®ning the boundaries of thephysician's clinical role in decision-making. The phys-ician in this model is assumed to be the primary source

of information to the patient on medical/scienti®cissues about the patient's disease and treatmentoptions. To ful®ll this role, the physician, at a mini-

mum, needs to give the patient all relevant informationfrom the highest quality research evidence on the ben-e®ts and risks of various treatments so that she will beenabled to make an informed decision. Beyond infor-

mation transfer, the physician has no further role inthe decision-making process. The remaining tasks ofdeliberation and decision-making are the patient's

alone. Eddy (1990, p. 442) describes the rationale forrestricting physician involvement in these latter twosteps as follows:

. . . the people whose preferences count are thepatients, because they are the ones who will have tolive (or die) with the outcomes. . . .Ideally, you and

I are not even in the picture. What matters is whatMrs. Smith thinks . . . . It is also quite possible thatMrs. Smith's preferences will di�er from Mrs.

Brown's preferences. If so, both are correct, because`correct' is de®ned separately for each woman.Assuming that both women are accurately informed

regarding the outcomes, neither should be per-suaded to change her mind.


17

Not only the direction of information exchange butalso the amount of information exchanged can vary

across decision-making models. So far, we havefocused on minimum amounts for each model buthave not speci®ed outer boundaries. The amount of in-

formation that the physician could convey to thepatient, for example, is, theoretically, in®nite. Thephysician could provide detailed information on issues

like the biology of the disease or detailed aspects ofthe molecular basis for the disease. However, in prac-tice, the amount of information exchanged will be

in¯uenced by time and money constraints, both ofwhich raise issues of equity and costs. Time spent byphysicians with a given patient, for example, depletesthe time available to them for other needy patients in

their practice. This issue seems particularly salient to ashared decision-making approach. Because the infor-mation exchange is two way rather than one way, as

are processes of deliberation and decision-making, thisapproach is likely to take more time than either thepaternalistic or informed approaches, each of which

requires less interaction and consensus building.Given the importance of information transfer from

physician to patient in both the shared and informed

models, it is not surprising that various sorts of de-cision-aids have been and are being developed to helpphysicians communicate treatment information topatients and to present the information in a standar-

dized way (Llewellyn-Thomas, 1995). These aids rangefrom high technology interactive videos (Barry et al.,1995; Deber, 1996; Flood et al., 1996; Liao et al.,

1996) to low technology ¯ip charts with audio tapes.(O'Connor et al., 1994). Decision boards are anotherform of communication aid that lie between the high

and low technology options (Levine et al., 1992;Sebban et al., 1995; Whelan et al., 1995; Elit et al.,1996). Each of these types of decision aids has beendeveloped, for the most part, in isolation from one

another and by di�erent research teams. Althougheach type has its own supporters and advocates, to ourknowledge, few, if any, empirical studies have been

undertaken that compare the di�erent decision aids interms of criteria such as e�ectiveness, e�ciency andpatient satisfaction (Entwistle et al., 1998).

In addition, one important role for decision-aids hasreceived little attention. This is their potential for re-lationship building between the physician and patient.

Exchanging information is a process which enables thephysician and patient to get to know each other and todetermine how well they can work together. This issueis particularly important for the patient. Through the

process of information exchange, the patient has theopportunity to assess the extent to which the physi-cian's practice style, attitudes and behaviour will

match her own expectations of and preferences forhow she wants the physician to interact with her.

Building trust is one part of this process but does notcapture all the important ingredients of the relation-

ship.In our current study of women with early stage

breast cancer, for example, we found that women

stressed the importance not only of ®nding a physicianthey could trust, but also one who would treat them asindividuals. Patients' assessments of these physician

attributes rested, in large part, on their perceptions ofthe physician's ability and willingness to contextualizethe decision-making process by framing the discussion

in terms of each patient's unique background, charac-teristics and life experience (Charles et al., 1999).Decision aids used within the context of the phys-

ician±patient relationship provide another piece of in-

formation for patients to use in assessing their level ofcompatibility and comfort with a given physician'spractice style. Some decision aids, however, can be

self-administered and used outside the context of thephysician/patient interaction. What e�ect, if any, infor-mation tools that are self-administered might have on

a patient's relationship with her physician is an issuerequiring further study.Decision aids which present scienti®c information to

patients about treatment bene®ts and risks are devel-oped to create more informed patients and to encou-rage èvidence-based decision-making'. This approachassumes that if only physicians knew how to transfer

scienti®c information to patients in an accurate andunbiased way, the latter could be ®lled up (like anempty glass) with new knowledge and thereby trans-

formed into informed and willing decision-makers.However, patients are not empty vessels. They come tothe medical encounter with their own beliefs, values,

fears, illness experiences and, increasingly, informationabout various treatment options. Moreover, patientsare not so much interested in average outcomes foraggregate groups of patients as they are in knowing

what this information means for themselves speci®-cally. Patients interpret information on average treat-ment outcomes in order to make them personally

meaningful within the decision-making context theyface ( AdelsaÈ rd and Sachs, 1996; Turney, 1996; Charleset al., 1998). In so doing, their own values and beliefs

act as ®lters in processing what information is allowedin and how it is understood (Williams and Calnan,1996). In this interpretive process, the intended mess-

age to the patient may be lost, altered or transformed(Parsons and Atkinson, 1992; Whelan et al., 1995;Charles et al., 1998). Research into decision aids andother communication mechanisms that focus only on

de®ning the speci®c message to be conveyed and themost appropriate means of doing so, fail to considerpatient factors that might also a�ect how information

is processed and understood. This latter type of datawould be useful to clinicians who want to be part of


18

the deliberation process, to better understand theirpatients and to recognize potential di�erences between

lay and medical world views (Mishler, 1984).

Deliberation

The deliberation stage of decision-making refers tothe process of expressing and discussing treatment pre-

ferences. The minimum requirement for who isinvolved in this process varies across decision-makingmodels. In the paternalistic approach, the treating

physician weighs the bene®ts and risks of each optionalone or in consultation with other physicians. Thetreating physician dominates the deliberation processwhile the patient passively listens. Physician dominance

is justi®ed by clinical judgement and experience. Thelabel paternalistic is an apt term for this model becauseit evokes the image of a parent±child relationship

where the authority ®gure (physician) has the right todecide what is best for the child (patient), even if thechild disagrees (Parsons, 1951).

The treating physician may verbally communicate tothe patient only the ultimate treatment decision, failingto reveal knowledge and values considered in the selec-

tion process and how these were weighted. Decision-making in this context can be completed fairly quicklyif the physician feels well informed to make the de-cision and unrestrained by the need to have patient

input into this process. Of course, the lack of patientinput is precisely the reason why this model is viewedby many as undesirable.

The de®ning characteristic of deliberation in theshared decision-making model is its interactionalnature (Charles et al., 1997a). This is both its major

strength and weakness. The emphasis on interactionensures patient input into the process; but it alsomakes the process more cumbersome and time con-suming. Both physician and patient are assumed to

have a legitimate investment in the treatment decision,the patient because her health is at stake and the phys-ician out of concern for the patient's welfare.

For a shared model to work, both physicians andpatients have to perceive that there are treatmentchoices. Otherwise, there is nothing to decide (Charles

et al., 1998). Patients typically face one of two alterna-tive treatment decisional contexts. The ®rst is a choicebetween two di�erent treatments; the second is a

choice between doing nothing (e.g. watchful waiting)and doing something (e.g. implementing a speci®cintervention such as radiation). In an earlier study, wefound that women with early stage breast cancer

attending a regional cancer centre for consultation re:adjuvant therapy did not perceive the latter situationas one of choice. Many women felt they had no choice

but to accept the treatment o�ered so that they couldreassure themselves that they had done everything

possible to ®ght the disease and to alleviate the possi-bility of post-decision regret should the disease return.

As one woman said: ``Doing nothing is no choice''(Charles et al., 1998).In a shared approach, each person needs to be will-

ing to engage in the decision-making process byexpressing treatment preferences, in addition to what-ever information they exchange. Some have argued

that if information is exchanged, this is su�cient toview the interaction as shared. We view information asonly the ®rst step in the overall treatment decision-

making process. It is the basic building block to enablea shared process to occur but it does not, in and ofitself, constitute that process.In a shared model, the interaction process to be

used to reach an agreement may be explicitly discussedat the outset of the encounter or may evolve implicitlyas the interaction unfolds. The process is likely to be

consensual if both parties start out fairly close togetherin their thinking about the preferred treatment. If theyare wider apart in their views, a process of negotiation

is likely to occur. Negotiating as equal partners, how-ever, is not easy for the patient because of the inherentinformation and power imbalance in the relationship.

Physicians, in the usual case, will have superior knowl-edge of the technical issues involved in treatment de-cision-making and perhaps years of clinical experiencewith similar types of patients. The physician bears the

o�cially legitimized title of èxpert' while the patientmay feel particularly vulnerable and frightened duringthe medical encounter. When education, income, cul-

ture and/or gender di�erences also exist between thephysician and patient, the patient may feel too inti-mated to freely and openly express her preferences, let

alone negotiate for them with the physician. Creating asafe environment for the patient so that she feels com-fortable in exploring information and expressingopinions is probably the highest challenge for phys-

icians who want to practise a shared approach(Guadagnoli and Ward, 1998). At the other end of thepatient spectrum are those who are well informed

about their illness and various treatment options andwho have no di�culty expressing preferences. Some ofthese patients may have already made the treatment

decision before entering the physician's o�ce. If thepatient's preference is di�erent from the physician'sand the physician is not able to change the patient's

view, then the process will become con¯ictual.In a shared model, both physicians and patients are

assumed to have an investment in the treatment de-cision. The physician can legitimately give a treatment

recommendation to patients and try to persuade themto accept the recommendation. However, physicianswould also have to concentrate on listening to and

understanding why patients might favour a di�erenttreatment option. Perhaps the decision will be resolved


19

through further discussion and clari®cation of values,preferences and information, but perhaps not. In the

latter case, physicians would have to decide whetherthey could endorse patients' preferences as part of anegotiated agreement in which patients' views count,

or whether the strength of their own views of the bestpossible treatment for each patient would precludeendorsement of any other option. If a physician can-

not, in good conscience, endorse the patient's prefer-ence, then there is no agreement on the decision toimplement even though the deliberation process was

shared. In this case, the patient would have to go else-where and begin the process over again with anotherphysician if she hopes to have her preferred treatmentimplemented (unless, of course, her preferred option

was to do nothing). This example illustrates thatpatients face constraints in that their preferences forspeci®c treatments can only be implemented if a phys-

ician agrees to do so. On the other hand, physiciansalso face constraints. A patient turned down by onephysician, can make the same treatment request to

another physician. A refusal from the ®rst physiciandoes not preclude her from receiving the desired ser-vice from the second.

Each of the above examples assumes that only twoparties are involved in the decision-making process.This is the most simple case but probably not theusual case. The patient may decide to share any or all

of the decision-making steps with persons other thanor in addition to the physician. For example, somewomen with early stage breast cancer in our study of

shared treatment decision-making shared the infor-mation exchange component of the process with theironcologist but consulted with family, friends or their

family physician in selecting the most appropriatetreatment for themselves. These latter individuals knewthe patient personally and, were sought out during thedeliberation and decision-making stages precisely for

this reason. Including others in the decision-makingprocess introduces additional complexity since itexpands both the nature and the number of decisions

to be made as well as increasing the need for co-ordi-nation so that consultations with all persons involvedcan occur. In addition, some decisions require third

party agreement as a necessary pre-condition for im-plementation. For example, a physician and patientmay decide that the latter would do better being cared

for at home, even though a high level of constant andclose supervision is required. If there is no care-giverwilling to step in and either organize or undertake thistask, the decision cannot be implemented.

These examples illustrate that, in many instances, agiven physician±patient interaction is only one slice ofa larger decision-making process that involves others

in key roles and that takes place outside the context ofthe medical encounter. To the extent that our concep-

tualizations of treatment decision-making fail to incor-porate these others or to recognize their in¯uence, they

fail to capture important slices of the reality of thisprocess. Concepts, particularly sensitizing concepts,serve as analytic guides, de®ning, at least initially, the

boundaries of what to look for empirically to betterunderstand a phenomenon or process of interest(Blumer, 1969; Charmaz, 1990; van den Hoonaard,

1997). If conceptualizations of treatment decision-mak-ing fail to incorporate a potential role for signi®cantothers outside the physician±patient dyad, empirical

research will focus solely on this micro-system, exclud-ing important external in¯uences. Focusing on thephysician±patient dyad may yield a lot of informationabout this particular slice of reality, but relatively little

about the importance of this slice in the overall treat-ment decision-making process.In the informed model, as noted earlier, the patient

proceeds through the deliberation and decision-makingprocess on her own. The physician's role is limited toproviding medical/scienti®c information that will

enable her to make an informed decision. Underlyingthis model are two assumptions. The ®rst is that aslong as patients possess current scienti®c information

on treatment bene®ts and risks, they will be able tomake the best decision for themselves. The second isthat physicians should not have an investment in thedecision-making process or in the decision made

(Eddy, 1990). To do so, would go beyond the bound-aries of an appropriate clinical role because the phys-ician might harm the patient by inadvertently steering

her in a certain direction which re¯ects the physician'sown bias. Underlying this concern is the assumptionthat the interests and motivations of the physician and

patient may not be the same. This consumer orientedmodel emphasizes patient sovereignty and patients'rights to make independent, autonomous choices (Quilland Brody, 1996).

The view that physicians have no legitimate role toplay in the discussion or recommendation of treat-ments may be di�cult for many physicians to accept

since it runs counter to decades of professional medicaltraining and practice in which clinical experienceexpertise and knowledge, have been assumed to be the

quintessential skills that physicians have to o�er. Theinformed model may meet patients' needs for auton-omy in decision-making (for those who value this goal)

but it may not meet the needs of physicians who wantto participate in treatment decision-making and whoconsider this a key part of their clinical role. Theidenti®cation of needs raises the issue of whose needs

and goals should be served in the medical encounter:the physician's, the patient's or both? Few (hopefully)would argue for the physician's needs alone. However,

there may well be disagreement over the latter twooptions.


20

The use of evidence-based clinical guidelines in treat-ment decision-making provides a useful context within

which to consider this issue. Increasingly, clinicalguidelines are being developed, based on the highestquality research evidence available, to inform treat-

ment decision-making. Underlying the evidence-basedapproach is an assumption that whatever treatment isshown by the evidence to be the most e�ective is the

best treatment and the `rational' choice to implement.Some physicians go further to argue that if aninformed patient with an expressed desire to `get well'

chooses a di�erent treatment, this choice must be theresult of ìrrational' thinking and it is the physician'sduty to try to change the patient's mind (Brock andWartman, 1990). In such situations, evidence may be

used by the physician to prescribe the `right' treatment.Consumer sovereignty takes second place to the physi-cian's own belief system about what ought to deter-

mine the treatment decision. This role for evidence isnot compatible with either a shared or an informedmodel of treatment decision-making. A role that is

compatible lies in using scienti®c information to helpcreate more informed patients and to enhance patientchoice.

When physicians and patients have di�erent ideasabout which decision-making model should be used tostructure the decision-making interaction, they areheaded for con¯ict. Decision-making using any of the

above models will be prone to setbacks if the physicianand patient are not in step with each other. In our ear-lier paper (Charles et al., 1997a), we made the analogy

that shared decision-making takes `two to tango'. Forthe two parties to dance together, the physician needsto know what kind of dance the patient prefers and

the steps that this involves. Otherwise, the dance willbe punctuated with false starts and missteps, creatingtension between the partners and impeding their abilityto work together. This analogy can be extended. For a

certain type of dance, it seems appropriate that thephysician take the lead (for example, in transferringtechnical information to patients). However, when the

music changes to another type of dance, the patientmay well take the lead, being more of an expert in thenew steps required to ®t with this particular beat (for

example, patient preferences for di�erent health states).In a shared treatment decision-making model both thephysician and patient can take turns `leading' speci®c

discussions depending on which person has moreexpertise and experience to contribute on a given issue.

Decision on the treatment to implement

The ®nal task in the decision-making process ischoosing a treatment to implement. In the paternalistic

and informed models, the decision-maker is one per-son; in the ®rst case, the physician and in the second,

the patient. However, in both cases the decision-makeris not totally autonomous because each faces con-

straints in actually implementing the decision. Thephysician must have the patient's informed consentbefore proceeding and the patient needs the physician's

agreement to implement her preferred treatment(unless no treatment or an alternative therapy is pre-ferred). In the shared model, both parties, through the

deliberation process, work towards reaching an agree-ment and both parties have an investment in the ulti-mate decision made. The extent to which patient

involvement in decision making is associated with agreater commitment to the agreed upon treatment isan important area for study.

Practical applications of the framework

Our revised and updated framework depicted inTable 1 is more ¯exible and incorporates a moredynamic perspective on treatment decision-making

than our earlier model. We think it is also clearer interms of practical applications for physicians andothers. First, the framework provides a description of

the various analytical stages in the decision-makingprocess. The framework can be used to educate phys-icians about these stages and about the de®ningcharacteristics of each model. The framework also

describes the general path each model follows andmore speci®cally, behavoural expectations of bothphysicians and patients for implementing each model.

Second, physicians can use the framework to helpexplain to patients the di�erent approaches that can beused to make treatment decisions. They can also use

the framework as a tool in assessing patient prefer-ences (as well as their own) in this regard.Third, the framework makes explicit the possibility

that not only can the decision-making approach used

in one physician±patient interaction change in the nextinteraction, it can also change within a single inter-action. For example, a physician who starts the consul-

tation with an informed approach may need to switchmid-stream to a more shared approach if it becomesevident that the patient does not want to make the de-

cision on her own. In this case, what started as aninformed approach changes to one in which the phys-ician plays a more active role in making the decision.

Alternatively, even though a physician may be morecomfortable with a paternalistic approach, a givenpatient may want more of a role in decision-makingthan the former expects or is used to. To respond to

this patient's preferences, the physician will need tomove towards a shared model. The framework canhelp physicians identify the changes required of them

to move from a particular decision-making approachadopted at the outset of a medical encounter to a


21

di�erent one that better meets the needs of a speci®cpatient in the later stages of decision-making. Said

another way, a single treatment decision-making pro-cess can combine elements from di�erent models atdi�erent stages in the overall process.

Fourth, the framework makes explicit that switchingmodels during the encounter is easier in some circum-stances than in others. In fact, some movements across

models are not possible within a single interaction unlessthe whole process is started anew. For example, a sharedmodel, where both the physician and patient decide on

the treatment to implement requires as a preconditionthat two way information exchange and deliberationhave already taken place. If the physician starts the de-cision-making process with a paternalistic approach, the

information gap would need to be ®lled before a switchto a more shared approach could be made.Fifth, the framework recognizes that, in reality,

there are multiple approaches that lie between thethree ideal types. Starting with the paternalistic model,for example, the more that each step moves from a

physician dominated encounter to one where thepatient's input is recognized, nourished and valued, themore the model evolves into a shared approach.

Similarly, as the physician's role fades into the back-ground in steps 2 and 3 of the decision-making pro-cess, the approach moves towards the informed model.It would be di�cult to pinpoint the exact point

where one model ends and another begins. We thinkof the three prominent approaches discussed above asmarkers or anchor points that re¯ect the most well

known and best described models, recognizing thatthere are many combinations in between. While eachcould be labeled, this would be a time consuming and

di�cult task. In addition, we doubt that this would bea useful exercise since labels seem to generate norma-tive judgements about what are `good' and `bad' waysof decision-making rather than focusing on the speci®c

situational context in which one approach would bemore appropriate than another. The identi®cation inthe framework, not only of general approaches to

treatment decision-making but also of combinations inbetween increases options for physicians. It also re-inforces the importance of ¯exibility so that physicians

are able to recognize and respond to changes in patientpreferences for the nature of the interaction as the de-cision-making process unfolds.

Sixth, identifying the interactional dynamics requiredof each model highlights the added `costs' of engagingin shared decision-making relative to the other models.While advocates of shared decision-making have

stressed the bene®ts of this approach in promotingpatient participation and patient-centred care, fewhave focused on its ®nancial costs. If shared decision-

making turns out to require more time on averagethan other approaches in order to facilitate interaction

and to build consensus, then physicians may respondby either advocating for increased fees or seeing fewer

patients. However, there may also be costs for notinvolving patients, in the form of repeat visits, secondopinions and doctor shopping. The potential system-

wide policy impacts of a shared decision-makingapproach have not yet received much systematicresearch or public policy attention.

Seventh, the framework can assist in the evaluationof di�erent components of the overall treatment de-cision-making process. For example, if the goal of

treatment decision aids is limited to improving infor-mation transfer then the framework can be used toidentify the speci®c decision-making stage which formsthe relevant context for the evaluation. However, if the

goal of decision aids goes beyond this to also incorpor-ate relationship building within a shared decision-mak-ing approach, then the relevant evaluation context will

have to be broadened. As another example, the identi-®cation of explicit and analytically distinct phases ofthe decision-making process makes it possible to target

speci®c questions regarding patient satisfaction to eachof these stages and to determine which contributesmost to a patient's overall satisfaction level.

Eighth, the framework can be used as an edu-cational tool in medical training to stimulate discussionabout professional orientations to and roles in treat-ment decision-making. For example,

1. Whose views of the meaning of shared decision-making should count: those of academics and

researchers who advocate speci®c normative modelsin professional journals, or those of physicians andpatients in non-academic settings whose views might

be quite di�erent?2. Should physicians try to in¯uence the treatment de-

cision-making process and outcome or should theclinical role be limited to transferring relevant treat-

ment information to the patient?3. If a patient and her physician both prefer a paterna-

listic approach, is there anything `wrong' with mak-

ing the decision in this manner?4. If a physician prefers that a patient make the treat-

ment decision (the informed model), should the

patient be `forced' to do so, even if she does notwant to?

5. Given the power, status and informational asymme-

try between physician and patient, is it realistic toexpect patients, even informed patients, to be ableto hold their own in negotiations with physiciansabout treatment preferences? What steps can phys-

icians take to create a safe environment where thiscan occur?

6. What is the most e�ective strategy available to

physicians to elicit patient preferences for involve-ment in treatment decision-making?


22

Conclusion

In this paper we have revisited and expanded our

earlier conceptual framework of di�erent treatment de-cision-making models. We think this framework ismore ¯exible than its predecessor and recognizes more

clearly the dynamic nature of treatment decision-mak-ing. Practical applications of the framework have alsobeen discussed. Over the course of our research we

have learned that treatment decision-making is a com-plex process that takes place over time and can involvemany individuals rather than an event that takes placeat a ®xed point in time and is restricted to the phys-

ician±patient dyad. Our thinking on treatment de-cision-making will continue to evolve as we move in aniterative process, empirically studying di�erent aspects

of this process and using the information to clarify ourconceptual thinking.

Acknowledgements

This research is supported by a grant from theNational Cancer Institute of Canada, Canadian Breast

Cancer Research Initiative. We would like to thanktwo anonymous reviewers for their thoughtful com-ments on an earlier draft of this paper. We are gratefulto Reviewer I, in particular, for pointing out that the

analogy we use of `taking two to tango' can beextended to incorporate the notion that the lead part-ner can change depending on the type of dance being

undertaken.

References

AdelswaÈ rd, V., Sachs, L., 1996. The meaning of 6.8: numeracy

and normality in health information talks. Social Science

& Medicine 43, 1179±1187.

Barry, M.J., Fowler, F.J., Mulley, A.G., et al., 1995. Patient

reactions to a program designed to facilitate patient par-

ticipation in treatment decisions for benign prostatic

hyperplasia. Medical Care 33, 771±782.

Beisecker, A., Beisecker, T., 1993. Using metaphors to charac-

terize doctor±patient relationships: paternalism versus con-

sumerism. Health Communication 5, 41±58.

Berwick, D.M., 1989. Health services research and quality of

care: assignments for the 1990s. Medical Care 27, 763±771.

Blumer, H., 1969. Symbolic Interactionism: Perspective and

Method. Prentice Hall, Englewood Cli�s, NJ.

Brock, D.W., Wartman, S.A., 1990. When competent patients

make irrational choices. New England Journal of Medicine

322, 1595±1599.

Brody, D.S., 1980. The patient's role in clinical decision-mak-

ing. Annals of Internal Medicine 93, 718±722.

Cahill, J., 1996. Patient participation: a concept analysis.

Journal of Advanced Nursing 24, 561±571.

Charles, C., DeMaio, S., 1993. Lay participation in health

care decision-making: a conceptual framework. Journal of

Health Politics, Policy and Law 18, 881±904.

Charles, C., Whelan, T., Gafni, A., Lindsay, E., 1999. Shared

treatment decision-making among physicians and women

with early stage breast cancer in the medical encounter:

what does it mean? Unpublished data. Grant funded by

the National Cancer Institute of Canada, (CBCRI)

Toronto, Ont., Canada.

Charles, C., Gafni, A., Whelan, T., 1997a. Shared decision-

making in the medical encounter: what does it mean? (Or,

it takes at least two to tango. Social Science & Medicine

44, 681±692.

Charles, C., Lomas, J., Giacomini, M., et al., 1997b. The role

of medical necessity in Canadian health policy. Four

meanings . . . and a funeral? The Milbank Quarterly 75,

365±394.

Charles, C., Redko, C., Whelan, T., Gafni, A., Reyno, L.,

1998. Doing nothing is no choice: lay constructions of

treatment decision-making among women with early-stage

breast cancer. Sociology of Health and Illness 20, 71±95.

Charmaz, 1990. `Discovering' chronic illness: using grounded

theory. Social Science & Medicine 30, 1161±1172.

Chassin, M.R., Brook, R.H., Park, R.E., 1986. Variation in

the use of medical and surgical services by the medicare

population. New England Journal of Medicine 314, 285±

290.

Chassin, M.R., Koseco�, J., Park, R.E., et al., 1987a. Does

inappropriate use explain geographic variations in the use

of health care services? Journal of the American Medical

Association 258, 2533±2537.

Chassin, M.R., Koseco�, J., Solomon, D.H., et al., 1987b.

How coronary angiography is used. Clinical determinants

of appropriateness. Journal of the American Medical


Coulter, A., 1997. Partnerships with patients: the pros and

cons of shared clinical decision-making. Journal of Health

Services Research and Policy 2, 112±121.

Deber, R., 1994. Physicians in health care management. 7.

The patient±physician partnership: changing roles and the

desire for information. Canadian Medical Association

Journal 151, 171±177.

Deber, R., 1996. Shared decision-making in the real world.

Journal of General Internal Medicine 11, 377±378.

Eddy, D.M., 1990. Anatomy of a decision. Journal of the

American Medical Association 263, 441±443.

Elit, L.M., Levine, M.N., Gafni, A., et al., 1996. Patients' pre-

ferences for therapy in advanced epithelial ovarian cancer:

development testing and application of a bedside decision

instrument. Gynecologic Oncology Journal 62, 329±335.

Emanuel, E.J., Emanuel, L.L., 1992. Four models of the phys-

ician±patient relationship. Journal of the American

Medical Association 267, 2221±2226.

Entwistle, V.A., Sowden, A.J., Watt, I.S., 1998. Evaluating

interventions to promote patient involvement in decision-

making by what criteria should e�ectiveness be judged?

Journal of Health Services Research Policy 3, 100±107.

Evans, R.G., 1984. Strained Mercy: the Economics of

Canadian Health Care. Butterworths, Toronto.

Flood, A.B., Wennberg, J.E., Nease, R.F., et al., 1996. The

importance of patient preference in the decision to screen


23

for prostate cancer. The Journal of General Internal

Medicine 11, 342±349.

Gafni, A., Charles, C., Whelan, T., 1998. The physician as a

perfect agent for the patient versus the informed treatment

decision-making model. Social Science & Medicine 47,

347±354.

Guadagnoli, E., Ward, P., 1998. Patient participation in de-

cision-making. Social Science & Medicine 47, 329±339.

Haug, M.R., Lavin, B., 1981. Practitioner or patient: whose

in charge? Journal of Health and Social Behavior 22, 212±

229.

Haug, M., Lavin, B., 1983. Consumerism in Medicine. Sage,

Beverly Hills.

Hughes, T.E., Larson, L.N., 1991. Patient involvement in

health care: a procedural justice viewpoint. Medical Care

29, 297±303.

Iscoe, N.A., Goel, V., Wu, K., et al., 1994. Variation in breast

cancer surgery in Ontario. Canadian Medical Association

Journal 150, 345±352.

Katz, S.J., Charels, C., Lomas, J., Welch, H.G., 1997.

Shooting inward as the circle closes. Journal of Health

Politics, Policy and Law 22 (6), 1413±1431.

Leape, L.L., Hilborn, L.H., Park, R.E., et al., 1993. The

appropriateness of use of coronary artery bypass graft sur-

gery in New York State. Journal of the American Medical


Levine, M.N., Gafni, A., Markham, B., et al., 1992. A bed-

side decision instrument to elicit a patient's preference con-

cerning adjuvant chemotherapy for breast cancer. Annals

of Internal Medicine 117, 53±58.

Liao, L., Jollis, J.G., DeLong, E.R., et al., 1996. Impact of an

interactive video on ischematic heart disease patient de-

cision making. Journal of General Internal Medicine 11,

373±376.

Llewellyn-Thomas, H.A., 1995. Patients' health-care decision

making: a framework for descriptive and experimental in-

vestigations. Medical Decision Making 15, 101±106.

Lomas, J., Anderson, G.M., Karin, D.P., et al., 1989. Do

practice guidelines guide practice? The e�ect of a consen-

sus statement on the practice of physicians. New England

Journal of Medicine 321, 1306±1311.

Lomas, J., 1990. Quality assurance and e�ectiveness in health

care: an overview (editorial). Quality Assurance in Health

Care 2, 5±12.

Lomas, J., Contandriopoulous, A.P., 1994. Regulating limits

to medicine: towards harmony in public- and self-regu-

lation. In: Evans, R.G., Barer, M.L., Marmor, T.R.

(Eds.), Why are Some People Healthy and Others Not?

The Determinants of the Health of Populations. Adline

De Gruyter, New York, pp. 253±283.

Lomas, J., Lavis, J., 1996. Guidelines in the mist. CHEPA

Working Paper #96-231996. Centre for Health Economics

and Policy Analysis, McMaster University, Hamilton,

Ontario.

Mishler, E.G., 1984. The Discourse of Medicine. Ablex

Publishing Corporation, New Jersey.

Nay®eld, S.G., Bongiovanni, G.C., Alciatti, M.H., et al.,

1994. Statutory requirements for disclosure of breast can-

cer treatment alternatives. Journal of the National Cancer

Institute 86, 1202±1208.

O'Connor, A.M., Tugwell, P., Wells, G., 1994. Testing a por-

table, self-administered, decision aid for post-menopausal

women considering long-term hormone replacement

therapy (HRT) to prevent osteoporosis and heart disease.

Medical Decision Making 14, 414±438.

Ontario Ministry of Health, 1994. Consent to Treatment: A

Guide to the Act. Ontario Queen's Printer for Canada,

Ontario.

Parsons, T., 1951. The Social System. The Free Press, New

York Chapter 10.

Parsons, E., Atkinson, P., 1992. Lay constructions of genetic

risk. Sociology of Health and Illness 14, 437±455.

Quill, T.E., 1983. Partnership in patient care: a contractual

approach. Annals of Internal Medicine 98, 228±234.

Quill, T.E., Brody, H., 1996. Physician recommendations and

patient autonomy: ®nding a balance between physician

power and patient choice. Annals of Internal Medicine

125, 763±769.

Roos, N.L., 1984. Hysterectomy: variations in rates across

small areas and across physicians' practices. American

Journal of Public Health 74, 327±335.

Roos, N.L., Wennberg, J.E., McPherson, K., 1988. Using

diagnosis-related groups for studying variations in hospital

admissions. Health Care Financing Review 9, 53±62.

Ryan, M., 1992. The economic theory of agency in health

care: lessons from non-economists for economistsHealth

Economics Research Unit Discussion Paper, 03/921992.

University of Aberdeen, Scotland.

Sebban, C., Browman, G., Gafni, A., et al., 1995. Design and

validation of a bedside decision instrument to elicit a

patient's preference concerning allogenic bone marrow

transplantation in chronic myeloid leukemia. American

Journal of Hematology 48, 221±227.

Turney, J., 1996. Public understanding of science. The Lancet

347, 1087±1090.

Thomasma, D.C., 1983. Beyond medical paternalism and

patient autonomy: a model of physician conscience for the

physician±patient relationship. Annals of Internal

Medicine 98, 243±248.

van den Hoonaard, W.C., 1997. Working with Sensitizing

Concepts: Analytical Field Research. In: Qualitative

Research Methods Series, 41. Sage, Thousand Oaks, CA.

Wennberg, J.E., 1990. Better policy to promote the evaluative

clinical sciences. Quality Assurance in Health Care 2, 21±

29.

Wennberg, J.E., Freeman, J.L., Culp, W.J., 1987. Are hospital

services rationed in new haven or over-utilized in Boston?

Lancet 1, 1185±1188.

Whelan, T.J., Levine, M.N., Gafni, A., Lukka, H., Mohide,

E.A., Patel, M., Streiner, D.L., 1995. Breast irradiation

post-lumpectomy: development and evaluation of a de-

cision instrument. Journal of Clinical Oncology 13, 847±

853.

Williams, S., Calnan, M., 1996. The `limits' of medicalization?

Modern medicine and the lay populace in `late' modernity.

Social Science & Medicine 52, 1609±1620.


24

Fax +41 61 306 12 34E-Mail [email protected]

Special Article

Psychother Psychosom 2008;77:219–226 DOI: 10.1159/000126073

Systematic Review of the Effects of Shared Decision-Making on Patient Satisfaction, Treatment Adherence and Health Status

E.A.G. Joosten a, b L. DeFuentes-Merillas a, b G.H. de Weert c T. Sensky e

C.P.F. van der Staak d C.A.J. de Jong a, b

a Novadic-Kentron, Network for Addiction Treatment Services, St-Oedenrode , b Nijmegen Institute forScientist-Practitioners in Addiction (NISPA), Nijmegen , c Julius Center for Health Sciences and Primary Health Care, UMC Utrecht , and d Academic Centre for Social Sciences, Radboud University Nijmegen, Nijmegen , The Netherlands; e Department of Psychological Medicine, Imperial College London, London , UK

ness has been done to date. It has been argued that SDM is particularly suitable for long-term decisions, especially in the context of a chronic illness, and when the intervention contains more than one session. Our results show that under such circumstances, SDM can be an effective method of reaching a treatment agreement. Evidence for the effective-ness of SDM in the context of other types of decisions, or in general, is still inconclusive. Future studies of SDM should probably focus on long-term decisions.

Copyright © 2008 S. Karger AG, Basel

Introduction

In recent decades, there has been an increasing em-phasis on patient involvement in treatment decisions [1] . The role of the clinician is no longer an authoritarian per-son ‘who knows what’s right for you’. The relationship between clinician and patient has become more of a part-nership [2] . Placing the patient at the centre of care [3] represents a new and important approach to improve the quality of medical care. Patient autonomy is seen as a ba-sic value and underlying premise for the provision of healthcare itself [4] . Furthermore, in Europe the World Health Organisation has highlighted the need to involve patients in the development and delivery of healthcare

Key Words

Shared decision-making � Adherence � Patient satisfaction � Quality of life � Well-being

Abstract

Background: In the last decade, the clinician-patient rela-tionship has become more of a partnership. There is growing interest in shared decision-making (SDM) in which the clini-cian and patient go through all phases of the decision-mak-ing process together, share treatment preferences, and reach an agreement on treatment choice. The purpose of this re-view is to determine the extent, quality, and consistency of the evidence about the effectiveness of SDM. Method: This is a systematic review of randomised controlled trials (RCTs) comparing SDM interventions with non-SDM interventions. Eleven RCTs met the required criteria, and were included in this review. Results: The methodological quality of the stud-ies included in this review was high overall. Five RCTs showed no difference between SDM and control, one RCT showed no short-term effects but showed positive longer-term ef-fects, and five RCTs reported a positive effect of SDM on out-come measures. The two studies included of people with mental healthcare problems reported a positive effect of SDM. Conclusions: Despite the considerable interest in ap-plying SDM clinically, little research regarding its effective-

Published online: April 16, 2008

E.A.G. Joosten, MA Nijmegen Institute for Scientist-Practitioners in Addiction Academic Centre for Social Sciences, Radboud University Nijmegen, PO Box 9104NL–6500 HE Nijmegen (The Netherlands) Tel. +31 24 361 1150, Fax +31 24 361 1152, E-Mail [email protected]

© 2008 S. Karger AG, Basel0033–3190/08/0774–0219$24.50/0

Accessible online at:www.karger.com/pps

Dow

nloa

ded

by:

Sta

nfor

d U

niv.

Med

. Cen

ter

171.

65.6

5.46

- 1

/6/2

014

11:1

8:54

PM

25

http://dx.doi.org/10.1159%2F000126073

Joosten /DeFuentes-Merillas /de Weert /Sensky /van der Staak /de Jong

Psychother Psychosom 2008;77:219–226220

and legislation has been passed in several countries aimed at strengthening the influence of patients [5] .

One method of fostering these modern priorities of the clinician-patient relationship is through the process of shared decision-making (SDM). SDM is defined as an approach in which the clinician and patient go through all phases of the decision-making process together and in which they share the preference for treatment and reach an agreement on treatment choice [2, 6–8] . Forms of de-cision-making can be regarded as a continuum with two extremes – the ‘traditional medical model’ and the ‘in-formed medical model’ [2, 9, 10] . Table 1 demonstrates where SDM fits between these extremes.

Charles et al. [11] have identified necessary criteria for or characteristics of SDM. The first characteristic is that SDM involves clinician and patient. Often the treatment decision involves more than one patient and one clini-cian. The involvement of family members in treatment decision-making may be important. Furthermore, steps are taken to ensure that clinician and patient are both in-volved in the process of decision-making. Additionally, both parties take steps to build a consensus about the pre-ferred treatment. At the very least, the clinician needs to explain the treatment alternatives and their possible con-sequences for the patient. The patient and clinician both bring information and values into their discussion. Fi-nally, the patient and clinician together discuss and eval-uate treatment options and together build a consensus on the treatment to implement.

Frequently, SDM studies comprise the use of decision aids. Decision aids are interventions designed to help people make specific and deliberative choices among op-tions by providing relevant information about the op-tions and outcomes relevant to a person’s health status [12] . A systematic review of decision aids concluded that they improve patients’ knowledge regarding treatment options and their condition [12] . Decision aids appear to have no effect on satisfaction with decision-making, anx-iety, and health outcomes.

Published SDM studies have reported improvements in patient satisfaction, treatment adherence, quality of life and well-being [13] when clinicians adopt a patient-centred approach [14] and when patients are more in-volved and perceive greater control over their treatment choice [11, 15–17] . SDM should foster a patient-centred approach and empower patients, therefore the outcomes mentioned above are appropriate to use in assessing the effectiveness of SDM. The aim of the present systematic review was to examine the extent, quality and consisten-cy of the published research evidence for the effectiveness of SDM with respect to these outcome variables.

Method

Inclusion and Exclusion Criteria The review included studies that met all the following criteria:

(1) studies in which a treatment decision needed to be made; (2) randomised controlled trial design; (3) involving patients aged 18

Table 1. Models of shared decision-making about treatment [2, 42]

Paternalistic model(‘traditional medical model’)

Shared decision-making ‘Informed medical model’

Role of the clinician

Active: Reports only selected information to the patient, chooses the therapy he considers best for the patient.

Active: Reports all information and treatment possibilities to the patient.Can recommend an option. Decides on the therapy together with the patient.

Passive: Reports all information and treatment possibilities to the patient. Withholds his recommendations. Makes no decision.

Role of the patient

Passive: Accepts the proposal of the clinician. Is obliged to cooperate in his recovery.

Active: Receives all information. Forms his own judgement on harms and benefits of treatment options. Discusses hispreferences with the clinician. Decideson the therapy together with the clinician.

Active: Receives all information. Forms his own judgement.Is free to choose between all options unbiased by his clinicians’ own option.Decides on the therapy alone.

Information One way (largely):Clinician ] patient

Two way:Patient } clinician

One way (largely):Clinician ] patient

Deliberation Clinician alone or with other clinicians

Clinician and patient(plus potential others)

Patient (plus potential others)

Who decides? Clinician Clinician and patient Patient

Dow

nloa

ded

by:

Sta

nfor

d U

niv.

Med

. Cen

ter

171.

65.6

5.46

- 1

/6/2

014

11:1

8:54

PM

26

Effects of SDM on Patient Satisfaction, Treatment Adherence and Health Status

Psychother Psychosom 2008;77:219–226 221

years or older faced with having to make a treatment decision;(4) comparing SDM with a control intervention, and (5) includ-ing one or more of the following outcome measures: degree of treatment adherence, patient satisfaction, well-being, and quality of life.

Excluded studies were those in which treatment decisions based on a choice between alternative treatment options did not explicitly involve shared decision-making between clinician and patient.

Search Strategy for Identification of Studies The literature was searched with the WebSPIRS 5 search en-

gine with the PsycINFO and Medline databases (from 1966 until July 2006). In addition, the Cochrane library, 2006, issue 2, was screened. All databases were searched from their dates of com-mencement. Reference lists of relevant studies were checked for further potential sources. The search was run with the following keywords: shared decision making, shared decisionmaking, shared decision-making, shared decision * , decision making 1 , randomised controlled trials 1 , adherence, patient compliance 1 , patient-partici-pation 1 , patient satisfaction 1 , well-being and quality of life 1 . All the decision-making terms were combined with OR, and all the out-come measures were combined with the decision-making terms with AND. Only studies that were published in the English lan-guage were included.

Methodological Quality Assessment Two reviewers (E.J. and L.D.F.) independently assessed the

methodological quality of the randomised controlled trials (RCTs). The list of criteria recommended in the guidelines for systematic reviews issued by the Cochrane Back Review Group [18] was used, but adapted for this SDM review by the addition of four criteria (marked A–D in table 3 ). These items, key character-istics of SDM identified by Charles et al. [11] , are: (A) SDM in-volves at least two participants – clinician and patient; (B) both parties share information; (C) both parties take steps to build a consensus about the preferred treatment, and (D) an agreement is reached on the treatment to implement.

An item was scored ‘positive’ (+) if the criterion was met, ‘neg-ative’ (–) if it was not met, or ‘unclear’ (?) if it was not clear wheth-er the criterion was met or not. Subsequently, all authors were consulted for additional information about the criteria that were scored ‘unclear’ in their studies. A total score was computed by counting the number of positive scores.

Results

Study Selection The search resulted in 328 references via PsychINFO,

659 in Medline, and 373 in Cochrane Library withthe keywords: shared decision * , shared decisionmaking, shared decision making and shared decision-making . These keywords in combination with satisfaction, adher-ence/compliance, quality of life or well-being resulted in

26 references via PsychINFO, 121 in Medline and 32 in Cochrane Library. After deleting duplicates from all the databases consulted, the search finally resulted in 137 different references.

The first selection was based on titles, keywords and abstracts, and resulted in selecting 34 studies in a single reviewer format. The other articles were not primary re-search studies but reviews, editorials, letters, and quality improvement reports. All 34 studies included at least two patient groups: SDM and Non-SDM. Of these 34 studies, 17 were RCTs and 17 were non-RCTs (observational, pro-spective cohort, cross-sectional, self-reported, explorato-ry, and case control studies). Of the 17 RCTs, 11 focussed on SDM and included as outcome measures adherence, patient satisfaction, quality of life or well-being. Of the six excluded RCTs, in one article [19] the patient made the treatment decision, one measured clinician satisfaction [20] , and four did not explicitly involve SDM of clinician and patient and/or were based on a choice between alter-native treatment options [21–24] .

Study Characteristics Characteristics of the included studies (n = 11) are

shown in table 2 . Nine studies involved SDM in physical healthcare: three in cancer, and the remainder in ulcer disease, ischemic heart disease, hormone replacement therapy, dentistry, and benign prostatic hypertrophy. Two studies examined the effects of SDM in mental healthcare (treatment for schizophrenia and depression).

It is important to take into account that the SDM in-terventions of the selected studies are heterogeneous (see also table 2 ). Most of the interventions used an interactive videodisc program, resulting in an individual summary of important points. A consultation with the clinician followed this program and results were discussed [25–28] . Other studies used a question/information sheet [29, 30] or card ranking discussion [31] to promote patient participation in the treatment decision. In addition, gen-eral practitioners were trained in SDM skills [32] and be-haviour change strategies were used to increase patient involvement [33] . Finally, two studies used SDM inter-ventions within a treatment program [34–36] .

Methodological Quality In general, the methodological quality of the studies

included in this review was high ( table 3 ). All studies had six ( 1 50%) or more positive scores on the validity criteria, the determined threshold for high quality [18] .

Most of the studies did not include a blinded care pro-vider (criterion 5). Furthermore, most patients were not 1 MeSH or thesaurus terms.

Dow

nloa

ded

by:

Sta

nfor

d U

niv.

Med

. Cen

ter

171.

65.6

5.46

- 1

/6/2

014

11:1

8:54

PM

27



blinded regarding their treatment allocation (criterion 4). Most of the studies had three or four positive scores on the criteria of the key characteristics of SDM. The de-cision-making in all studies involved at least two parties (criterion A). However, in some studies it was not clear whether patient and clinician took steps to build a con-sensus (criterion C). In addition, none of the studies had a direct measure of the quality of the SDM interventions.

Problems arise in interpreting studies that yielded nega-tive results – are these results attributable to failure of effective SDM, or to poor SDM technique or applica-tion?

Effectiveness of SDM Five of the identified RCTs [31, 26–28, 30] reported no

difference between SDM and control group on the out-

Table 2. Study characteristics of included studies

Population Intervention Outcome measures Follow-up

Results (significantoutcomes betweenconditions)study (first

author)condition n duration decision aid clinician

training

Greenfield 1985 [33]

Ulcer disease 45 Singlesession

No (education of the patient)

No Patient satisfaction with care; increase of knowledge;physical limitations

6 to 8weeks

Patient satisfaction:no diff erencesIncrease of knowledge:SDM < controlPhysical limitations:SDM > control

Morgan 2000 [25]

Ischemic heart disease 240 Singlesession

Interactive videodisc presenting treatment options

No Patient satisfaction;increase of knowledge

6 months Patient satisfaction:no differencesIncrease of knowledge: SDM > control

Gattellari 2001 [30]

Cancer 335 Singlesession

Question sheet No Satisfaction withthe consultation

3, 6, 9 and 12 months

Satisfaction:no differences

Murray2001 [26]

Benign prostatic hypertrophy(primary care)

112 Singlesession

Interactive videodisc technique with booklet and printed summary

No Quality of life;anxiety; general health status

3 and 9 months

Quality of life, anxiety,and general health status:no differences

Murray 2001 [27]

Hormone replacement therapy (primary care)

205 Singlesession

Interactive videodisc technique with booklet and printed summary

No Quality of life;anxiety; general health status

3 and 9 months

Quality of life, anxiety,and general health status:no differences

Malm 2003 [34]

Schizophrenic disorders 84 Multiple sessions

No (treatmentprogram)

Yes Costumer satisfaction

2 years Costumer satisfaction:SDM > control

Ruland 2003 [28]

Cancer 52 Singlesession

Computerized support system with assessment summaries

No Patient satisfaction

Patient satisfaction:no differences

Von Korff 2003 [35]; Ludman 2003 [36]

Depression 386 Multiple sessions

No (treatmentprogram)

Yes Adherence;depression

3, 6, 9 and 12 months

Adherence anddepression outcomes:SDM > control

Edwards 2004 [31]

GP patients with known atrial fibrillation, prosta-tism, menorrhagia or menopausal symptoms

747 Single session

No (training) Yes Satisfaction 1 month Satisfaction:SDM < control

Van Roosmalen 2004 [32]

BRCA 1/2 mutation carriers

88 Multiple sessions

Card ranking(treatment options)

No Well-being 3 and 9 months

Well-being:SDM > control

Johnson 2006 [29]

Dentistry 70 Single sessions

EndoDB decision aid for treatment choices when root canal therapy or extraction is indicated

Yes Satisfaction;anxiety;increase of knowledge

Satisfaction: no differencesAnxiety: no differencesIncrease of knowledge: SDM > control

Dow

nloa

ded

by:

Sta

nfor

d U

niv.

Med

. Cen

ter

171.

65.6

5.46

- 1

/6/2

014

11:1

8:54

PM

28



come measures. These RCTs had in common that they involved decision-making in physical healthcare. Fur-thermore, these studies were dealing with a single deci-sion or measurement after one consultation. In contrast with the former studies, van Roosmalen et al. [32] , al-though showing no short-term effects, found a positive effect in the long term. The five remaining studies [25, 26, 29, 34–36] reported improved outcomes attributable to SDM. Of these five, Malm et al. [34] and Von Korff et al. [35] /Ludman et al. [36] involved decision-making in mental healthcare.

The most frequently used outcome measure, in seven studies, was patient satisfaction. The study by Malm et al. [34] is the only RCT that showed a positive outcome for patient satisfaction. The SDM intervention in this study concerned a treatment program, in contrast to the studies that report no differences between the conditions [25, 28–31, 33] . These studies involved a single decision or measurement after one consultation.

Other outcome measures were psychological and physical well-being (e.g. quality of life, anxiety, and de-pression). Two [32, 35, 36] out of five studies [26, 27, 29] showed positive effects of SDM with regard to these out-come measures. Van Roosmalen et al. [32] found a posi-tive effect with respect to well-being. The study by Von Korff et al. [35] /Ludman et al. [36] applied depression outcome measures. These studies are heterogeneous.

Nevertheless, these studies involved patients making longer-term decisions and/or having chronic diseases. Another similarity between these studies is the duration of the intervention. The intervention in the study by van Roosmalen et al. [32] contains three sessions added to the treatment as usual and a 12-month intervention was used in the study by Von Korff et al. [35] /Ludman et al. [36] .

One selected study had adherence as outcome mea-sure [35, 36] . The patients in the intervention condition were significantly more likely to adhere to the medica-tion (antidepressant) at 9- to 12-month follow-up mea-surement.

Finally, three selected studies [25, 29, 33] – in addition to the included outcome measures degree of treatment adherence, patient satisfaction, well-being, and quality of life – considered patient knowledge as outcome measure. The studies by Morgan et al. [25] and Johnson et al. [29] found no difference between intervention and control groups with regard to satisfaction. However, patients in the intervention group had significantly higher knowl-edge scores. In contrast with these studies, Greenfield et al. [33] reported that knowledge of ulcer disease was sig-nificantly greater for controls after intervention. On the other hand, this study showed that patients in the exper-imental group reported significantly fewer physical limi-tations due to ulcer disease than controls.

Table 3. Methodological quality of randomized controlled trials

Study (first author) Criteria validity/reliability Criteria key characteristics of SDM Total

1 2 3 4 5 6 7 8 9 10 11 total A B C D total

Greenfield, 1985 [33] ? ? – ? + + + + + + – 6 + + ? ? 2 8Morgan, 2000 [25] + + + – – – + – + + – 6 + + + + 4 10Gattellari, 2001 [30] + + + – + + + + – + + 9 + + – ? 2 11Murray, 2001 [26] + + + – – – + + + + + 8 + + ? + 3 11Murray, 2001 [27] + + + – – – + + + + + 8 + + ? + 3 11Malm, 2003 [34] + + + – + + + + + + + 10 + + + + 4 14Ruland, 2003 [28] + – + + – + + + + + – 8 + + + + 4 12Von Korff, 2003 [35];

Ludman, 2003 [36]b + + + – – + + + + + + 9 + + + + 4 13Edwards, 2004 [31]c + + + + + + + + + + – 10 + + + + 4 14Van Roosmalen, 2004 [32]a + – + – – – + + + + + 7 + ? ? ? 1 8Johnson, 2006 [29] + – + + – + + + + + – 8 + + + + 4 12

1 = Adequate randomisation procedure; 2 = concealment of treatment allocation; 3 = similarity of baseline characteristics; 4 = blinding of patients; 5 = blinding of care provider; 6 = outcome assessor blinded to the intervention; 7 = co-interventions avoided or equal; 8 = compliance; 9 = withdrawal/dropout rate; 10 = similarity of timing outcome assessment; 11 = intention-to-treat analyses; A = the decision-making involves at least two parties (clini-cian-patient); B = both parties share the information during the intervention; C = both parties take steps to build a consensus; D = agreement is reached on the treatment to implement.

a See also van Roosmalen et al. [43]. b See also Katon et al. [44, 45]. c See also Elwyn et al. [46].

Dow

nloa

ded

by:

Sta

nfor

d U

niv.

Med

. Cen

ter

171.

65.6

5.46

- 1

/6/2

014

11:1

8:54

PM

29



Discussion

SDM has been proposed as an important advance in modern clinical practice, and clinicians have been urged to adopt it in order to foster relationships with their pa-tients that are more appropriate to the modern age. To our knowledge, this is the first systematic review of the effectiveness of SDM.

The methodological quality of the studies included in this review was high, and most studies included three or all four key characteristics of SDM. The two RCTs in mental healthcare scored positively on all four SDM cri-teria.

Limitations of the review include that it focussed only on English language publications. Furthermore, the het-erogeneity of the samples, settings, and measurements might affect the generalisation of the results, although this heterogeneity also highlights the fact that SDM is a generic intervention, not dependent on the treatment set-ting or specialty. Moreover, a crucial point is that no study was excluded from the review on the basis of its ratings on the key characteristics of the SDM elements.

The studies with positive results are various. Studies that showed improvement in satisfaction, adherence, de-pression, and well-being had in common that the SDM interventions concerned treatment programs or con-tained more than one session. In addition, these success-ful studies involved patients making longer-term deci-sions and/or having chronic diseases, while most of the studies that did not show significant outcomes involved single or specific decisions. SDM can be regarded most appropriately as a collaborative process rather than one or two isolated events, and it is therefore not surprising that individual decisions have no significant measurable effect on factors that SDM might be expected to influ-ence. A longer interaction between clinician and patient is perhaps necessary for patient’s attitudes regarding de-sire for information and participation in medical deci-sions to manifest themselves in information-seeking be-haviour [37, 38] .

People with chronic illnesses (e.g. people with schizo-phrenia, diabetes, asthma) have to change their lifestyle and habits to improve their health status. Montori et al. [39] indicated that treatment decisions in chronic care, relative to acute care decisions, are more likely to require a more active patient role in carrying out the decision and also offer extensive chances to revisit and reverse deci-sions. Acute care decisions may involve minimal patient participation, are often urgent, and may be irreversible. While longer-term decisions in chronic illness usually of-

fer much scope for patient autonomy, urgent decisions about acute care usually involve the clinician in a more paternalistic role.

In addition to the included outcome measures (degree of treatment adherence, patient satisfaction, well-being, and quality of life), increase of knowledge was often used in the selected studies. Knowledge is an important and frequently named outcome measure. Patients need to be informed about complex decisions. They need to compre-hend the treatment options and their benefits and harms in order to consider and discuss these with their clinician [12] . These studies had in common that they involved sin-gle or specific decisions. However, the results of these studies were contradicting.

It is interesting that the two studies regarding mental healthcare both reported a positive effect of SDM [34–36] . Patients in mental healthcare are particularly en-couraged to focus on treatment issues that might affect their lifestyle and preferences [9] . Also, both of these studies focussed on people with chronic illnesses, and it has been demonstrated that chronically ill patients who are actively involved in health decision-making are more likely to enact and adhere to health behaviours, and to engage in other health-promoting or health-maintaining behaviours [40] .

Conclusions

The good-quality research identified in this review in-dicates that SDM can be an effective and useful way of reaching a treatment agreement when patients have to make long-term decisions. Furthermore, research shows that SDM interventions are effective when they concern treatment programs or contain more than one session.

Several lessons can be learned from the studies re-viewed, to apply to future research on SDM. First, it is probably inappropriate to evaluate SDM in the context of single, acute decisions. Not only do the published studies reveal no benefits of SDM under such circumstances, but perhaps such benefits should not be expected in any case. Using a decision aid is unlikely to produce the benefits of SDM on its own. The wider literature on SDM indicates that SDM is often complex and usually very time con-suming [13, 41] . Selecting the best treatment is not always easy. It is likely that SDM has to take place in a structured, more frequent, and longer interaction. Second, given the complex nature of SDM, it is probably appropriate for tri-als to include multiple outcome measures. In addition, increase of knowledge seems to be an important topic in

Dow

nloa

ded

by:

Sta

nfor

d U

niv.

Med

. Cen

ter

171.

65.6

5.46

- 1

/6/2

014

11:1

8:54

PM

30



SDM studies. Unfortunately, results of the reviewed stud-ies were contradictory and further research is needed to confirm the idea that SDM increases knowledge. Third, as well as assessing appropriate outcomes, it is essential for future studies to examine the process of SDM. This is important to measure the quality of the intervention. Without this, if a trial produces a negative result, it is im-possible to tell whether this outcome is due to the poor quality of SDM in the study, or to the ineffectiveness of SDM in a particular context or application. In addition, investigation of the process of SDM would be expected to lead to better understanding of the key elements of SDM. Fourth, theoretic considerations suggest that SDM should be effective in decision-making with people with mental disorders [9, 10] , and little research that is currently avail-able supports this. More research is clearly needed in this area.

The available evidence indicates that SDM can be ef-fective in the context of chronic illness and when the in-tervention contains more than one session. However, considering the growing clinical interest in SDM, it is surprising and disappointing how little randomised con-trolled studies have been published regarding its efficacy. There is therefore an urgent need for further research.

Acknowledgements

We wish to thank the authors of the original papers for their comments on this review. The Dutch Ministry of Welfare and Sports (VWS) and the Dutch Organization for Health Research and Development (ZonMW) funded this project (grant No. 985-10-018). These agencies had no role in the conduct or interpreta-tion of the study.

References

1 Say RE, Thomson R: The importance of pa-tient preferences in treatment decisions – challenges for doctors. BMJ 2003; 327: 542–545.

2 Charles C, Whelan T, Gafni A: What do we mean by partnership in making decisions about treatment? BMJ 1999; 319: 780–782.

3 Grol R: Improving the quality of medical care. Building bridges among professional pride, payer profit, and patient satisfaction. JAMA 2001; 286: 2578–2585.

4 Sensky T: Withdrawal of life-sustaining treatment. BMJ 2002; 325: 175–176.

5 Crawford MJ, Aldridge T, Bhui K, Rutter D, Manley C, Weaver T, Tyrer P, Fulop N: User involvement in the planning and delivery of mental health services: a cross-sectional sur-vey of service users and providers. Acta Psy-chiatr Scand 2003; 107: 410–414.

6 Jordan JL, Ellis SJ, Chambers R: Defining shared decision making and concordance: Are they one and the same? Postgrad Med J 2002; 78: 383–384.

7 Montgomery AA, Fahey T: How do patients’ treatment preferences compare with those of clinicians? Qual Health Care 2001; 10(suppl I):i39–i43.

8 Frosch DL, Kaplan RM: Shared decision making in clinical medicine: past research and future directions. Am J Prev Med 1999; 17: 285–294.

9 Hamann J, Leucht S, Kissling W: Shared de-cision making in psychiatry. Acta Psychiatr Scand 2003; 107: 403–409.

10 Faller H: Shared decision making: Ein An-satz zur Stärkung der Partizipation des Pati-enten in der Rehabilitation. Rehabilitation (Stuttg) 2003; 42: 129–135.

11 Charles C, Gafni A, Whelan T: Shared deci-sion-making in the medical encounter: What does it mean? (or it takes at least two to tan-go). Soc Sci Med 1997; 44: 681–692.

12 O’Connor AM, Stacey D, Entwistle V, Llewellyn-Thomas H, Holmes-Rovner M, Tait V, Tetroe J, Fiset V, Barry M, Jones J: De-cision aids for people facing health treatment or screening decisions. The Cochrane Data-base of Systematic Reviews 2003; Issue 1.Art No: CD001431. DOI 10.1002/14651858.CD001431.

13 Ong LML, DeHaes CJM, Hoos AM, Lammes FB: Doctor-patient communication: a re-view of the literature. Soc Sci Med 1995; 40: 903–918.

14 Rimondini M, Del Piccolo L, Goss C, Mazzi M, Paccaloni M, Zimmermann C: Commu-nication skills in psychiatry residents – How do they handle patient concerns? Psychother Psychosom 2006; 75: 161–169.

15 Montgomery AA, Harding J, Fahey T: Shared decision making in hypertension: the impact of patient preferences on treatment choice. Fam Pract 2001; 18: 309–313.

16 Towle A, Godolphin W: Framework for teaching and learning informed shared deci-sion making. BMJ 1999; 319: 766–771.

17 Savage R, Armstrong D: Effect of a general practitioner’s consulting style on patients’ satisfaction: a controlled study. BMJ 1990; 301: 968–970.

18 Van Tulder M, Furlan A, Bombardier C, Bouter L, and the Editorial Board of the Co-chrane Collaboration Back Review Group: Updated method guidelines for systematic reviews in the Cochrane Collaboration Back Review Group. Spine 2003; 28: 1290–1299.

19 Whelan T, Levine M, Willan A, Gafni A, Sanders K, Misky D: Effect of a decision aid on knowledge and treatment decision mak-ing for breast cancer surgery. JAMA 2004; 292: 435–441.

20 Sullivan MD, Leigh J, Gaster B: Training in-ternists in shared decision making about chronic opioid treatment for non-cancer pain. J Gen Intern Med 2006; 21: 360–362.

21 Volk RJ, Cass AR, Spann SJ: A randomized controlled trial of shared decision making for prostate cancer screening. Arch Fam Med 1999; 8: 333–340.

22 Wieringa-DeWaard M, Bindels PJE, Vos J, Bonsel GJ, Stalmeier PFM, Ankum WM: Pa-tient preferences for expectant management vs. surgical evacuation in first-trimester un-complicated miscarriage. J Clin Epidemiol 2004; 57: 167–173.

23 Myers ED, Branthwaite A: Out-patient com-pliance with antidepressant medication. Br J Psychiatry 1992; 160: 83–86.

24 Llewellyn-Jones RH, Baikie KA, Smithers H, Cohen J, Snowdon J, Tennant CC: Multifac-eted shared care intervention for late life de-pression in residential care: randomised controlled trial. BMJ 1999; 319: 676–682.

25 Morgan MW, Deber RB, Llewellyn-Thomas HA, Gladstone P, Cusimano RJ, O’Rourke K, Tomlinson G, Detsky AS: Randomized, con-trolled trial of an interactive videodisc deci-sion aid for patients with ischemic hearth disease. J Gen Intern Med 2000; 15: 685–693.

26 Murray E, Davis H, See Tai S, Coulter A, Gray A, Haines A: Randomised controlled trial of an interactive multimedia decision aid on benign prostatic hypertrophy in pri-mary care. BMJ 2001; 323: 493–498.

Dow

nloa

ded

by:

Sta

nfor

d U

niv.

Med

. Cen

ter

171.

65.6

5.46

- 1

/6/2

014

11:1

8:54

PM

31

http://dx.doi.org/10.1002%2F14651858



27 Murray E, Davis H, See Tai S, Coulter A, Gray A, Haines A: Randomised controlled trial of an interactive multimedia decision aid on hormone replacement therapy in pri-mary care. BMJ 2001; 323: 490–493.

28 Ruland CM, White T, Stevens M, Fancuillo G, Khilani SM: Effects of a computerized system to support shared decision making in symptom management of cancer patients: preliminary results. J Am Med Inform Assoc 2003; 10: 573–579.

29 Johnson BR, Schwartz A, Goldberg J, Koer-ber A: A chairside aid for shared decision making in dentistry: a randomized con-trolled trial. J Dent Educ 2006; 70: 133–141.

30 Gattellari M, Butow PN, Tattersall MHN: Sharing decisions in cancer care. Soc Sci Med 2001; 52: 1865–1878.

31 Edwards A, Elwyn G, Hood K, Atwell C, Rob-ling M, Houston H, Kinnersley P, Russell I; Study Steering Group: Patient-based out-come results from a cluster randomized trial of shared decision making skill development and use of risk communication aids in gen-eral practice. Fam Pract 2004; 21: 347–354.

32 Van Roosmalen MS, Stalmeier PFM, Verhoef LCG, Hoekstra-Weebers JEHM, Oosterwijk JC, Hoogerbrugge N, Moog U, van Daal WA: Randomized trial of a shared decision-mak-ing intervention consisting of trade-offs and individualized treatment information for BRCA1/2 mutation carriers. J Clin Oncol 2004; 16: 3293–3301.

33 Greenfield S, Kaplan S, Ware JE Jr: Expand-ing patient involvement in care, effects on patient outcome. Ann Intern Med 1985; 102: 520–528.

34 Malm U, Ivarsson B, Allebeck P, Falloon IRH: Integrated care in schizophrenia: a two-year randomized controlled study of two community-based treatment programs. Acta Psychiatr Scand 2003; 107: 415–423.

35 Von Korff M, Katon W, Rutter C, Ludman E, Simon G, Lin E, Bush T: Effect on disability outcomes of a depression relapse prevention program. Psychosom Med 2003; 65: 938–943.

36 Ludman E, Katon W, Bush T, Rutter C, Lin E, Simon G, Von Korff M, Walker E: Behav-ioural factors associated with symptom out-comes in a primary care-based depression prevention intervention trial. Psychol Med 2003; 33: 1061–1070.

37 Beisecker AE, Beisecker TD: Patient infor-mation-seeking behaviours when communi-cating with doctors. Med Care 1990; 28: 19–28.

38 Deber RB: Physicians in health care manage-ment. 7. The patient-physician partnership: changing roles and the desire for informa-tion. Can Med Assoc J 1994; 151: 171–176.

39 Monitori VM, Gafni A, Charles C: A shared treatment decision-making approach be-tween patients with chronic conditions and their clinicians: the case of diabetes. Health Expect 2006; 9: 25–36.

40 Goldring AB, Taylor SE, Kemeny ME, Anton PA: Impact of health beliefs, quality of life, and physician-patient relationship on the treatment intentions of inflammatory bowel disease patients. Health Psychol 2002; 21: 219–228.

41 Davis RE, Dolan G, Thomas S, Atwell C, Mead D, Nehammer S, Moseley L, Edwards A, Elwyn G: Exploring doctor and patient views about risk communication and shared decision-making in the consult. Health Ex-pect 2003; 6: 198–207.

42 Charles C, Gafni A, Whelan T: Decision making in the physician-patient encounter: revisiting the shared treatment decision-making model. Soc Sci Med 1999; 49: 651–661.

43 Van Roosmalen MS, Stalmeier PFM, Verhoef LCG, Hoekstra-Weebers JEHM, Oosterwijk JC, Hoogerbrugge N, Moog U, van Daal WA: Randomised trial of a decision aid its timing for women being tested for a BRCA1/2 muta-tion. Br J Cancer 2004; 90: 333–342.

44 Katon W, Rutter C, Ludman EJ, Von Korff M, Lin E, Simon G, Bush T, Walker E, Unutzer J: A randomized trial of relapse prevention of depression in primary care. Arch Gen Psy-chiatry 2001; 58: 241–247.

45 Katon W, Von Korff M, Lin E, Simon G, Walker E, Unützer J, Bush T, Russo J, Lud-man E: Stepped collaborative care for pa-tients with persistent symptoms of depres-sion: a randomized trial. Arch Gen Psychiatry 1999; 56: 1109–1115.

46 Elwyn G, Edwards A, Hood K, Robling M, Atwell C, Russell I, Wensing M, Grol R; Study Steering Group: Achieving involve-ment: process outcomes from a cluster ran-domized trial of shared decision making skill development and use of risk communi-cation aids in general practice. Fam Pract 2004; 21: 337–346.

Dow

nloa

ded

by:

Sta

nfor

d U

niv.

Med

. Cen

ter

171.

65.6

5.46

- 1

/6/2

014

11:1

8:54

PM

32

Shared Treatment Decision Making Improves Adherenceand Outcomes in Poorly Controlled Asthma

Sandra R. Wilson1, Peg Strub2, A. Sonia Buist3, Sarah B. Knowles1, Philip W. Lavori4, Jodi Lapidus3,William M. Vollmer5, and the Better Outcomes of Asthma Treatment (BOAT) Study Group*

1Palo Alto Medical Foundation Research Institute, Palo Alto, California; 2The Permanente Medical Group, San Francisco, California; 3Oregon Health

and Science University, Portland, Oregon; 4Stanford University School of Medicine, Stanford, California; and 5The Kaiser Permanente Center for

Health Research, Portland, Oregon

Rationale: Poor adherence to asthma controller medications resultsin poor treatment outcomes.Objectives: To compare controller medication adherence and clinicaloutcomes in 612 adults with poorly controlled asthma randomizedto one of two different treatment decision-making models or tousual care.Methods: In shared decision making (SDM), nonphysician cliniciansand patients negotiated a treatment regimen that accommodatedpatient goals and preferences. In clinician decision making, treat-ment was prescribed without specifically eliciting patient goals/preferences. The otherwise identical intervention protocols bothprovided asthma education and involved two in-person and threebrief phone encounters.Measurements and Main Results: Refill adherence was measured usingcontinuous medication acquisition (CMA) indices—the total days’supply acquired per year divided by 365 days. Cumulative controllermedication dose was measured in beclomethasone canister equiv-alents. In follow-up Year 1, compared with usual care, SDM resultedin: significantly better controller adherence (CMA, 0.67 vs. 0.46; P ,

0.0001) and long-acting b-agonist adherence (CMA, 0.51 vs. 0.40;P 5 0.0225); higher cumulative controller medication dose (canisterequivalent, 10.9 vs. 5.2; P , 0.0001); significantly better clinicaloutcomes (asthma-related quality of life, health care use, rescuemedication use, asthma control, and lung function). In Year 2,compared with usual care, SDM resulted in significantly lower rescuemedication use, the sole clinical outcome available for that year.Compared with clinician decision making, SDM resulted in: signifi-cantly better controller adherence (CMA, 0.67 vs. 0.59; P 5 0.03) andlong-acting b-agonist adherence (CMA, 0.51 vs. 0.41; P 5 0.0143);higher cumulative controller dose (CMA, 10.9 vs. 9.1; P 5 0.005); andquantitatively, but not significantly, better outcomes on all clinicalmeasures.Conclusions: Negotiating patients’ treatment decisions significantlyimproves adherence to asthma pharmacotherapy and clinical out-comes.Clinical trials registered with www.clinicaltrials.gov (NCT00217945and NCT00149526).

Keywords: randomized controlled trial; asthma control; patient–clinician

communication

Among patients with asthma, and others with chronic condi-tions that require pharmacotherapy, only about half take their

medications at therapeutically effective doses (1, 2). Estimatednonadherence rates for asthma controller medications rangefrom 30 to 70% (3–6), including in patients with so-calleddifficult asthma who might appear to require treatment withstill more potent medications (7). Poor adherence exacerbatesairway inflammation, and may result in suboptimal asthmacontrol, functional limitations, decreased quality of life, excesshealth care use, and even death (8, 9).

A recent Cochrane Review (12) of adherence studies, in-cluding those that appeared to demonstrate improved adherenceto some types of medications (e.g., antihypertensives) (10, 11),found serious methodologic problems, thereby limiting anyconclusions regarding intervention efficacy. Furthermore, noneof the included studies focused on asthma medication adherence.Consequently, there is a general lack of evidence of effectiveadherence interventions targeting adults with asthma, and spe-cifically, poorly controlled asthma.

Observational studies suggest that failure to elicit and addresspatients’ individual circumstances and goals/preferences regard-ing their regimen may contribute to treatment nonadherence(13). Asthma treatment guidelines recommend that cliniciansconsider patients’ treatment goals, but little is known aboutclinician adherence to these recommendations or the effects oftheir doing so (8).

Charles and colleagues (14, 15) hypothesized that a sharedtreatment decision-making (SDM) process, in which the patient

AT A GLANCE COMMENTARY

Scientific Knowledge on the Subject

The effects of shared patient–clinician decision makingregarding asthma care on treatment decisions, treatmentadherence, and asthma-related clinical outcomes have notbeen experimentally evaluated.

What This Study Adds to the Field

In a randomized controlled trial, patients with poorlycontrolled asthma who shared in making decisions abouttheir treatment showed significantly better adherence toasthma controller medications and to long-acting b-agoniststhan patients who participated in either of two controlconditions. As a result of both their medication choicesand better adherence, patients with shared decision-making(SDM) received a higher cumulative dose of anti-inflamma-tory medication over a 1-year period. Compared to usualcare, SDM was also associated with significantly betterasthma-related quality of life, fewer asthma-related medicalvisits, lower use of rescue medication, higher likelihood ofwell controlled asthma, and better lung function.

(Received in original form June 16, 2009; accepted in final form December 17, 2009)

* A complete list of members may be found at the end of the article.

Supported by National Institutes of Health grants R01 HL69358 and R18

HL67092.

Correspondence and requests for reprints should be addressed to Sandra R.

Wilson, Ph.D., Palo Alto Medical Foundation Research Institute, 795 El Camino

Real, Ames Building, Palo Alto, CA 94301. E-mail: [email protected]

This article has an online supplement, which is accessible from this issue’s table of

contents at www.atsjournals.org

Am J Respir Crit Care Med Vol 181. pp 566–577, 2010

Originally Published in Press as DOI: 10.1164/rccm.200906-0907OC on December 17, 2009

Internet address: www.atsjournals.org

33

has actively participated, will result in a greater commitmentand adherence to the selected regimen than to a regimenselected by the physician alone. The authors described fourkey defining features of the SDM model, namely, that bothclinician and patient: (1) share relevant information; (2) expresstreatment preferences; (3) deliberate the options; and (4) agreeon the treatment to implement.

There is a paucity of strong evidence from appropriatelycontrolled trials that supports the inference from observationalstudies that SDM regarding treatment of a chronic disease (inwhich self-management is essential) actually results in patientsaccepting and adhering to the regimen and improves bothtreatment adherence and disease outcomes (16, 17). Althougha recent review of SDM by Joosten and colleagues (18) iden-tified 11 randomized controlled trials that met at least one ofCharles and colleagues’ criteria, none concerned asthma, andnearly half involved a one-time treatment decision, rather thandecisions typical of chronic disease management. Furthermore,only one included medication adherence as an outcome, andnone investigated clinical outcomes, other than a limited mea-sure of patient well being. Although Joosten and colleaguesnoted the potential effectiveness of the SDM process, theyrecommended additional research that should include multipleclinical outcomes.

Better Outcomes of Asthma Treatment (BOAT) was athree-arm, multisite, randomized, controlled trial in 612 patientswith poorly controlled asthma. The two experimental interven-tion arms were designed in the context of asthma care manage-ment, which refers to a period of targeted review of asthmatreatment and control and asthma self-management educationby a nonphysician health professional. The primary hypothesiswas that patients with poorly controlled asthma who receivedcare management using an SDM approach would exhibitgreater adherence to controller medications, better asthma-related quality of life, and lower health care utilization for acutesymptoms than patients who received usual care (no asthmacare management). A contingent secondary hypothesis wasthat, given a demonstrated benefit over usual care, patientswho participated in SDM would demonstrate better outcomesthan patients who received the identical care management,except that treatment was determined by the care manager andphysician alone (clinician decision making [CDM]). Secondaryclinical outcomes included short-acting b-agonist (SABA) use,lung function, and asthma control.

Some results presented here have been previously reportedin the form of abstracts based on preliminary results (19–23).

METHODS

The study has been approved annually by the institutional reviewboards of the Kaiser Foundation Research Institute in Oakland,California, and of the Kaiser Permanente (KP) Center for HealthResearch in Portland, Oregon, and Honolulu, Hawaii. Substantialadditional information about the study methods and timeline areavailable in the online supplement.

Patient Recruitment and Eligibility Criteria

The target population was patients whose asthma was not wellcontrolled, and whose adherence to their asthma regimen was likelyto be inadequate. KP members, aged 18–70 years, with evidencesuggestive of poorly controlled asthma, were identified at five clinicalsites using computerized records of overuse of rescue medications (acontroller/[controller 1 rescue medication] ratio <0.5 and at least threeb-agonist dispensings in the past year) or a recent asthma-relatedemergency department (ED) visit or hospitalization. Exclusion criteriaincluded intermittent asthma (brief exacerbations or symptoms less thanonce/wk), primary diagnosis of chronic obstructive pulmonary disease or

emphysema, insufficient pulmonary function reversibility (for ex-/currentsmokers and those without regular controller use), regular use of oralcorticosteroids, and current asthma care management.Spirometry. Spirometry was performed at enrollment using standard-ized research methods (24) and equipment that met American Tho-racic Society standards.

Randomization

A computer-based adaptive randomization algorithm (25) was usedto ensure concealment from randomization staff and better-than-chance balance among the three groups on age (18–34, 35–50, and 51–70 yr), sex, race/ethnicity, hospitalization in the prior two years (yes/no), and frequency of asthma controller use in the past week (none,1–3, >4 d).

Intervention Protocol

The SDM and CDM interventions were identical in format, content,and all patient education handouts and worksheets, except for theprocess by which treatment was decided.

Format. Session 1 of the intervention is outlined in Figure 1,highlighting the unique features of the SDM and CDM protocols. Insession 2 (z1 mo after session 1), and in three brief phone calls 3, 6,and 9 months later, patient progress was assessed and medications wereadjusted, as necessary, using the assigned care management approach.Except as noted subsequently here, both protocols used identicalstandardized interventionist scripts and materials.

The patient’s asthma history was elicited using a standardizedpatient information form, the patient’s level of asthma control wasclassified, and asthma education was provided. Once treatment wasnegotiated (SDM) or decided (CDM), patients were instructed in thecorrect use of the relevant inhaler medication devices using methodspreviously shown by this team to improve inhaler technique andeliminate common usage errors (26, 27). At the end of session 1, awritten asthma management and action plan was created, and potentialbarriers to medication adherence were elicited and addressed usingmotivational interviewing techniques (28). Any subsequent changesmade at session 2 or in a follow-up phone call were documented in theplan.

Treatment decision process. In the CDM model, the care managerprescribed an appropriate regimen based on the patient’s level ofasthma control, and explained that decision to the patient. The SDMmodel implemented the four key defining features described byCharles and colleagues (14, 15). The care manager elicited the patient’sgoals for treatment and relative priorities regarding symptom control,regimen convenience, avoidance of side effects, and cost. The patientwas then shown a list of the full range of regimen options for all levelsof asthma severity, based on the then-current national asthma guide-lines (29) and KP pharmacopeia. These options differed with respect tothe number and type(s) of medications, dosing, and schedule. Usinga simple worksheet, the patient and clinician then compared the prosand cons of all of the options the patient wished to consider, whichincluded the option of continuing the patient’s current de facto regimen(i.e., how they were using their current asthma medications) to arrive ata treatment that best accommodated the patient’s and care manager’sgoals.

For both groups, a SABA was always prescribed for rescue use asneeded. If indicated, treatment of allergic rhinitis and/or gastro-esophageal reflux disease was prescribed (CDM) or negotiated(SDM).

Care Manager Training and Intervention Quality Control

A total of 16 KP nurses, respiratory therapists, and pharmacists, as wellas nurse practitioners and physician assistants, most of whom alreadyserved as asthma care managers, were recruited to serve as study caremanagers and assigned to the SDM or CDM program. SDM and CDMcare managers were trained separately and worked independently.

For quality control purposes, audiotapes of both sessions of 10% ofthe patients were scored on a detailed performance checklist to de-termine whether the two protocols were delivered as intended. In addi-tion, patients were given a stamped postcard to return to the researchoffice after session 1 to report their perceived role in the treatmentdecision.

Wilson, Strub, Buist, et al.: Shared Treatment Decision Making 567

34

Coordination with Patient’s Physician

Care managers documented each encounter in the patient’s chart,where it also was available to the patient’s physician. The care man-agers discussed their recommendations with the physician if they or thephysician had any questions about the new regimen. For care managerswho were not licensed to prescribe, the physician reviewed and wrotethe prescription.

Usual Care Control Condition

Usual asthma care at KP medical centers was based on a stepped-careapproach to pharmacotherapy with the aim of long-term asthma con-trol, as recommended by the National Asthma Education PreventionProgram’s Expert Panel Report 2 (29). At some KP sites, physiciansalso had the option to refer patients to an asthma care managementprogram, typically of less than 6 months’ duration, in which a licensedhealth professional (nonphysician) provided asthma education andaddressed adherence and other medication use and self-managementissues in a manner similar to, but less structured than, the CDM in-

tervention. However, asthma care management was neither a requiredaspect of usual care nor necessarily available at all BOAT sites, andcurrent participation in that program was an exclusion criterion for thestudy. Once enrolled in BOAT, usual care and SDM or CDM patients(after the intervention phase) still had access to KP’s existing caremanagement services, if available, based on their physician’s referral.

Outcome Data

Pharmacy data. Medication acquisition data were extracted fromKP dispensing records for 1 year prerandomization and 2 years post-randomization. Fill/refill adherence was measured using a continuousmedication acquisition (CMA) index for each year, calculated as thetotal days’ supply acquired in a given year divided by 365 days (30–32).The index represents the proportion of the prescribed medicationsupply acquired by the patient during each 365-day period, and maypotentially overestimate, but not underestimate, actual use.

Acquisition indices were calculated separately for: (1) all asthmacontrollers (inhaled corticosteroids [ICS], leukotriene modifiers, cro-

Figure 1. Outline of intervention

protocols, with unique features

of each highlighted.

568 AMERICAN JOURNAL OF RESPIRATORY AND CRITICAL CARE MEDICINE VOL 181 2010

35

molyn sodium, and theophylline); (2) for ICS alone; and (3) for long-acting b-agonists (LABAs). Combination ICS-LABA medications(i.e., fluticasone-salmeterol) contributed to both the ICS and LABAindex calculations.

A second pharmacy measure, beclomethasone diproprionate can-ister equivalents, was the estimated total number of canisters ofbeclomethasone to which the asthma controller medications dispensedin a given year were equivalent (i.e., equipotent in terms of their anti-inflammatory effectiveness). Per the method of Schatz and colleagues(33), each dispensing of a controller medication (excluding theophyl-line), in any form or quantity, was assigned a weight representing itsequivalent in fractional or multiple canisters of beclomethasone 80 mg.The weighted values were summed for each patient for each 365-dayperiod to obtain a cumulative measure of controller medication dosedispensed.

A third pharmacy measure—a clinical outcome—used a separateset of weights to standardize the amounts of all SABAs dispensed interms of their bronchodilator effectiveness in canisters of albuterol,regardless of SABA type and delivery mode. The weights weresummed to obtain the total number of albuterol canister equivalentsacquired by the patient in each study year.

Other outcomes. The primary clinical outcomes were asthma-relatedquality of life and asthma health care utilization. Asthma-related qualityof life for the prior 2 weeks was assessed by patient self-report at baselineand follow-up Year 1, using the five-item Symptom Subscale of theJuniper Mini Asthma Quality of Life Questionnaire (34).

Health care utilization data included the date, diagnoses, facility,and service (ED, hospital in-patient, urgent care, or out-patient de-partment) of all visits to KP or contracted facility, including visits withInternational Classification of Diseases, Ninth Revision code prefix 493.These data were extracted from KP databases for 1 year prerandom-ization and 1 year postrandomization to calculate the annual asthma-related visit rate for each patient. BOAT intervention sessions werenot included in these rates.

In addition to SABA use (see above), secondary clinical outcomesincluded self-reported asthma control and lung function measured atbaseline and follow-up Year 1. Asthma control in the preceding 4weeks was assessed using the four-item Asthma Therapy AssessmentQuestionnaire (ATAQ) (35). Lung function measures included FEV1

expressed as percent of the predicted value based on age-, sex-, andrace-specific norms (36), and the ratio of FEV1 to FEV6, expressed as apercentage.

Statistical Analysis

For each outcome, the primary hypothesis was that patients whoparticipated in SDM would demonstrate a significant advantagerelative to patients under usual care. The statistical significance ofthe secondary hypothesis, that SDM would demonstrate an advantageover CDM, was considered only if the primary null hypothesis wasrejected. Given the conditional sequence of the hypotheses, noadjustment for multiple comparisons was needed to preserve the type1 error rate of the secondary comparison (SDM vs. CDM) at the a levelof 0.05. Because BOAT was designed to test an SDM model, therewere no a priori hypotheses regarding differences between the CDMand usual care groups; however, results of these comparisons areinformative and are presented for completeness.

Multivariable generalized linear regression analysis was used toestimate the intervention effect on each outcome (except ATAQ) atfollow up, controlling for the baseline value of that outcome, site, andthe randomization balancing variables. To estimate the odds ratios ofhaving well-controlled asthma (ATAQ score 5 0) at follow-up Year 1relative to usual care, multivariable logistic regression analysis wasused. Baseline estimates, overall and by group, are presented withoutadjustment for any other variable. Missing data were not imputed:baseline and follow-up analyses were restricted to those patients withcomplete data for the analytic model variables at both time points.

The ratings scores of the care managers, patients, and qualitycontrol evaluator regarding the treatment decision process were com-pared using the Wilcoxon test. Group differences in patient character-istics and asthma medication regimens were tested using either x2 testsor t tests. All group differences were tested using a two-sided a of 0.05,and all analyses used SAS software version 9.2 (37).

RESULTS

Recruitment

Initially, 5,414 patients were identified as being potentiallyeligible (Figure 2), of whom 2,534 were contactable and pro-vided informed consent for preliminary eligibility screening(38). The final sample size was 612 (n 5 204/group): Honolulu,n 5 114; Oakland/Richmond, n 5 180; Portland, n 5 196; andSan Francisco, n 5 122. At the Year 1 follow-up clinic visit, 551patients completed the patient assessment and lung functiontests. Baseline and follow-up pharmacy and utilization datawere extracted for all patients from existing clinical /adminis-trative records.

Baseline Characteristics

As intended by the selection procedures and eligibility criteria,the sample consisted primarily of persons whose asthma, at base-line, was poorly or very poorly controlled (83.9%) (Table 1) whenclassified based on symptoms, rescue medication use, and lungfunction per guidelines of the Global Initiative for Asthma (39).

Intervention Process

Length and quality. Session 1 lasted an average of 77 (617)minutes for the CDM group and 106 (622) minutes for theSDM group. Session 2 averaged 31 (618) minutes and 32 (619)minutes, respectively. The follow-up phone contacts togetheraveraged 30 (625) minutes per patient for the CDM group and35 (625) minutes per patient for the SDM group. Interventionfidelity was high (see the online supplement).

Intervention cost estimate. Using KP salary guidelines, anaverage $54 per-hour rate for care manager salary and benefits(z$112,000/yr) was assumed and applied to the average lengthof the intervention sessions and follow-up telephone contacts,plus an average estimated 20 minutes additional time per pa-tient for documentation and visit reminders. The estimated costper patient treated was $174 using the SDM model (z3.2 h) and$142 using the CDM model (z2.6 h), a cost difference of $32.

Treatment decisions. At the conclusion of session 1, therewere no differences in the proportions of the SDM and CDMgroups whose regimen included an ICS or who were prescribeda LABA or allergic rhinitis medication (Table 2). However, forabout 13% more SDM patients than CDM patients, the de-cision process resulted in selection of a higher dose fluticasonepropionate preparation (220 mg) rather than the higher strengthbeclomethasone dipropionate (80 mg), KP’s formulary-preferred ICS, or than the lower-dose fluticasone propionatepreparation (110 mg).

Medication Acquisition

Controller use. During the prerandomization year, approxi-mately 18% of patients did not acquire any controller medica-tion, and overall adherence, per CMA values, was very poor. Infollow-up Year 1, the adjusted mean acquisition index for allcontroller medications was significantly higher in the SDMgroup (CMA, 0.67) compared with both the usual care (CMA,0.46; P , 0.0001) and CDM groups (CMA, 0.59; P 5 0.029)(Figure 3A), and also significantly higher in the CDM than inthe usual care group (0.59 vs. 0.46; P 5 0.0008). Similarly, infollow-up Year 1, the adjusted mean index for ICS alone wassignificantly higher in the SDM group (CMA, 0.59) than in boththe usual care (CMA, 0.37; P , 0.0001) and CDM groups(CMA, 0.52; P 5 0.017), and significantly higher in the CDMthan in the usual care group (0.52 vs. 0.37; P , 0.0001).

In follow-up Year 2, the SDM group’s adjusted mean CMAindices for all controllers, and for ICS separately, remained


36

higher than at baseline, but were no longer significantly higherthan the usual care or CDM groups’ values (Figure 3A).

Controller regimen anti-inflammatory potency. In follow-upYear 1, the SDM group acquired more than twice as manybeclomethasone canister equivalents than the usual care group(10.9 vs. 5.2; P , 0.0001), and also significantly more than theCDM group (10.9 vs. 9.1; P 5 0.005; Figure 3B). The differencebetween CDM and usual care was also significant (9.1 vs. 5.2;P , 0.0001). In follow-up Year 2, the adjusted mean canisterequivalents acquired by the SDM group remained greater thanits baseline value (7.1 vs. 4.9), and continued to be significantlygreater than those of the usual care (mean, 4.6; P 5 0.0002) andCDM groups (mean, 5.8; P 5 0.04; Figure 3B). There was nolonger a significant difference between CDM and usual care.

LABA use. In the prerandomization year, 22.2% of thepatients acquired a LABA at least once, and 11.0% of thoseprescribed a LABA acquired an ICS-LABA combination. Infollow-up Year 1, significantly higher proportions of both SDMand CDM patients acquired a LABA compared with usual carepatients (Table 3), and among those, patients in the SDM groupwere more likely than those in the CDM group to acquirea combination preparation (41.1% vs. 23.2%; P 5 0.005).Among patients whose regimen included a LABA, the adjusted

mean LABA acquisition index was significantly higher in theSDM group (CMA, 0.51) than either the usual care (CMA, 0.40;P 5 0.0225) or CDM groups (CMA, 0.41; P 5 0.0143; Figure3C). There was no significant difference between the CDM andusual care groups.

For those on a LABA in follow-up Year 2, the adjusted meanLABA acquisition index remained significantly higher in theSDM group (CMA, 0.52) than in either the usual care (CMA, 0.42;P 5 0.0296) or CDM group (CMA, 0.43; P 5 0.0346) (Figure 3C),with no significant difference between the CDM and usual caregroups. Patients in the SDM group also continued to be signifi-cantly more likely to acquire a LABA at least once than patientsunder usual care (Table 3), and significantly more likely to beusing a combination ICS-LABA preparation than patients underCDM.

Clinical Outcomes

Asthma-related quality of life. At follow-up Year 1, both theSDM (mean, 5.5) and CDM groups (mean, 5.4) had significantlyhigher adjusted mean symptom subscale scores than the usualcare group (mean, 5.1; respective P 5 0.0003 and 0.009), but didnot differ significantly from each other (Figure 4A). Further-more, 70.3% of the SDM group had a score increase of greater

Figure 2. Case progress through the

Better Outcomes of Asthma Treatment(BOAT) study: identification, eligibility

determination, initial assessment, ran-

domization, intervention, and follow

up. 1Patients were identified as poten-tially eligible based on their recent hos-

pitalization or emergency department

visit and a medication ratio of 60.50,

indicating an overuse of rescue medi-cation. 2Not contactable includes pa-

tients whose primary care physicians

(PCPs) were not notified, PCP did notrespond, PCP did not assent, letters

were not sent/received, or calls were

not successfully completed. 3Non-

screenable patients were not screenedfor multiple reasons, including disinter-

est in participating in the study. 4In-

cludes persons who passively refused

by failing to keep two or more enroll-ment appointments. 5Reasons for in-

eligibility include failure to meet

spirometry criterion and other psycho-social and medical prerandomization

exclusion criteria (e.g., drug rehabilita-

tion or currently receiving asthma care

management, etc.). 6Excludes five post-randomization exclusions for previously

undiscovered bahavioral/mental health

problems: shared decision making

(SDM), n 5 1; clinician decision making(CDM), n 5 4.


37

than 0.50 points, compared with 55.6% of the UC group and61.1% of the CDM group.

Health care use. During the prerandomization year, approx-imately 35% of patients had no asthma-related visits. In follow-up Year 1, both the SDM and CDM groups had significantlylower visit rates (1.0/yr and 1.1/yr) than the usual care group(1.4/yr; P 5 0.0161 and 0.0147, respectively) (Figure 4B).

SABA use. In follow-up Year 1, the SDM group acquiredsignificantly fewer albuterol canister equivalents (adjustedmean, 6.5) than the usual care group (adjusted mean, 8.1; P 5

0.002), but not the CDM group (adjusted mean, 7.1; P 5 0.09)(Figure 3D). The CDM group also used significantly less SABAthan usual care (7.1 vs. 8.1; P 5 0.038). In follow-up Year 2,SABA use continued to decrease for all three groups as KPinstituted policies limiting the number of SABA refills thatphysicians could authorize per prescription. The SDM groupcontinued to use significantly less SABA than the usual caregroup (4.7 vs. 6.3; P 5 0.0141) and less than the CDM group (4.7

vs. 6.0; borderline P 5 0.06), but with no significant differenceremaining between CDM and usual care.

Asthma control. At follow-up Year 1, ATAQ scores de-creased in each group (SDM D 5 20.80, CDM D 5 20.54, UCD 5 20.46), but the SDM group had nearly twice the odds ofreporting no asthma control problems (ATAQ 5 0) than theusual care group (odds ratio, 1.9; 95% confidence intervals[CIs], 1.3–2.9; P 5 0.002) (Figure 4C), but not significantlygreater odds than the CDM group. The CDM group also hadgreater odds of no control problems than the usual care group(odds ratio, 1.6; 95% CI, 1.1–2.4; P 5 0.0239).

Lung function. At follow-up Year 1, the adjusted meanpercent predicted FEV1 for the SDM group was significantlygreater than the usual care group (76.5% vs. 73.1%; P 5

0.0068), but not the CDM group (76.5% vs. 75.8%; P 5 0.47)(Figure 5A). The adjusted mean FEV1:FEV6 ratio was alsosignificantly greater for the SDM group compared with theusual care group (72.8% vs. 70.0%; P 5 0.0005), but not theCDM group (72.8% vs. 71.8%; P 5 0.09) (Figure 5B). However,there was no significant difference between the CDM and usualcare groups (71.8% vs. 70.0%; P 5 0.07).

Patient-Perceived Roles in Treatment Decision Making

On the postcards mailed back after session 1, patients in theSDM group anonymously rated their influence on the treatmentselection as being approximately the same as the care manager’sinfluence (mean rating, 3.1 6 0.6 on the five-point scale), butwith neither being more influential than the other. The ratingsof patients in the SDM group were significantly different fromthose in the CDM group, with the latter feeling that their caremanagers had a greater influence (mean, 2.5 6 0.9; P , 0.0001)than they themselves did.

DISCUSSION

Patients who shared in decisions regarding their asthma treat-ment were significantly more likely to adhere to ICS and othercontroller therapy, and also to LABA medications, than patients

TABLE 1. BASELINE CHARACTERISTICS OF BETTER OUTCOMESOF ASTHMA TREATMENT PARTICIPANTS, BY GROUP

Characteristics UC CDM SDM

Demographic characteristics

Mean age, years* 45.1 6 12.4 46.9 6 12.1 45.7 6 13.3

Sex*

Female 117 (57.4) 114 (55.9) 115 (56.4)

Male 87 (42.6) 90 (44.1) 89 (43.6)

Ethnicity*

Caucasian 127 (62.3) 124 (60.8) 128 (62.8)

African American 30 (14.7) 34 (16.7) 32 (15.7)

Asian 22 (10.8) 18 (8.8) 20 (9.8)

Hispanic 8 (3.9) 9 (4.4) 9 (4.4)

Pacific Islander 17 (8.3) 16 (7.8) 15 (7.4)

American Indian 0 (0.0) 3 (1.5) 0 (0.0)

Education

Less than high school

diploma

6 (2.9) 2 (1.0) 6 (2.9)

High school diploma/some

college

116 (56.9) 132 (65.0) 114 (55.9)

4-yr college degree or higher

education

82 (40.2) 69 (34.0) 84 (41.1)

Family income .$40,000/yr 134 (69.1) 139 (70.9) 133 (66.8)

Ever told by doctor they had COPD 11 (5.4) 14 (6.9) 4 (2.0)

Current smoker 33 (16.2) 33 (16.2) 31 (15.2)

Asthma characteristics

Level of control

Very poorly controlled 85 (42.1) 82 (40.2) 79 (38.7)

Poorly controlled 83 (41.1) 87 (42.7) 96 (47.1)

Moderately well controlled 29 (14.4) 24 (11.8) 17 (8.3)

Well controlled 5 (2.5) 11 (5.4) 12 (5.9)

Asthma controller medication use*

None 50 (24.5) 43 (21.1) 42 (20.6)

1–3 d/wk 38 (18.6) 44 (21.6) 41 (20.1)

>4 d/wk 116 (56.9) 117 (57.4) 121 (59.3)

Hospitalized for asthma in past 2 yr* 76 (37.3) 69 (33.8) 71 (34.8)

Daytime symptom frequency

,1/wk 11 (5.4) 12 (5.9) 14 (6.9)

>1/wk but ,daily 111 (54.4) 114 (55.9) 101 (49.5)

Daily 82 (40.2) 78 (38.2) 89 (43.6)

Nocturnal symptoms

<23/mo 113 (55.4) 116 (56.9) 114 (55.9)

.23/mo but ,53/mo 22 (10.8) 24 (11.8) 29 (14.2)

>53/mo 69 (33.8) 64 (31.4) 61 (29.9)

FEV1 % predicted

.80% 76 (37.6) 79 (38.7) 70 (34.3)

60–80% 62 (30.7) 65 (31.9) 78 (38.2)

,60% 64 (31.7) 60 (29.4) 56 (27.5)

Definition of abbreviations: CDM 5 clinician decision making; COPD 5 chronic

obstructive pulmonary disease; SDM 5 shared decision making; UC 5 usual care.

Values are expressed as n (%) or mean (6SD), n 5 204 per group.

* Randomization balancing variable.

TABLE 2. PRESCRIBED MEDICATIONS AT THE END OF SESSION1, BY GROUP

CDM* SDM*

Medication n (%) n (%) P Value

Any controller† 181 (97.8) 186 (97.4) 1.0‡

Any ICS 178 (96.2) 181 (94.8) 0.50x

Beclomethasone 80 108 (60.7) 90 (49.7)

Fluticasone 220k 53 (29.8) 78 (43.1) 0.032x

Other ICS{ 17 (9.6) 13 (7.2)

LABA** 91 (49.2) 92 (48.2) 0.84x

SABA 185 (100.0) 191 (100.0) N/A

Allergic rhinitis medication 48 (26.0) 39 (20.4) 0.20x

GERD medication 5 (2.7) 0 (0.0) 0.028‡

Definition of abbreviations: CDM 5 clinician decision making; ICS 5 inhaled

corticosteroid; GERD 5 gastroesophageal reflux disease; LABA 5 long-acting

b-agonist; N/A 5 not applicable; SABA 5 short-acting b-agonist; SDM 5 shared

decision making.

* SDM group, n 5 191; CDM group, n 5 185.† Controllers include ICSs, leukotriene modifiers, and theophylline, but not

LABAs. No patients in this sample were prescribed oral prednisone for daily/

alternate day use.‡ Fisher’s exact test for cell sizes <5.x P values estimated from Pearson’s x2.k Includes flovent 220 preparations in combination with a LABA.{ Includes ICS-LABA combination, budesonide, aerobid, beclomethasone 40,

fluticasone 110, and beclomethasone with strength unspecified (n 5 2).

** Includes single preparations (salmeterol and folmoterol) and ICS-LABA

combination preparations (fluticasone-salmeterol 100, fluticasone-salmeterol

250, and fluticasone-salmeterol 500).


38

who experienced either usual care or who received care manage-ment in which the clinician played the primary role in choosingthe treatment regimen. By virtue of both (1) their greater fill/refill adherence and (2) the pattern of their regimen choices,patients in the SDM group also acquired a significantly higheraverage daily dose of asthma controller medication (a largernumber of beclomethasone canister equivalents) than eitherpatients under usual care or active control patients. Addition-ally, patients who shared in making treatment decisions hadsignificantly better clinical outcomes on all six measures—

asthma-related quality of life, asthma health care utilization,use of rescue medication, lung function, and the likelihood ofwell-controlled asthma—compared with those receiving usualcare. Although the SDM approach, and the behavioral andregimen changes it induced, were not associated with signifi-cantly better clinical outcomes compared with the CDM ap-proach, the differences were consistently in a direction favoringSDM on both objectively measured and patient-reported out-comes. Furthermore, the clinician decision model only resultedin significantly better clinical outcomes compared with usual

Figure 3. Group differences in pharmacy outcomes. (A) Groupdifferences in controller medication acquisition for each study

period. Controller medications include inhaled corticosteroids

(ICS), leukotriene modifiers, and theophylline, and exclude long-

acting b-agonists (LABAs) and oral prednisone. Prerandomizationvalues are unadjusted, and follow-up values are adjusted for

baseline CMA and the balancing variables. The 95% confidence

intervals (CIs) are shown. Group differences and 95% CIs forcontroller CMA values at follow-up Year 1: shared decision making

(SDM)-usual care (UC) 5 0.21 (95% CI, 0.13–0.28); SDM–clinician

decision making (CDM) 5 0.08 (95% CI, 0.01–0.15); CDM-UC 5

0.13 (95% CI, 0.05–0.20). At follow-up Year 2: SDM-UC 5 0.03(95% CI, 20.05–0.11); SDM-CDM 5 0.04 (95% CI, 20.04–0.12);

CDM-UC 5 20.01 (95% CI, 20.09–0.07). The number of patients

per group at each time point: SDM, n 5 204; CDM, n 5 204; UC,

n 5 204. (B) Group differences in ICS canister equivalents for eachstudy period. Controller medications include inhaled corticoste-

roids and leukotriene modifiers, and exclude theophylline, LABAs

and oral prednisone. Prerandomization values are unadjusted, andfollow-up values are adjusted for baseline ICS canister equivalents

and the balancing variables. The 95% CIs are shown. Group

differences and 95% CIs for ICS canister equivalents at follow-up

Year 1: SDM-UC 5 5.8 (95% CI, 4.5–7.0); SDM-CDM 5 1.8 (95%CI, 0.57–3.1); CDM-UC 5 3.9 (95% CI, 2.6–5.2). At follow-up

Year 2: SDM-UC 5 2.5 (95% CI, 1.2–3.8); SDM-CDM 5 1.4 (95%

CI, 0.04–2.7); CDM-UC 5 1.1 (95% CI, 20.18–2.4). The number

of patients per group at prerandomization and follow-up Year 2 is:SDM, n 5 204; CDM, n 5 202; UC, n 5 204; and at follow-up Year

1 is: SDM, n 5 204; CDM, n 5 202; UC, n 5 203. (C) Group

differences in LABA acquisition, among those on a LABA, for each

study period. LABA medications include fluticasone-salmeterol100, fluticasone-salmeterol 250, fluticasone-salmeterol 500, and

formeterol. Prerandomization values are unadjusted and follow-up

values are adjusted for baseline CMA and the balancing variables.The 95% CIs are shown. Group differences and 95% CIs for LABA

CMA at follow-up Year 1: SDM-UC 5 0.11 (95% CI, 0.02–0.20);

SDM-CDM 5 0.09 (95% CI, 0.02–0.17); CDM-UC 5 0.01 (95%

CI, 20.08–0.10). At follow-up Year 2: SDM-UC 5 0.11 (95% CI,0.01–0.20); SDM-CDM 5 0.09 (95% CI, 0.01–0.18); CDM-UC 5

0.01 (95% CI, 20.08–0.11). The number of patients per group at

each time point is: prerandomization, SDM, n 5 40; CDM, n 5 44;

UC, n 5 52; follow-up Year 1: SDM, n 5 112; CDM, n 5 108; UC,n 5 59. (D) Group differences in short-acting b-agonist (SABA)

use for each study period. For each patient, the number of canister-

equivalents is the mean of the sum of all SABAs dispensed to thatpatient, each weighted relative to one canister of a standard albute-

rol canister. Prerandomization values are unadjusted, and follow-

up values are adjusted for baseline albuterol canister equivalents

and the balancing variables. The 95% CIs are shown. Groupdifferences and 95% CIs for SABA use at follow-up Year 1: SDM-

UC 5 21.6 (95% CI, 22.5 to 20.78); SDM-CDM 5 20.73 (95%

CI, 21.6 to 0.12); CDM-UC 5 20.89 (95% CI, 21.7 to 20.05).

At follow-up Year 2: SDM-UC 5 21.2 (95% CI, 22.1 to 20.24);SDM-CDM 5 20.91 (95% CI, 21.9–0.04); CDM-UC 5 20.28

(95% CI, 21.2–0.67). The number of patients per group at each

time point is: SDM, n 5 204; CDM, n 5 204; UC, n 5 204.


39

care on four of the six clinical outcomes, and not in significantlyless SABA use or a higher FEV1:FEV6 ratio. Only amongpatients in the SDM group was SABA use (the only clinicaloutcome available for the second follow-up year) significantlylower than that of usual care in follow-up Year 2.

The greater advantage of the SDM than the CDM modelover usual care, as well as the greater persistence of its ef-fectiveness in reducing SABA use, support a treatment prefer-ence for the SDM approach. However, a rigorous, long-termcost-benefit analysis is required to determine whether theseclinical benefits are accompanied by cost savings that offset thecost of the CDM or the additional cost of the SDM intervention.

There was no evidence that the SDM approach resulted ina significant proportion of patients avoiding corticosteroids orelecting inadequate doses. In fact, patient involvement resultedin higher proportions receiving the highest-dose fluticasone(220 mg) over the highest-dose beclomethasone (80 mg), andthe combination ICS-LABA over separate preparations. Bothtendencies appeared to be due to the greater convenience of theregimen (i.e., the need for fewer puffs of fluticasone [220 mg/d]than beclomethasone [80 mg/d] to achieve an equipotent dose),and the convenience of a single inhaler in the case of thecombination preparation. Without the patient’s active involve-ment, the CDM care managers tended to choose the formulary-recommended ICS and separate ICS and LABA preparations.

Significance of Findings

An SDM approach is consistent with the concept of patient-centered care, and this study demonstrates that it is an impor-tant component with significant potential to not only changepatient behavior through increased adherence, but also to im-prove clinical outcomes. The present findings have significantimplications for asthma treatment and research, and potentiallyfor the treatment of a wide range of other chronic conditions.

The findings also provide previously unavailable informationon the average degree of clinical improvement, on a range ofoutcome measures, that is associated with a specific averageincrease in the cumulative annual ICS dose. This finding mayhelp in evaluating the clinical importance of other interventionsdirected at improving medication adherence that may lack someor all of the clinical outcome measures obtained in the presenttrial.

The observation of a mean improvement in the quality of lifescore of 0.40 points, attributable to the SDM model, is less thanthe putative 0.50 minimal clinically important difference on thatmeasure (40). However, questions exist regarding the method-

ology used to establish that minimal clinically importantdifference value (41, 42). The fact that the SDM group reportedsignificantly higher quality of life at follow up, and that morethan 70% of the group experienced a score improvement ofgreater than 0.50 points, is additional evidence that the clinicalbenefits of the intervention were evident to the patients.

Methodological significance. Concern with the quality ofclinician–patient communication dates back at least 4 decades(43). Until now, observational studies have been the norm. Fewcontrolled experimental studies have been conducted of modifi-cations in communication around the treatment decision process,as distinct from other aspects of clinician–patient communica-tion, and none of those that have been conducted concernedasthma. Most have emphasized one-time or acute treatmentdecisions, rather than the ongoing decisions associated withchronic conditions. Previous research also has generally focusedon patient satisfaction, and has shown little evidence of signifi-cantly changing patient behavior or improving clinical outcomes.Furthermore, lack of assessment of the quality of the interven-tions, as delivered, has severely limited the interpretability of thelargely negative trials (16).

Attributing the observed adherence, regimen potency, andclinical benefits to the patients’ active participation in theirtreatment decisions is justified because the SDM and CDMinterventions were identical in all respects, except the treatment

Figure 3. (Continued).

TABLE 3. PRE- AND POSTRANDOMIZATION PERCENTAGE OFPATIENTS DISPENSED A LONG-ACTING b-AGONIST, BY GROUP

Dispensed

a LABA*

UC

n (%)

CDM

n (%)

SDM

n (%) P Value†

Prerandomization‡ SDM-UC: P 5 0.16

Yes 52 (25.5) 44 (21.6) 40 (19.6) SDM-CDM: P 5 0.62

No 152 (74.5) 160 (78.4) 164 (80.4) CDM-UC: P 5 0.35

Follow-up Year 1† SDM-UC: P , 0.0001

Yes 59 (28.9) 108 (52.9) 112 (54.9) SDM-CDM: P 5 0.69

No 145 (71.1) 96 (47.1) 92 (45.1) CDM-UC: P , 0.0001

Follow-up Year 2 SDM-UC: P 5 0.002

Yes 63 (30.9) 90 (44.1) 93 (45.6) SDM-CDM: P 5 0.77

No 141 (69.1) 114 (55.9) 111 (54.4) CDM-UC: P 5 0.006

Definition of abbreviations: CDM 5 clinician decision making; LABA 5 long-

acting b-agonist; SDM 5 shared decision making; UC 5 usual care.

* Includes fluticasone-salmeterol 100, fluticasone-salmeterol 250, fluticasone-

salmeterol 500, salmeterol, and formoterol.† P values estimated from Pearson’s x2.‡ UC 5 204; CDM 5 204; SDM 5 204.


40

decision process. This experimental difference was also reflectedin the perception of those in the SDM group that they had a greaterrole in the treatment decisions than did the patients in the CDMgroup. Previous controlled trials of SDM have given insufficientattention to the choice of the control condition. Joosten andcolleagues’ review (18) did not consider the appropriateness ofthe control condition as a design criterion; most studies reviewedsimply compared their intervention to the current standard of care.Without an active control for features of the intervention otherthan the treatment decision process (e.g., providing patient

education), it is difficult to know the extent to which any positiveresults are attributable specifically to the patient’s involvement inthe treatment choice. The contribution of the BOAT study isenhanced by the existence of such a control, which allowed theelucidation of the unique contribution of the shared decisionprocess itself.

Asthma care management. The target population of patientswith poorly controlled asthma was a specific subset of patientswith asthma within a very large managed health care systemthat had a long-standing commitment to high-quality asthma

Figure 4. Group differences in asthma-related quality of life,

asthma-related health care utilization, and asthma control. (A)

Group differences in asthma-related quality of life. The five-item

Mini Asthma Quality of Life Questionnaire (MAQLQ) subscale isscored on a symptom scale of 0 (All of the time) to 7 (None of the

time). Subscale items include: shortness of breath, bothered by

coughing, chest tightness or heaviness, difficulty sleeping, and

chest wheeze. Pre randomization values are unadjusted andfollow-up values are adjusted for the baseline score and the

balancing variables. 95% confidence intervals (CIs) are shown.

Group differences and 95% CIs for MAQLQ score at follow-upYear 1 are: shared decision making (SDM)-usual care (UC) 5 0.39

(95% CI, 0.18–0.60); SDM–clinician decision making (CDM) 5

0.11 (95% CI, 20.11–0.32); CDM-UC 5 0.28 (95% CI, 0.07–

0.50). The number of patients per group with no missing valuesat either time point is SDM n 5 182, CDM n 5 180, UC n 5 189.

(B) Group differences in the annual rate of asthma-related health

care utilization (visits/yr). Prerandomization values are unadjusted,

and follow-up values are adjusted for baseline asthma-relatedhealth care utilization and the balancing variables. The 95% CIs

are shown. Group differences and 95% CIs for medical visits at

follow-up Year 1 are: SDM-UC 5 20.36 (95% CI, 20.66 to 20.07);SDM-CDM 5 0.01 (95% CI, 20.29–0.30); CDM-UC 5 20.37

(95% CI, 20.67 to 20.07). The number of patients per group at

each time point is: SDM, n 5 204; CDM, n 5 204; UC, n 5 204. (C)

Odds ratios of well controlled asthma at each time period. Theusual care group was the referent value when estimating the odds

for the SDM and CDM groups. The CDM group also served as the

referent value for the SDM group. The Asthma Therapy Assessment

Questionnaire (ATAQ) scale is from 0 (no control problems; ‘‘wellcontrolled’’) to 4 (four control problems). Well controlled asthma is

based on an ATAQ score 5 0. Pre-randomization odds ratios are

unadjusted and follow-up odds ratios are adjusted for the baseline

ATAQ score and the balancing variables. 95% CIs are shown. Thenumber of patients per group with no missing values at either time

point is: SDM, n 5 182; CDM, n 5 180; UC, n 5 189.


41

care, education of patients with asthma, and physician adher-ence to asthma treatment guidelines, and that, at some sites,offered asthma care management as an optional part of usualmedical care. Virtually all of these patients had medicationbenefits with modest copayments that varied with the provisionsof their insurance plans. Nevertheless, in the baseline year,these patients had acquired only about one-third of the days’supply of medication that had been prescribed for them, andwere experiencing frequent symptoms and activity limitations.Nearly one-fifth were not using an asthma controller at all. Ourfindings reveal that care management using a clinician decisionmodel was clearly beneficial in terms of medication adherenceand many clinical outcomes, and suggest that the likelihood ofachieving the hoped-for benefits, and their magnitude, is in-creased by specifically involving the patient in the choice oftreatment.

Need for ongoing reinforcement. The fall-off in asthma con-troller adherence/acquisition that was observed during follow-upYear 2 in both care management conditions is not surprising, andsuggests that further follow up and reinforcement may beimportant to sustain the benefits of a shared decision processand of care management in general. For both models, theinterventions typically occurred very early in follow-up Year 1,with no external reinforcement of the intervention processes afterthe patients’ 9-month follow-up intervention phone calls.

Primary care providers and other clinicians at KP who mayhave seen patients subsequently had no access to the interven-

tion materials, and hence were very unlikely to have useda comparable shared treatment decision approach. Patients inboth care management conditions were also less likely thanpatients under usual care to have asthma-related medical visitsduring follow-up Year 1, which would also reduce the oppor-tunity for reinforcement.

The fall-off in adherence may also suggest that, havingexperienced a clinical benefit in Year 1, patients began to ‘‘stepdown’’ therapy on their own. There is a need for furtherinvestigation into the pattern and causes of the decline inmedication adherence over time, and whether periodic reviewby a care manager or physician can sustain both adherence andclinical benefits.

Intervention cost versus benefits. Compared with usual care,in follow-up Year 1 the SDM care management interventionincreased the total days’ supply of controller medicationacquired by the patient by an average of 77 days and by 9.6beclomethasone canister equivalents, increased the quality oflife score by 0.4 points, decreased asthma-related physicianvisits per year by 0.4 visits, reduced albuterol acquisition by 1.6canister equivalents, increased the FEV1:FEV6 ratio by anaverage of 2.8 percentage points, and doubled the likelihoodof having well controlled asthma. Although the SDM interven-tion required a per-patient investment of $174 for care managertime, and resulted in some increase in cost to the patient andhealth care system for medications, it also resulted in decreasedcosts for asthma-related provider visits. The study was not

Figure 5. Group differences in lung function. (A) Group differ-

ences in percent predicted FEV1 for each study period. Preran-

domization values are unadjusted, and follow-up values are

adjusted for baseline percent predicted FEV1 and the balancingvariables. The 95% confidence intervals (CIs) are shown. Group

differences and 95% CIs for percent predicted FEV1 at follow-up

Year 1 are: shared decision making (SDM)–usual care (UC) 5 3.2

(95% CI, 0.87–5.4); SDM–clinician decision making (CDM) 5

0.85 (95% CI, 21.4–3.1); CDM-UC 5 2.3 (95% CI, 0.04–4.6).

The number of patients per group with no missing values at

either time point is: SDM, n 5 165; CDM, n 5 170; UC, n 5 172.(B) Group differences in the ratio of FEV1:FEV6 for each study

period. The ratio is expressed as a percentage. Prerandomization

values are unadjusted, and follow-up values are adjusted for the

baseline FEV1:FEV6 ratio and the balancing variables. The 95% CIsare shown. Group differences and 95% CIs for the ratio of

FEV1:FEV6 at follow-up Year 1: SDM-UC 5 1.9 (95% CI, 0.84–

3.0); SDM-CDM 5 0.92 (95% CI, 20.14–2.0); CDM-UC 5 0.98

(95% CI, 20.07–2.0). The number of patients per group with nomissing values at either time point is: SDM, n 5 165; CDM, n 5

170; UC, n 5 172.


42

powered to detect specific differences in the more costly EDvisits and hospitalizations; hence, any cost savings in this regardare unknown, and should be the focus of future research.

Strengths

The SDM intervention included all four defining features of theSDM model (mutual information sharing, expressing treatmentpreferences, discussing the options, and agreeing on treatment).The study design tested the hypothesized benefits of this modelin a randomized, controlled trial with a very strong active, aswell as a passive, control group. Care managers’ adherence totheir respective intervention protocols was objectively assessed,and confirmed the fidelity of intervention delivery, and it wasdocumented that the interventions resulted in differing percep-tions of patients’ own influence on the treatment decisions.

Other strengths include the use of objective measures ofmedication acquisition and refill adherence and health care uti-lization, available for all patients during follow up, high-qualityspirometry, and multiple validated patient-centered measures.

Limitations

As an initial efficacy trial, this study was not powered to detectdifferences in ED visits or hospitalization rates—the most costlytypes of utilization; hence, a true cost–benefit analysis was notperformed. The results of this study are also limited to adultpatients; it remains to be determined whether the effects of ashared decision process can be generalized to pediatric patients(i.e., to treatment decisions made by parents on their child’sbehalf). Finally, in settings in which different treatment optionshave more pronounced differential cost implications for patients(e.g., non–managed care organizations), or in which asthmamanagement guidelines support different patterns of medicationusage (e.g., greater use of combination products), the priorities ofadult patients may be more or less consonant with clinicianrecommendations than was observed here.

Although pharmacy dispensing data were obtained for bothfollow-up years, the inability to continue active follow-up and toextract health care utilization data through follow-up Year 2 is amodest limitation. However, even a 1-year follow up of multiplebehavioral, clinical, and health care utilization outcomes greatlyexceeds the duration of most previous studies.

Conclusions

An SDM approach to the selection of asthma pharmacotherapy,in the context of asthma care management, is efficacious inimproving both medication adherence and clinical outcomes. Anappropriately powered study to determine the cost-effectivenessof this approach is warranted, as are further studies of the effec-tiveness of this approach in patients with other poorly controlledchronic diseases and in both younger and older patients.

Conflict of Interest Statement: S.R.W. received up to $1,000 from Asthmatx, Inc.,for consulting related to assessment of patient-centered outcomes of bronchialthermoplasty; P.S. does not have a financial relationship with a commercial entitythat has an interest in the subject of this manuscript; A.S.B. received $1,001–$5,000 from GlaxoSmithKline, $1,001–$5,000 from AstraZeneca, $1,001–$5,000from Merck, $1,001–$5,000 from Sepracor, and $1,001–$5,000 from ScheringPlough in advisory board fees; S.B.K. does not have a financial relationship witha commercial entity that has an interest in the subject of this manuscript; P.W.L.received $5,001–$10,000 from Pfizer, Inc., in consultancy fees, and $5,001–$10,000 from the Palo Alto Medical Foundation Research Institute in consultancyfees; J.L. does not have a financial relationship with a commercial entity that has aninterest in the subject of this manuscript; W.M.V. received up to $1,000 fromMerck in advisory board fees, more than $100,001 from GlaxoSmithKline, morethan $100,001 from Pfizer, more than $100,001 from Novartis, $50,001–$100,001 from Sepracor, and $50,001–$100,000 from Boheringer Ingelheim inunrestricted educational grants, holds more than $100,001 in stock ownership oroptions in Vanguard Health Care, a mutual fund, and is employed by Kaiser

Permanente, a managed care organization that has contracts with entities who areinterested in the American Thoracic Society.

Acknowledgment: The authors gratefully acknowledge the contributions of theData and Safety Monitoring Board members Stephen Lazarus, M.D., BruceBender, Ph.D., and Theodore Colton, Sci.D.; International Advisory Groupmembers Amiram Gafni, Ph.D., Elizabeth Juniper, M.C.S.P., M.P.H., CynthiaRand, Ph.D., Sean Sullivan, M.D., and Kevin Weiss, M.D.; Hawaii Better Out-comes of Asthma Treatment (BOAT) Clinical Site Director, Matthew Lau, M.D.;care managers Susan Amina, N.P., Ed Birnbaum, R.R.T., Daniel Cheung,Pharm.D., Kathy Jessen, R.R.T., C.P.F.T., Joe Kann, P.A., Sharmin Khajavi,Pharm.D., Tammy Law, Pharm.D., Jeanne Lewis-Hughes, R.N., Iris Maitland,R.N., Len Moriyama, R.R.T., Tanya Ramsey, C.P., Nancy Siegal, P.A., Ida Treist-man, R.N., Gary Vita, C.P., and Jan Vita, Pharm.D.; recruitment, assessment andfollow-up staff Veronica Luna, Karen Kriete, Amy Stone-Murai, Jodi Thirtyacre;pulmonary function measurement specialists Robert Jensen, M.D., and RobertCrapo, M.D.; Kaiser Division of Research (Oakland) investigator Stephen Van DenEeden, Ph.D., and data extractor, analyst, and programmer, Jun Shan, Ph.D.;Kaiser Center for Health Research (Portland) data extractors, analysts, andprogrammers Mike Allison, Jane Wallace, and Paul Cheek; Kaiser Center forHealth Research (Hawaii) data extractor, analyst, and programmer MarkSchmidt; and Palo Alto Medical Foundation Research Institute data analystsand programmers Qiwen Huang, M.S., Yinge Qian, M.S., and Shinu Verghese,M.S. The administrative assistance of Amy Waterbury, Kathy Stamm, and ShelleyClark, and the document production assistance of Kathy Stamm, in thepreparation of this manuscript are gratefully acknowledged.

Other members of the Better Outcomes of Asthma Treatment (BOAT) Study Group:Faith Bocobo, M.D., Don German, M.D., and Alaina Poon, Pharm.D., thePermanente Medical Group (San Francisco, CA); Myngoc Nguyen, M.D., thePermanente Medical Group (Oakland, CA); John Hoehne, M.D., the PermanenteMedical Group (Richmond, CA); Nancy Brown, Ph.D., Palo Alto MedicalFoundation (Palo Alto, CA); Christine Fukui, M.D., Hawaii Permanente MedicalGroup (Honolulu, HI); and Joan Holup, M.A., the Kaiser Permanente Center forHealth Research (Portland, OR).

References

1. Bosley CM, Parry DT, Cochrane GM. Patient compliance with inhaled

medication: does combining beta-agonists with corticosteroids im-prove compliance? Eur Respir J 1994;7:504–509.

2. Sackett DL, Snow JC. The magnitude of compliance and non-compli-

ance. In: Haynes RB, Taylor WD, Sackett DL, editors. Compliancein health care. Baltimore: Johns Hopkins University Press; 1979. pp.11–22.

3. Gong H, Simmons MS, Clark VA, Tashkin DP. Metered-dose inhaler

usage in subjects with asthma: comparison of nebulizer chronolog anddaily diary recordings. J Allergy Clin Immunol 1988;82:5–10.

4. Baum D, Creer TL. Medication compliance in children with asthma.

J Asthma 1986;23:49–59.5. Coutts JA, Gibson NA, Paton JY. Measuring compliance with inhaled

medication in asthma. Arch Dis Child 1992;67:332–333.6. Milgrom H, Bender B, Ackerson L, Bowry P, Smith B, Rand C.

Noncompliance and treatment failure in children with asthma.J Allergy Clin Immunol 1996;98:1051–1057.

7. Gamble J, Stevenson M, McClean E, Heaney LG. The prevalence of

nonadherence in difficult asthma. Am J Respir Crit Care Med 2009;180:817–822.

8. National Asthma Education and Prevention Program. Expert panel

report-3: guidelines for the diagnosis and management of asthma.Bethesda, MD: National Institutes of Health, National Heart, Lung,and Blood Institute; 2007. NIH Publication No. 08-4051.

9. Gillissen A. Patients’ adherence in asthma. J Physiol Pharmacol 2007;58:

205–222.10. Friedman RH, Kazis LE, Jette A, Smith MB, Stollerman J, Torgerson J,

Carey K. A telecommunications system for monitoring and counsel-ing patients with hypertension: impact on medication adherence andblood pressure control. Am J Hypertens 1996;9:285–292.

11. Piette JD, Weinberger M, McPhee SJ, Mah CA, Kraemer FB, Crapo

LM. Do automated calls with nurse follow-up improve self-care andglycemic control among vulnerable patients with diabetes? Am JEmerg Med 2000;108:20–27.

12. Haynes RB, Ackloo E, Sahota N, McDonald HP, Yao X. Interventions

for enhancing medication adherence. Cochrane Database Syst Rev[serial on the Internet]. 2008 [accessed 2009 February 19];CD000011.Available from http://www.thecochranelibrary.

13. Osterberg L, Blaschke T. Adherence to medication. N Engl J Med 2005;

353:487–497.


43

14. Charles C, Gafni A, Whelen T. Decision-making in the physician–patient encounter: revisiting the shared treatment decision-makingmodel. Soc Sci Med 1999;49:651–661.

15. Charles C, Gafni A, Whelen T. Shared decision-making in the medicalencounter: what does it mean? (Or it takes at least two to tango). SocSci Med 1997;44:681–692.

16. Von Korff M, Katon W, Rutter C, Ludman E, Simon G, Lin E, Bush T.Effect on disability outcomes of a depression relapse preventionprogram. Psychosom Med 2003;65:938–943.

17. Ludman E, Katon W, Bush T, Rutter C, Lin E, Simon G, Bush T.Behavioural factors associated with symptom outcomes in a primary-care based depression prevention intervention trial. Psychol Med2003;33:1061–1070.

18. Joosten EA, DeFuentes-Merillas L, de Weert GH, Sensky T, van derStaak CP, de Jong CA. Systematic review of the effects of shareddecision-making on patient satisfaction, treatment adherence, andhealth status. Psychother Psychosom 2008;77:219–226.

19. Wilson SR, Strub P, Knowles SB, Buist AS, Huang Q, Nguyen M.Ethnicity, income, and education as potential modifiers of the effectsof shared treatment decision-making (SDM) between asthma caremanager and patient on asthma controller medication adherence: theBetter Outcomes of Asthma Treatment (BOAT) trial [abstract]. JAllergy Clin Immunol 2009;123:S72(265).

20. Wilson SR, Strub P, Buist AS, Vollmer W, Huang Q, Bocobo F, NguyenM. Does involving patients in treatment decisions yield long termimprovements in adherence to asthma controller medications andreduce albuterol use? The BOAT trial Year 2 follow-up [abstract].Proc Am Thorac Soc 2008;177:A232.

21. Wilson SR, Knowles S, Qian Y, Buist AS, Strub P, Lapidus J, Bocobo F,Nguyen M. Does involving patients in treatment decisions reduce useof asthma rescue medications [abstract]? Proc Am Thorac Soc 2007;175:A270.

22. Wilson SR, Strub P, Buist AS, Brown NL, Lapidus J, Luna V, VergheseS. Does involving patients in treatment decisions improve asthmacontroller medication adherence [abstract]? Proc Am Thorac Soc2006;3:A469.

23. Wilson SR, Buist AS, Holup J, Brown NL, Lapidus J, Luna V, VergheseS. Shared decision making vs. management by guidelines: impact onmedication regimen [abstract]. Proc Am Thorac Soc 2005;2:A907.

24. American Thoracic Society. Standardization of spirometry, 1994 update.Am J Respir Crit Care Med 1995;152:1107–1136.

25. Pocock S. Clinical trials: a practical approach. New York: Wiley; 1983.pp. 84–86.

26. Wilson SR, Scamagas P, German DF, Hughes GW, Lulla S, Coss S,Chardon L, Thomas RG, Starr-Schneidkraut N, Stancavage FB et al.A controlled trial of two forms of self-management education foradults with asthma. Am J Med 1993;94:564–576.

27. Buist AS, Vollmer WM, Wilson SR, Frazier EA, Hayward AD. A ran-domized clinical trial of peak flow versus symptom monitoring in olderadults with asthma. Am J Respir Crit Care Med 2006;174:1077–1087.

28. Rollnick S, Miller WR. What is motivational interviewing? Behav CognPsychother 1995;23:325–334.

29. National Asthma Education and Prevention Program. Expert panelreport-2: guidelines for the diagnosis and management of asthma.Bethesda, MD: National Institutes of Health, National Heart, Lung,and Blood Institute; 1991. NIH Publication No. 91-3642.

30. Steiner JF, Prochazka AV. Pharmacoepidemiology report: the assess-ment of refill compliance using pharmacy records: methods, validity,and applications. J Clin Epidemiol 1997;50:105–116.

31. Steiner JF, Koepsell TD, Fihn SD, Inui TS. A general method ofcompliance assessment using centralized pharmacy records: descrip-tion and validation. Med Care 1988;26:814–823.

32. Choo PW, Rand CS, Inui TS, Lee ML, Cain E, Cordeiro-Breault M,Canning C, Platt R. Validation of patient reports, automated phar-macy records, and pill counts with electronic monitoring of adherenceto antihypertensive therapy. Med Care 1999;37:846–857.

33. Schatz M, Nakahir R, Crawford W, Mendoza G, Mosen D, Stibolt TB.Asthma quality-of-care markers using administrative data. Chest2005;128:1968–1973.

34. Juniper EF, Guyatt GH, Cox FM, Ferrie PJ, King D. Development andvalidation of the mini-asthma quality of life questionnaire. Eur RespirJ 1999;14:32–38.

35. Vollmer WM, Markson LE, O’Connor E, Sanocki LL, Fitterman L, BergerM, Buist S. Association of asthma control with health care utilizationand quality of life. Am J Respir Crit Care Med 1999;160:1647–1652.

36. Hankinson JL, Odencrantz JR, Fredan KB. Spirometric reference valuesfrom a sample of the general US population. Am J Respir Crit CareMed 1999;159:178–187.

37. SAS Institute Inc. Software release 9.2. Cary, NC: SAS Institute Inc.;2006–2008.

38. Moher D, Schulz KF, Altman DG. for the CONSORT Group. TheCONSORT statement: revised recommendations for improving thequality of reports of parallel group randomized trials. BMC Med ResMethodol 2001;1:2.

39. Pocket Guide for Asthma Management and Prevention. Global initia-tive for asthma workshop report: global strategy for asthma manage-ment and prevention, 2004, NIH Publication No. 02-3659 [accessed2007 October 30]. Available from: http://www.ginasthma.org / Guide-lineitem.asp?l152&l251&intId590

40. Juniper EF, Guyatt GH, Willam A, Griffith LE. Determining a minimalimportant change in a disease-specific quality of life questionnaire.J Clin Epidemiol 1994;47:81–87.

41. Norman GR, Stratford P, Regehr G. Methodological problems in theretrospective computation of responsiveness to change: the lesson ofCronbach. J Clin Epidemiol 1997;50:869–879.

42. Guyatt GH, Juniper EF, Walter SD, Griffith LE, Goldstein RS.Interpreting treatment effects in randomized trials. BMJ 1998;316:690–693.

43. Korsch BM, Negrete VF. Doctor–patient communication. Sci Am 1972;227:66–74.


44

Online Data Supplement

Shared treatment decision-making improves adherence and outcomes in poorly controlled

asthma

Sandra R. Wilson, PhD

Peg Strub, MD

A. Sonia Buist, MD

Sarah B. Knowles, PhD, MPH

Philip W. Lavori, PhD

Jodi Lapidus, PhD

William M. Vollmer, PhD

and the BOAT Study Group

45

METHODS

The study has been reviewed and approved annually since 2002 by the IRBs of the Kaiser

Foundation Research Institute in Oakland, Calif., and of the Kaiser Permanente Center for Health

Research in Portland, Ore., and Honolulu, Hawaii. Randomization occurred from January 2003

to January 2005; final extraction of two years’ follow-up pharmacy dispensing data occurred in

mid-2007.

Patient Recruitment and Eligibility Criteria

More than 5,000 Kaiser Permanente (KP) members aged 18-70 years were identified at

Portland, Honolulu, and San Francisco, Oakland, and Richmond, Calif. KP clinical sites

(estimated aggregate patient population 3.8 million) through asthma registries as having some

evidence suggesting asthma that was not well controlled, specifically, pharmacy evidence of

overuse of rescue medications relative to asthma controllers, and/or a recent asthma-related

emergency department (ED) visit or hospitalization. Of the 4,285 that were contactable, 2,534

provided informed consent and were screened in person; the remaining 1,751 could not be

screened for multiple reasons including disinterest in participating. The in-person screening

yielded 1,170 who met the basic eligibility criteria. Six hundred and twelve ultimately also met

the study criteria for reversibility of airway obstruction and were randomized and completed the

study. See Figure 2 in the manuscript for the CONSORT diagram showing the detailed

recruitment process.

Exclusion criteria included intermittent asthma, recent asthma-related hospitalization

(past 3 months), primary diagnosis of chronic obstructive pulmonary disease or emphysema,

insufficient pulmonary function reversibility, regular use of oral corticosteroids, current

enrollment in a KP asthma care management program, pregnancy, hospitalization during the

1 46

preceding three months, inability to speak/understand English, use of a non-KP pharmacy,

expectation of dropping/losing KP membership during the study, or inability to meet baseline

data collection requirements.

Spirometry. To be fully eligible, a 12% increase in FEV1 pre- to post- bronchodilator

had to be demonstrated by all patients who did not use a controller medication regularly. Current

or ex-smokers who regularly used a controller were required to exhibit at least an 8% increase.

There was no reversibility criterion for never-smokers who regularly used a controller.

Informed Consent Procedures

All potentially eligible participants gave informed consent prior to the baseline

assessment and lung function testing. Brief operational descriptions of all three study arms were

provided in the informed consent document; the two experimental group descriptions specified

what the participants and care managers would do but did not use labels (e.g., shared decision

making) in these descriptions.

Randomization

A computer-based adaptive randomization algorithmE1 was used to ensure better than

chance balance, within site, among the three experimental groups. Five balancing variables were

used: age (18-34, 35-50, 51-70 years), gender, race/ethnicity (White/Caucasian, African-

American, Hispanic, Asian, Pacific Islander, American Indian/Alaska Native), hospitalization in

the two years preceding enrollment (yes/no), and frequency of asthma controller use in the past

week (none, 1-3, ≥4 days).

The assignment probabilities were set equal (1/3:1/3:1/3) for the first 15 randomizations

at each site and were equal thereafter whenever no net group assignment imbalance existed (after

summing the marginal imbalance scores across the five dimensions). When an imbalance did

2 47

exist such that there was one smallest group, the assignment probabilities were 6/8 for the

smallest and 1/8 for each of the other two. When the two smallest groups were tied, the

assignment probabilities were 3/8 for the two smallest and 2/8 for the largest group. These

probabilities tended to offset the direction of imbalance, following Efron’s rationale.E2

Allocation was concealed from staff randomizing patients. Randomization was

implemented by having a designated, non-blinded research staff member at the site enter the

relevant patient descriptors into the randomization module on the BOAT website, which

immediately performed the randomization, stored the result, and returned the patient’s study

assignment for implementation of the experimental assignment as indicated (e.g., referral to the

designated care manager). All other study personnel, with the exception of the care managers,

were blinded to patient’s study assignment.

Compensation

BOAT patients were not compensated by free access to medication, nor were they

provided with any financial incentive for participation in the intervention sessions. Patients were

provided with a modest incentive for completion of the study assessments at baseline and follow-

up. All KP Northern California (KPNC) patients received $30 if they completed both baseline

assessments; those at KP Portland and Hawaii received BOAT-logo hip/bum/fanny packs (per

differing policies of the respective IRBs about incentives for participation in research

assessments). Once randomized, KPNC patients received $35 for each follow-up assessment (at

the Year 1 and Year 2 follow-up visits) with a $25 bonus for completing both; KP Portland and

Hawaii patients received inexpensive BOAT-logo items (magnet, pens, etc.) for completing the

assessment at follow-up Year 1. Follow-up retention was high and comparable at all sites (see

Figure 2 in the manuscript for the CONSORT diagram).

3 48

Intervention Development

Intervention development had several critical goals: 1) to create two distinct care

management approaches – one representing an operational definition of “shared decision

making”E3, E4 (SDM) and the other an operational definition of “clinician decision making”

(CDM); 2) to give patients in both groups a comparable understanding of asthma and the basis of

its medical treatment; 3) to include specific activities in the SDM protocol to elicit the patient’s

goals and priorities, present the various possible treatment regimens and their relative merits (in

terms of control of inflammation, side effects, convenience, and cost to the patient), and ensure

that the patient and clinician discussed and compared the various regimens relative to the

patient’s priorities in order to arrive at a treatment decision to which both agreed; and 4) to train

care managers (interventionists) with diverse professional and clinical backgrounds and personal

styles to implement their assigned intervention (either SDM or CDM) in a consistent manner.

Prior to drafting the protocols, BOAT intervention development staff and co-investigators

observed patient encounters with asthma care managers already employed in the KP system in

order to determine how care management was being implemented per national and KP

guidelines, assess the variations among care managers, and determine the extent to which any

elements of structured shared decision-making were already a part of their approach. The BOAT

International Advisory Group also provided input on the shared decision model and on issues

that needed to be considered to operationalize this model. Drafts of the interventions were

prepared and reviewed by several care managers and study medical advisors (pulmonary, allergy,

or general internal medicine specialists). Care managers then pre-tested the interventions with

28 patients similar to those to be recruited into the clinical trial. These encounters were either

observed by developers or audio-taped with a subsequent review by the development team.

4 49

Multiple rounds of revision and pre-testing occurred until satisfactory versions of the scripts

were obtained.

Intervention Protocol

Format. The final SDM and CDM protocols were identical in format and in most

components (Figure 1 in the manuscript). Both involved two in-person visits with an asthma

care manager (Session 1, intended to take 50-60 minutes; Session 2, typically held one month

later, approximately 20-30 minutes), plus three brief follow-up phone calls at approximately 3-

month intervals. Both protocols involved:

o a detailed asthma history;

o education about asthma and assessment of the patient’s understanding;

o explanation of how asthma control can be objectively determined from symptoms and lung

function;

o comparison of this objective assessment of asthma control based on lung function test results

and reported symptoms (the dial as shown in Figure E-1) with the patient’s initial perception

of their asthma control (using the same dial without the objective symptom and lung function

criteria);

o explanation of what the patient’s current level of control, given the particular asthma

treatment regimen the patient has been implementing, implies about the severity of their

asthma (Figure E-2);

o explanation of severity and control as the basis for medical treatment and of the rationale for

an emphasis on controllers and minimal use of rescue medications;

o instruction on proper technique for use of inhalers/spacers;

5 50

o preparation of a written Asthma Management and Action Plan at the end of Session 1 based

upon the regimen selected;

o establishment of a follow-up plan to evaluate and adjust the regimen, as necessary, at Session

2 (after the patient had implemented the plan for approximately one month) and in

succeeding 3-, 6-, and 9-month follow-up telephone calls from the care manager to the

patient.

As highlighted in Figure 1 in the manuscript, in the SDM protocol the patient’s goals and

preferences regarding treatment, and any barriers they might confront in implementing treatment,

were specifically elicited and taken into account in negotiating the treatment regimen (see

Treatment decision process below). This was not done in the CDM protocol. The CDM care

manager’s goal was to prescribe treatment following medication guidelines and encourage the

patient to carry out that treatment.

The intervention protocols were in the form of “scripts” used by the care managers to

structure their interaction with the patient. They included the use of the following visual

aids/worksheets: 1) Patient Information Form for gathering data from the patient regarding

recent asthma symptoms, perceptions of control, medication use, alternative treatments used,

environmental triggers, and, in the case of the SDM intervention, patient goals for treatment and

priorities regarding factors that were important to the patient (control, convenience, side effects,

and cost); 2) Asthma Controllers/Relievers, a poster with pictures of actual medication delivery

devices to help patients identify their asthma medications; 3) How to Use Your Inhaler/Inhaler

Skills Checklist form (double-sided); 4) Your Lungs and Asthma, a poster with illustrations and

key teaching points; 5) How Well Controlled is Your Asthma, a pictorial dial with segments

defining each level of control (Figure E-1); 6) Asthma Severity Classification, a chart depicting

6 51

a 3-step process for assessing asthma severity in currently medicated patients, based on the

National Asthma Education and Prevention Program’s (NAEPP) Expert Panel Report-2

consensus guidelines.E5 Care managers subsequently used the patient’s asthma severity

classification to discuss treatment recommendations, per prescribing guidelines (Figure E-2); 7)

an Asthma Management and Action Plan filled out for the patient at the end of the session; and

8) four 1-page Weekly Asthma Diary forms.

The SDM intervention used one additional resource, the Medication Planner worksheet

(Figure E-3; see further description of its use in Treatment decision process below).

Treatment options. Identical treatment guidelines for asthma of different levels of

severity were given to all the care managers (both SDM and CDM). All treatment regimens

listed were consistent with the severity-based guidelines from the National Asthma Education

and Prevention Program’s Expert Panel Report-2 and the KP system-wide guidelines and

pharmacopeia. The various regimens were constructed from four different inhaled

corticosteroids (ICS) (beclomethasone, fluticasone, budesonide, and flunisolide), either alone or

in conjunction with another controller (a leukotriene modifier or theophylline) or a long acting

beta-agonist (LABA). Options that involved a LABA always included an ICS, either separately

or as a combined ICS-LABA preparation. For ICSs, different strengths and dosing regimen

options (number of puffs and frequency – once or twice daily) were included as appropriate to

each level of severity. Regimens roughly similar in the degree of control they offered were

clustered on the list, and the clusters were ordered from those offering most to those offering

least control. For treatment of mild persistent asthma, 11 treatment options were listed; for

moderate persistent asthma, 33 options; and for severe persistent asthma, seven options – a total

of 51 options altogether, encompassing most of the logical possibilities.

7 52

The treatment guidelines also included treatment recommendations for allergic rhinitis or

gastroesophageal reflux disease where relevant, and suggested the option of a prescription for

oral prednisone for the patient to have on hand in the event they experienced severe symptoms.

Additionally, all regimen options included a short acting beta-agonist (SABA) prescribed for use

as a rescue medication, as needed. For patients who reported exercise-induced asthma, the

regimen could include the use of a SABA prior to exercise.

Treatment decision process. The treatment guidelines described above were used in

different ways in the two intervention conditions. CDM care managers were given a form

containing guidelines for prescribing, with options grouped by the severity of asthma to which

they were appropriate. They explained to the patients that their treatment recommendation was

based on the level of control of the patient’s asthma in relation to their current regimen, but they

did not show the form to the patient or discuss alternatives regimens unless stimulated to do so

by the patient. They also explored barriers to the implementation of treatment and encouraged

the patient to adopt the prescribed regimen. The CDM care managers were instructed to answer

any patient questions, but neither encouraged nor discouraged questions about alternate

medications.

SDM care managers, on the other hand, discussed multiple treatment options with the

patient in light of the individual’s goals and preferences. Control of inflammation, convenience,

avoidance of side effects, cost, and any other considerations mentioned by the patient were

listed, in the order of their priority for the patient, in the first column of the Medication Planner.

The form presenting the full range of treatment options for asthma was shown to the patient.

These differed in the nature of the medication and medication combination options, range of

dosages, and schedule. Those options the patient wished to consider were entered at the top of

8 53

each succeeding column of the Planner. In all cases, the first treatment option listed on the

Planner was the de facto regimen, i.e., for the patient to make no change and to continue to use

or not use their current medications as they were. This very frequently meant a regimen that was

not appropriate to the severity of their asthma or its current level of control. At least two

alternative regimens, or as many as the patient wished to consider, were always considered from

among the full range of options available. The patient and clinician then compared the pros and

cons of each option in light of the patient’s goals and priorities, noting these pros and cons in the

appropriate cells of the Planner and discussing these until the patient and care manager came to a

joint decision.

It was apparent when a patient wanted to consider a treatment option that was not

consistent with guideline recommendations because of the way the options were categorized on

that form (by severity level and the nature of the medication(s), and the care manager would

have pointed this out, as appropriate, by indicating that the option provided less potential to

control the asthma than other options, or that it offered a greater level of control but also

potentially greater risk of side effects and/or less convenience (e.g., due to increased dosing

requirements for the same effect). Sharing such evaluative information was part of the

clinician’s professional contribution to the discussion. The patient’s contribution was in sharing

their own goals and preferences and evaluating the information provided. If a patient chose a

potentially very inadequate or dangerous regimen, a procedure was in place to have them sign a

form indicating they were aware of this, but this was rarely necessary.

Patients did not receive free access to medication. All prescribed medications were based

on the KP pharmacopeia and costs were discussed per the individual patient’s KP membership

plan (e.g., medication co-pay), which varied depending on their plan’s medication benefits.

9 54

During the session, the patient’s costs for different regimens could be accessed electronically by

the clinician.

Further information on the SDM and CDM protocols is available from the corresponding

author.

Interventionist Training

To be eligible to be a care manager, individuals had to be licensed, non-physician health

professionals. BOAT recruited nurses (n=3), respiratory therapists (n=3), and pharmacists (n=7),

nurse practitioners (n=1) and physician assistants (n=2), most of whom also served as care

managers within the KP system, for total of 16 BOAT interventionists. Care managers’

professional discipline was not related to the number of patients they saw, and patient

assignment to a specific care manager, within site and experimental group (SDM or CDM), was

unrelated to either patient or care manager characteristics: 38 patients saw care managers who

were nurses, 121 saw respiratory therapists, 114 saw pharmacists, 38 saw the nurse practitioner,

and 64 saw a physician assistant.

Assignment of care managers to the SDM or CDM conditions was done purposively to

achieve the best possible overall balance between the two conditions in terms of disciplinary

backgrounds and amount of clinical experience. Care managers assigned to SDM and those

assigned to CDM were trained separately and saw only the protocol for their assigned condition.

Assignments of care managers to SDM or CDM were not strictly random because of the inherent

limitations on the availability of appropriately credentialed clinicians who had requisite

experience in adult asthma management (including in most cases, experience in asthma care

management) and who were employed and available for the amount of time required to see the

numbers of study participants at a given clinical site. Initially, five SDM and five CDM care

10 55

managers were trained. The initial 1-day small group training session included a general

introduction to the BOAT study and an overview of the assigned intervention protocol; practice

of the successive segments of the intervention (including demonstrations by the development

team); and instruction on the use of spirometry results in the classification of asthma severity

(ASB). This was followed by individual trainee practice with real or simulated patients,

observation and subsequent debriefing by the development team; and instructions on the plan

and procedures for pilot tests with four real patients in their home clinics. Each session with

these four patients was tape recorded and reviewed against a checklist by the primary

intervention developer/QC evaluator. Detailed feedback was given to each care manager on each

session with a practice patient. After four initial patients, and when performance was deemed

acceptable, the care manager was certified as ready to see study participants.

Subsequently, intervention QC procedures were implemented, with a review of

audiotapes and ongoing individual feedback to the care managers. Monthly conference calls

were held with the care managers assigned to a given experimental condition to discuss their

experiences, provide reinforcement, identify any general problems or situations not explicitly

covered in the existing protocols, and develop common solutions, which were subsequently

documented by issuing formal supplements to the SDM and/or CDM protocols, as appropriate.

A total of 11 supplements were issued. Some supplements were required when special patient

circumstances were encountered for the first time and others were occasioned by external factors

such as changes to the system’s formulary, medication availability, or release of new information

about potential side effects. No supplements were required after the initial months of

intervention implementation.

11 56

Personnel changes and the addition of a third and fourth clinical site in Northern

California required training six additional care managers individually. The pattern of training

and the training activities were essentially the same as described above. In total, six SDM and

10 CDM care managers were trained. The larger number of CDM than SDM care managers was

due to the additional sites as well as unanticipated staff turnover (position changes within KP,

spouse job relocation, family medical leave, etc.). In cases where care managers had limitations

on their scope of practice, arrangements were made to have proposed prescriptions reviewed and

signed by a KP study physician.

Communication with primary care provider (PCP). Communication and coordination

of care between the care manager and patients’ PCPs throughout the BOAT trial was facilitated

by the fact that all clinical contacts at KP, including those with the care managers, are

documented in the patient’s chart and the basic facts regarding the patient’s regimen and clinical

encounters were available in KP’s electronic database and accessible to the patient’s physicians

and other authorized individuals involved in the patient’s care.

There was no intent that the patient’s physicians be fully blinded to intervention

assignment, nor obviously could the care manager be blinded to the assignment of those patients

assigned to him/her. Care managers were blinded to both the participation and study

assignments of other patients, however. The patient’s physicians could see documentation of

care management encounters in the patient’s record if and when they had occasion to view that

record during the trial. They were not specifically informed which patients were enrolled or

which had been assigned was to the SDM or CDM arms, or whether a given care manager was

using the SDM or CDM protocol, but it is possible that they became aware of this if they had any

question about the basis for a proposed treatment change for which the care manager was

12 57

seeking approval. Both SDM and CDM care managers sought approval for such changes, when

required by the nature of the change and what was permitted by the individual care manager’s

professional scope of practice with regard to prescribing new medications or adjusting dosages

per protocol.

Usual Care Control Condition

Usual asthma care at the BOAT KP medical centers was based on a stepped-care

approach to pharmacotherapy with the aim of long-term asthma control, as described by the

NAEPP’s Expert Panel Report-2.E5 Specific prescribing guidelines are also based on the NAEPP

recommendations as well as the KP formulary.

As part of usual asthma care, KP physicians have the option to refer patients for

specialized care by an allergist or pulmonary specialist and/or to refer patients to an adult asthma

care management program, but the availability at specific sites varies and is subject to resource

limitations. Consequently, the asthma care management program is not an integral part of the

medical consultations occurring in the course of usual care. Where available, the program is

based on physician-referral only and is an intensive 6-month program in which a non-physician

advanced health care professional (usually a respiratory therapist, nurse, or clinical pharmacist;

not a physician) provides education about asthma and works with the patient to improve

treatment adherence, and identify and resolve any other barriers to attaining asthma control.

Care managers are trained to provide asthma education and to use motivational interviewing and

other strategies to identify and address barriers to regimen adherence. They can adjust treatment

by protocol within KP asthma guidelines and in consultation with the patient’s physician.

All care managers – both in the existing KP program and BOAT interventions –

documented each visit and phone contact in the patient’s chart, where it was available to the

13 58

patient’s physician. The care managers discussed their recommendations with the physician if

they or the physician had any questions about the patient’s new regimen. For those care

managers whose scope of practice did not allow independent prescribing, the physician approved

and wrote the prescription.

Intervention Quality Control (QC) Procedures

Session 1 and Session 2 were audio-taped for approximately 10% of each care manager’s

patients (the first four and, subsequently, every ninth patient). Ratings based on these taped

sessions were used for ongoing QC and feedback to the care managers throughout the trial, but

especially during the initial learning period. One study group co-author (NLB) evaluated all

taped sessions. A second QC rater evaluated 32 randomly selected Session 1 tapes – four from

each of the eight care managers (four SDM and four CDM) who saw the greatest number of

study patients. For both protocols, the Adherence to Protocol Summary Score and Decision

Roles Score scales were used to evaluate protocol adherence and the patient’s and care

manager’s roles in the treatment decision. The raters scored each protocol element, and gave an

overall adherence summary rating for the five major protocol sections (Set the Stage, Gather

Information, Provide Education, Discuss Treatment, and Wrap Up).

The Adherence to Protocol Summary Score (APSS) was calculated as the mean of the

five summary ratings for the major sections of the Session 1 protocol. Each section summary

rating scale was anchored at the end points: 1=‘Inadequate’ and 5=‘Word for word (without

reading).’ If both protocols were delivered as intended, mean scores across care managers would

be approximately 5 with no group difference.

The Decision Roles Score reflected the relative contribution of care manager and patient

to treatment decisions. After each patient encounter, both the care manager and patient rated

14 59

their individual role on a five point scale: [Decision made] 1= ‘By the care manager alone,’ 2=

‘Mostly by the care manager,’ 3= ‘By the patient and care manager equally,’ 4= ‘Mostly by the

patient,’ or 5= ‘By the patient alone.’ If the interventions were delivered as intended, the SDM

and CDM groups would have systematically different decision role ratings – with SDM

encounters being rated ≥3 and CDM encounters rated between 1 and 2.

Evaluation by care manager. Following each session, all care managers entered into the

BOAT data collection website any regimen changes or new treatment decisions reached, as well

as their evaluation of their own and the patient’s relative contribution to those decisions.

Evaluation by patient. At the close of each session, the patient was given a pre-

addressed, stamped postcard to mail to the local research coordinating center with a rating of

his/her own and the care manager’s relative contribution to the treatment decisions. The

postcard asked “Who made the decisions in your meeting with the care manager about what your

asthma treatment would be” and answers were recorded on a 5-point scale from 5=“I alone made

the decision” to 3= “The care manager and I participated equally in making the decision,” to

1=“The care manager alone made the decision.” The postcard contained only the participant’s

study ID number, no identifying information, and was not seen by the care manager.

Further validation of the QC rater’s scoring. The purpose of the QC validation by the

second rater was to determine the general extent to which the ratings were consistent with those

of the primary QC rater with regard to the SDM and CDM care managers’ adherence to their

respective protocols and the relative contribution of the care manager and patient to treatment

decisions in SDM vs. CDM encounters.

Results of intervention QC procedures

15 60

Protocol adherence. The protocol adherence scores assigned by the QC rater were high

(SDM=4.0, CDM=3.9) and did not differ significantly between the two groups (P=0.47). A

value near 4.0 indicates that all of the intended topics were covered, virtually all of them

completely and thoroughly, that the interventionist followed the intervention script closely and

covered topics in the prescribed sequence, that the handouts and other teaching aides/worksheets

were used appropriately, and that accurate information was given in response to participant

questions.

Roles of patients and care managers in treatment decisions. The QC evaluator’s ratings

of the audio-taped encounters provided evidence that both the SDM and CDM protocols were

delivered as intended. Patients, care managers, and the evaluator all accorded the patient a

significantly greater influence on the treatment decisions in SDM than CDM sessions; all

p<0.0001). The mean decision role scores for SDM patients, as assigned by care managers, the

patients, and the QC evaluator, all were ≥ 3.0, meaning that the SDM patients were consistently

viewed as having an equal or somewhat greater influence on treatment selection than the care

managers, with neither having a dominant influence on the decision. In contrast, for CDM

sessions, all three felt that the patient had less influence on the treatment decisions than did the

care manager (mean rating=2.5).

Further validation of the QC rater’s scoring. The association between the QC and a

second raters’ evaluations of protocol adherence was moderately strong (SDM Spearman

Rho=0.50, CDM Rho=0.55; combined groups Rho=0.48, P=0.005); for their decision role

ratings, the combined association was very strong (SDM Rho=0.36, CDM Rho=0.27; combined

Rho=0.79, P< 0.0001). A correlation coefficient was the appropriate metric since the APSS is a

quasi-continuous score (i.e., the average of five 5-point ratings). The fact that the correlation

16 61

using the ratings from both SDM and CDM sessions was substantially larger than the individual

group correlations was due to the much larger variance in decision role ratings when the two

experimental groups were combined, which reflects the intended differences between the SDM

and CDM groups.

Data Collection

Baseline and Year 1 follow-up data for all patients were obtained by patient interview,

lung function testing, and data extraction from KP pharmacy and health care utilization records.

Pharmacy dispensing data were available at all sites for follow-up Year 2. However, due to

resource limitations, it was not possible to extract health care utilization data for all patients for

the follow-up Year 2, since in Hawaii, this extraction required manual chart review.

Patient Characteristics

The patient questionnaire was administered via interview at the pre-randomization and

Year 1 follow-up clinic visits. Trained interviewers administered all questionnaires for the study

to avoid problems associated with low literacy. It collected information on sex, age, basic

demographic characteristics, currently prescribed asthma medications, including their prescribed

dosages and schedules, the patient’s use of their prescribed daily controller medication, daytime

and nocturnal asthma symptoms, and whether or not the patient had been hospitalized in the past

year.

Asthma-related quality of life. Asthma-related quality of life was assessed using the 5-

item symptom sub-scale of the Juniper Mini Asthma Quality of Life Questionnaire (MAQLQ) at

pre-randomization and 12-months post-randomization. The symptom sub-scale included:

shortness of breath, bothered by coughing, chest tightness or heaviness, difficulty sleeping, and

17 62

chest wheeze. Each item on the MAQLQ was asked by the interviewer relative to the last two

weeks, with response choices ranging from 1=maximum impairment to 7=no impairment.

It has previously been asserted by the developers of the AQLQ that a group means

difference of > 0.50 points is the minimal difference in order to conclude that a clinical benefit

has been demonstrated. However, more recent research suggests that the methodology by which

the minimal clinically important difference was determined is subject to serious question, and

that smaller group differences may indeed be clinically important.E6, Members of the

development team have also modified their approach to focus on the magnitude of changes at the

individual patient level, and hence the proportion of patients whose scale scores changed by

more than a specific criterion amount.E7

Asthma control. Asthma control was self-reported using the Asthma Therapy

Assessment Questionnaire (ATAQ), a four-item questionnaire that assesses level of asthma

control in the past four weeks. Interviewers asked respondents to rate whether they had or did

not have problems on each of four dimensions (self-perception of asthma control, nighttime

awakening due to asthma symptoms, lost time at work due to asthma, and use of asthma reliever

medications). The total number of problems reported yielded scores between 0=no control

problems and 4=four control problems. At follow-up, few participants had ATAQ scores over 1,

so the 4-point scale was collapsed into binary form corresponding to an ATAQ score of 0 versus

scores of 1 through 4.

Asthma regimen and severity and control classification. During the pre-randomization

clinic visits, each patient’s regimen (medications and the dosages and schedules on which they

were actually being taken) was classified as to the asthma severity for which that regimen would

be appropriately prescribed (see Figure E-2), per the Global Initiative on Asthma guidelines.

18 63

This regimen was termed the de facto regimen and was not necessarily – and in many instances

was clearly not – the regimen as prescribed by the patient’s physician. For example, a patient

who only used rescue medications and not his prescribed daily controller would be classified as

having a de facto regimen appropriate to mild intermittent asthma.

In a small proportion of cases, patients reported taking medications on a regimen or

schedule that did not exactly match any of the standard guidelines-recommended regimens (e.g.,

taking fewer puffs per day of an ICS than would be appropriate for a patient with Moderate

Persistent asthma, but also taking a LABA) or else the strength of the preparation was missing.

In these cases, an appropriate intermediate regimen was defined (e.g., a regimen between mild

and moderate or between moderate and severe).

The criteria and algorithm for determining the level of asthma control were based on the

reported frequency of nighttime and daytime symptoms and the FEV1 percent predicted value

from baseline spirometry (see Figure E-1). When the level of control implied by participant’s

symptoms differed from the level of control implied by the FEV1, the lower of the two levels of

control was used, as is recommended in the guidelines.

Lung Function

Lung function was assessed at the pre-randomization and Year 1 follow-up clinic visits

per standardized research spirometry methodsE8,E9,E10 using equipment that met or exceeded

American Thoracic Society requirements (SensorMedic 1022 or OMI 2000, 8 liter dry rolling

seal spirometer with a digital electronic assembly). This system employs software that

determines the acceptability and reproducibility of each maneuver during the testing session and

alerts the technician to possible problems (e.g., slow initiation of the maneuver, early cessation

of airflow, coughing). Spirometry was performed before and 20 minutes after four puffs (200µg)

19 64

of albuterol, a SABA. Pulmonary function parameters included: 1) forced expiratory volume in

1 second (FEV1) expressed as percent predicted relative to age, gender, using race-specific

norms;E11 and 2) the ratio of FEV1 to forced expiratory volume in 6 seconds (FEV6), expressed

as a percentage ( FEV1/FEV6).

Technicians received two days of small group or individual training consisting of basic

didactic instruction, demonstrations, and observed practice until they could perform acceptable

maneuvers. At their home sites they performed additional tests on volunteers. These data were

transmitted to the pulmonary function reading center at LDS Hospital in Salt Lake City, UT (Dr.

R. Jensen), for evaluation. Research assistants who met criterion performance in terms of the

quality of their maneuvers were certified to perform study spirometry.

Spirometry quality was monitored by the pulmonary function reading center at the LDS

Hospital in Salt Lake City, UT. Each maneuver from each participant session was graded (A=

4.0, B=3.0, etc.), and the best, reproducible maneuver for each participant was used in the

analysis. Scores were summarized by technician across participants and provided the basis for

additional training, correction or reinforcement as needed.

Verifying KP membership

KP membership was verified for each patient for each study year to ensure that the

estimates of medication adherence and health care utilization for each year were accurate. A

comparison of KP membership records with pharmacy dispensing and medical visit records

revealed that, even after eliminating apparent “gaps” in membership of two months or less, most

patients who still appeared to have membership during the BOAT study period were still

receiving medical care and filling prescriptions at KP facilities, suggesting some unreliability in

the electronic membership records. Furthermore, using KP electronic records, a sensitivity

20 65

analysis was conducted in which we compared the results of the multivariate analytic model

predicting the groups’ mean CMA values for patients with at least 11 months apparent

membership with the results when the small proportion of patients with less than 11 months

membership was included. The difference in results was negligible. Consequently, a decision

was made to consider the full 365-day observation period for the pre-randomization and follow-

up year(s) for all patients when calculating medication adherence, regimen potency, and health

care utilization rates.

Intervention Process and QC Outcomes

Data on the SDM and CDM patients’ asthma regimens as a result of the intervention

were available from their Asthma Management and Action Plan. Information from the care

managers’ Post-Session Report forms, completed after each in-person encounter and follow-up

phone call, included the starting and ending times of the encounter, any additions, deletions, or

other changes made in the patient’s regimen during the session, and the care manager’s rating of

the relative contributions made by the care manager and the patient to the treatment decisions. A

similar rating was requested of the patient using the postcard described earlier.

21 66

REFERENCES

E1. Pocock S. Clinical trials: a practical approach. New York: Wiley; 1983;84-86.

E2. Efron B. Forcing sequential experiments to be balanced. Biometrika 1971;58:403-417.

E3. Charles C, Gafni A, Whelen T. Decision-making in the physician-patient encounter:

revisiting the shared treatment decision-making model. Soc Sci Med 1999;49:651-661.

E4. Charles C, Gafni A, Whelen T. Shared decision-making in the medical encounter: what

does it mean? (Or it takes at least two to tango). Soc Sci Med 1997;44:681-692.

E5. National Asthma Education and Prevention Program. Expert Panel Report-2: Guidelines

for the diagnosis and management of asthma. National Institute of Health, National

Heart, Lung and Blood Institute, NIH Publication No. 91-3642, 1991.

E6. Norman GR, Stratford P, Regehr G. Methodological problems in the retrospective

computation of responsiveness to change: the lesson of Cronbach. J Clin Epidemiol

1997;50:869-879.

E7. Guyatt GH, Juniper EF, Walter SD, Griffith LE, Goldstein RS. Interpreting treatment

effects in randomized trials. BMJ 1998;316:690-693.

E8. American Thoracic Society. Standardization of Spirometry: 1987 update. Am Rev Respir

Dis 1987;136: 1285-98.

E9. American Thoracic Society, Standardization of Spirometry, 1994 update. Am J Respir

Crit Care Med 1995;152:1107-36.

E10. Miller MR, Hankinson J, Brusasco V, Burgos F, Casaburi R, Coates A, Crapo R, Enright

P, van der Grinten CPM, Gustafsson P, Jensen R, Johnson DC, MacIntyre N, McKay R,

Navajas D, Pedersen OF, Pellegrino R, Viegi G, Wanger J. Standardization of

Spirometry. Eur Respir J 2005:26;319-38.

22 67

E11. Hankinson JL, Odencrantz JR, Fedan KB. Spirometric reference values from a sample of

the general U.S. population. Am J Respir Crit Care Med 1999;159:178-87.

23 68

Figure E1

24 69

Figure E2

25 70

Figure E3

26 71

Comparative effectiveness of asthma interventionswithin a practice based research networkTapp et al.

Tapp et al. BMC Health Services Research 2011, 11:188http://www.biomedcentral.com/1472-6963/11/188 (16 August 2011)

72

STUDY PROTOCOL Open Access

Comparative effectiveness of asthma interventionswithin a practice based research networkHazel Tapp1, Lisa Hebert2* and Michael Dulin1

Abstract

Background: Asthma is a chronic lung disease that affects more than 23 million people in the United States,including 7 million children. Asthma is a difficult to manage chronic condition associated with disparities in healthoutcomes, poor medical compliance, and high healthcare costs. The research network coordinating this projectincludes hospitals, urgent care centers, and outpatient clinics within Carolinas Healthcare System that share acommon electronic medical record and billing system allowing for rapid collection of clinical and demographicdata. This study investigates the impact of three interventions on clinical outcomes for patients with asthma.Interventions are: an integrated approach to care that incorporates asthma management based on the chroniccare model; a shared decision making intervention for asthma patients in underserved or disadvantagedpopulations; and a school based care approach that examines the efficacy of school-based programs to impactasthma outcomes including effectiveness of linkages between schools and the healthcare providers.

Methods/Design: This study will include 95 Practices, 171 schools, and over 30,000 asthmatic patients. Five groups(A-E) will be evaluated to determine the effectiveness of three interventions. Group A is the usual care controlgroup without electronic medical record (EMR). Group B practices are a second control group that has an EMRwith decision support, asthma action plans, and population reports at baseline. A time delay design during yearone converts practices in Group B to group C after receiving the integrated approach to care intervention. Fourpractices within Group C will receive the shared decision making intervention (and become group D). Group E willreceive a school based care intervention through case management within the schools. A centralized database willbe created with the goal of facilitating comparative effectiveness research on asthma outcomes specifically for thisstudy. Patient and community level analysis will include results from patient surveys, focus groups, and asthmapatient density mapping. Community variables such as income and housing density will be mapped forcomparison. Outcomes to be measured are reduced hospitalizations and emergency department visits; improvedadherence to medication; improved quality of life; reduced school absenteeism; improved self-efficacy andimproved school performance.

Discussion: Identifying new mechanisms that improve the delivery of asthma care is an important step towardsadvancing patient outcomes, avoiding preventable Emergency Department visits and hospitalizations, whilesimultaneously reducing overall healthcare costs.

Keywords: asthma, comparative effectiveness research, shared decision making, integrated approach to care

BackgroundSignificanceAsthma is a chronic lung disease that affects more than23 million people in the United States, includingapproximately 7 million children [1,2] The burden of

asthma in the U.S. is high, accounting annually for 2million emergency department visits, 504,000 hospitali-zations, 13.6 million physician office visits, and over4,200 deaths while resulting in $15 billion in directmedical costs [3-5]In North Carolina during 2007, the lifetime prevalence

of asthma in adults was 12.1%, impacting over 800,000individuals [6] The prevalence of asthma was dispropor-tionately higher in African American and Native

* Correspondence: [email protected] Physicians Network, Carolinas HealthCare System, PO Box 32861,Charlotte, NC 28232, USAFull list of author information is available at the end of the article

Tapp et al. BMC Health Services Research 2011, 11:188http://www.biomedcentral.com/1472-6963/11/188

© 2011 Tapp et al; licensee BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative CommonsAttribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction inany medium, provided the original work is properly cited.

73

mailto:[email protected]

http://creativecommons.org/licenses/by/2.0

American populations as well as low income popula-tions [7] Asthma prevalence for the state also reached17.8% in children under the age of 17 [7] Unfortunately,not only is asthma prevalence increasing in the Caroli-nas, but many patients suffering with asthma lack ade-quate control of their symptoms. Indeed, over 50% ofadults noted asthma symptoms more than once perweek and 21% had daily symptoms [7] This resulted inone-third of adults losing at least one day of workbecause of their asthma over the prior 12 months, and24% of asthma patients having at least one EmergencyDepartment (ED) or urgent care visit related to theirasthma during this same period of time [7] Rates of EDutilization for asthma management were also almost200% higher for minority children than non-minoritychildren.Our state’s asthmatic patients also frequently lack a

usual source of care, and 45% of asthmatics went with-out a visit to their regular physician over the past year[7] Patients with asthma also report a lower quality oflife, with 18% of all asthmatic patients rating their healthoverall as poor [6] Hospitalization rates are also higherfor asthmatic patients and a significant health expensefor the state. Over $88 million was spent on hospitaliza-tions for asthma in 2004 costing on average $8,259 perhospital stay.The 2009 Institute of Medicine (IOM) report identi-

fied two priorities in the need for comparative effec-tiveness research on asthma including the need tostudy an integrated approach to care and shared deci-sion making [8] The agency for healthcare quality andresearch (AHRQ) has also placed asthma on the prior-ity list of conditions for comparative effectivenessresearch with particular interest in impacting popula-tions that are low-income, minority groups, women,children, elderly individuals, and individuals with spe-cial health care needs, such as those who live in inner-city and rural areas. This attention to asthma is relatedto the burden of the disease on the U.S. population,disparities on outcomes for asthmatic patients, and thelack of knowledge of how to improve adherence tomedications and subsequently improve asthma out-comes [9-17] Gaps in our knowledge of the optimalmedical management and predictors of asthma out-comes (including environmental triggers) are also pre-sent and need to be addressed by comparativeeffectiveness studies [18-24].Potential solutions include more comprehensive

asthma management strategies that build upon existingsuccesses in the development of integrated care systems;special emphasis on targeting high-risk populations; theuse of self-management and shared decision makingapproaches to care; and school-based interventions thatcan be linked with primary care providers.

The National Asthma Education and Prevention Pro-gram (NAEPP) asthma guidelines emphasize that thegoal for optimal asthma control is to reduce bothimpairment and risk, and recommend a stepwiseapproach to pharmacotherapy [25,26]. Measurement ofasthma control can be complex, encompassing physicalexamination, objective tests, and patient history [26].Well-controlled asthma is characterized by: experienceof symptoms and use of a rescue medication twice aweek or less, no early morning or nighttime awakenings,no limitations on activities of daily living, normal forcedexpiratory volume in 1 second (FEV1) or peak expiratoryflow (PEF) test results, and controlled asthma as deter-mined by physician and patient assessments [27].Prevention and reduction of asthma exacerbations are

key to reducing risk associated with asthma. NAEPPguidelines define asthma exacerbation as an “acute orsub-acute episode of progressively worsening shortnessof breath, cough, wheezing, and chest tightness–or somecombination of these symptoms” [25] Although mostasthma exacerbations are treated in an outpatient set-ting, they present considerable difficulty for patients,including increased healthcare utilization, lost work pro-ductivity, school absences and increased healthcare costs[28].One individualized behavioral approach to asthma dis-

ease management is the use of an asthma action plan,developed through asthma education activities toincrease a patient’s knowledge and skills with regard toasthma control [25,29-31] The National Institute ofAllergy and Infectious Diseases (NIAID) inner-cityasthma studies program found that a home-basedapproach tailored to an individual child’s specific riskfactors provided a more effective intervention strategy,especially with the inclusion of a written asthma actionplan [32,33].

Methods/DesignThis study received ethics approval from the Institu-tional Review Board of Carolinas HealthCare System.

Description of all interventionsThis study will include 95 Practices, 171 schools, andover 30,700 asthmatic patients. The interventions to becompared will include (Figure 1, Table 1): Group A:control practices providing usual care; Group B: controlpractices with a centralized electronic medical record(EMR), decision support tools, and population manage-ment tools; Group C: intervention practices have all thetools of Group B, but use an integrated system based onthe Chronic Care Model to improve asthma outcomes;Group D: intervention practices will have all tools avail-able to Group C with the addition of implementation ofa shared decision making approach to care; and Group


Page 2 of 10

74

E: patients will receive additional care managementthrough school-based nurses and care managers.Three main interventions are planned:

1) An integrated approach to care (IAC) for asthmamanagementHere our objective is to compare the effectiveness of anintegrated approach to asthma management based onthe chronic care model (CCM) with a non-integratedepisodic care model (usual care control). This modelbased on the CCM, includes decision support, an elec-tronic Asthma Action Plan, population management

tools, training in the practice redesign and rapid cycleprocess improvement by a quality improvement coach,and linkages to community resources. This approachwill be compared to control practices (A) and practicesthat receive the tools without training with a coach (B).2) A Shared Decision Making (SDM) approachThe objective is to compare the effectiveness of a com-bined shared decision making and IAC approach (C)with the integrated approach alone and with usual care(control). This model is based on the successful inter-vention developed by Wilson and colleagues that

Figure 1 Study Design. This study will include 95 Practices, 171 schools, and over 30,700 asthmatic patients. Five groups (A-E) will be evaluatedto determine the effectiveness of each intervention. Group A is the usual care control group. Group B practices will provide a second controlgroup with an EMR with decision support, Asthma action plans, and Population reports (EAP) at baseline. A time delay design will be used as allpractices in Group B will receive the Integrated Approach to Care (IAC) intervention with 10-12 practices receiving the intervention every 2months over a 1 year period of time. Four practices within Group C will receive the Shared Decision Making (SDM) approach intervention with 1practice receiving the intervention every 4 months over a 2 year period. Group E will receive School Based Care (SBC) through casemanagement within the schools.


Page 3 of 10

75

changes the dialog between patients and physicians topositively impact patient medication compliance andasthma self-management. This model will be implemen-ted in select practices that have already implementedthe IAC approach to determine if additional benefit canbe gained using SDM. Practices receiving the SDMintervention also primarily serve disadvantaged popula-tions including the uninsured, Medicaid patients, dis-abled Medicare patients, and minority groups.3) A School-Based Care (SBC) ApproachHere our objective is to compare the effectiveness of aschool-based approach to care with the integratedapproach to care and with and without the shared deci-sion making approach to care and against the control.This intervention will provide an electronic data capturesystem to a robust CDC funded school-based interven-tion to assist with evaluation and to link the school-based care team with primary care providers. This sys-tem has identified almost 8,500 asthmatic children whowill be a part of this intervention. Outcomes for chil-dren will be compared to children in control practicesas well as practices in the IAC and SDM approach.Clinical OutcomesThe major clinical outcomes goals are: reduced hospita-lizations and emergency department visits; improvedadherence to medication; improve quality of life;reduced school absenteeism; improved self-efficacy andimproved school performance. Associated quality goalsare to improve percentage of patients with persistentasthma who are prescribed a controller medication; andthose with asthma who receive a flu vaccination. Qualitygoals will be measured by the percentage of patientswho reach the goal out of all patients identified with thedisease. The three interventions that will be undertakeneach have the potential to impact these goals. The datacollected for the evaluation of the three approaches toasthma management will also be leveraged to identifyother important variables that impact asthma outcomesincluding: pharmaceutical management, co-morbidities(Tobacco Use/Exposure, GERD, Allergic Rhinitis,

Obesity, Sleep Apnea), patient demographics (age, sex,race/ethnicity, insurance status), and community levelvariables (neighborhood quality of life, build environ-ment and transportation elements, pollution sources,and housing density).

SettingThe research network coordinating this project (TheMecklenburg Area Partnership for Primary CareResearch, MAPPR) includes the hospitals, urgent carecenters, and outpatient clinics within Carolinas Health-care System (CHS) that share a common ElectronicMedical Record (EMR) and billing system allowing forrapid collection of clinical and demographic data (Figure2). This vertically integrated hospital system providescare to over 1.2 million patients including over 38,000patients with a diagnosis of asthma. Between 2008 and2009, these patients were responsible for almost 17,500hospitalizations and 68,000 clinic visits. The researchnetwork includes a group of ambulatory clinics thattogether provide over 85% of care to the uninsured

County Containing CHS Facility

North Carolina

South Carolina

Figure 2 Map of North and South Carolina Showing the Scopeof Carolinas Healthcare System (CHS). Each county shaded ingreen is home to a CHS facility including the 32 hospitals and 95clinics that will take part in this project.

Table 1 Asthmatic Patients Seen in the CHS System 2008

Number of Patients Number of Clinic Visits

Total Number of Unique Patients with an Asthma Diagnosis 38,634 77,582

African American Race 14,168 34,551

Hispanic Ethnicity 2,043 4,596

Age < 18 11,058 21,357

Number within Mecklenburg County 16,458 41,961

Uninsured, Medicaid, or Medicare 13,564 31,022

Number of Hospitalizations for Asthma 10,321 NA

Number Emergency Room Visits 30,121 NA

CHS School Children with Asthma 8,500 NA


Page 4 of 10

76

patients within the surrounding community. Inclusion ofthe 92,000 patients who receive care within these clinicsallow the evaluation of the effectiveness of interventionsfor asthma management within a large population ofpoor and/or underserved community members. TheMAPPR network includes the Mecklenburg CountyHealth Department and the county school system(Charlotte Mecklenburg Schools, CMS). Indeed, CHSemploys the 121 school nurses who assist with manage-ment of children with asthma in the school setting.

Development of the InterventionThe approach is to implement the three separate butinterrelated interventions across the community. Theprimary intervention, the integrated approach to asthmamanagement, has been implemented in 10 practices,and through this project will be deployed in 65 addi-tional practices. In addition, an active federally fundedschool-based system of care will be enhanced throughthis project allowing improved evaluation and compari-son of outcomes between the school’s system of careand other intervention as well as allowing the creationof a school to healthcare provider link.Database DevelopmentThe overall strategy for this study is to create a centra-lized database for evaluation of comparative effective-ness of five different groups for management of asthmapatients. The new database will draw information fromthe Carolinas Healthcare Systems billing and clinicaldatabases, Medicaid claims data, school system data,chart abstraction, community level datasets, patient sur-veys and focus groups.A centralized database will be created with the goal of

facilitating comparative effectiveness research on asthmaoutcomes specifically for this study. The databasenamed ACER (Asthma Comparative EffectivenessResearch Database) will be setup using Microsoft SQLServer 2005 (Redmond, WA) and designed to supportSAS (SAS Institute, Cary, NC) that is the standard ana-lysis software used within the institution. The data willflow from 8 different sources including: the healthcaresystem’s billing data, the healthcare system’s clinicaldata, school data from the Institute for Social Capital,school nurse data, Medicaid claims data, patient leveldata (from surveys and focus groups), and communitylevel data from the Center for Metropolitan Studies(Figure 3).Any patient that has had a diagnosis of asthma

recorded for billing purposes at any visit throughout thesystem since 2008 will be identified for the study as hav-ing asthma. This system recorded 38,634 uniquepatients for 2008 resulting in 77,582 visits to primarycare practices. The unique medical records for each ofthese patients will be identified and the healthcare

utilization patterns for these patients will be monitoredprospectively through the course of the study. Addi-tional data elements drawn from this dataset includepatient demographics (age, gender, sex, race/ethnicity,and insurance type) as well as concurrent diagnoses.For this study, clinical data from the EMR will be

drawn from 75 of the study clinics (65 clinics in GroupB and 10 clinics in Group C). All patients with asthmawill be identified from billing data and the same uniquemedical record number will be used to pull thesepatient’s clinical data. Data elements that can be pulledinclude: Tobacco Use/Exposure, assessment of daytime/nighttime symptoms, controller medication for persis-tent asthma, action plan given/updated and flu vaccineup to date.The clinics in this study that are not connected to the

EMR will be used as control practices. Clinical datafrom these clinics is abstracted from charts by the CHSquality team and will be ongoing during the 3 years ofthis study for comparison. Data collection will occur ona quarterly basis and will include: Tobacco Use/Expo-sure, assessment of daytime/nighttime symptoms, con-troller medication for persistent asthma, action plangiven/updated, and flu vaccine up to date.A local not for profit institute has been working with

the Charlotte Mecklenburg School System for the past 4years to store data for community-wide research. TheInstitute stores data on school absenteeism and schoolperformance including End of Grade testing results. Theschool data will be added to the ACER server for thestudy to examine changes in school performance andabsenteeism that are related to the study interventions.The management of children with asthma within the

school system through school-based nurses has enor-mous potential to improve asthma outcomes. This studywill support the development of an electronic data cap-ture system for school nurses. Data elements collectedwill include clinical measures including peak flows,asthma medications utilization, and the Asthma ActionPlan.The North Carolina Medicaid System is one of the

most successful in the country in terms of providinghigh quality of care at a low cost [34] This system,Community Care Partners of North Carolina, works byincentivizing primary care physicians to provide preven-tative care through the assistance of case managers [35]CCPGM and Carolinas Healthcare System have a datasharing system allowing access to data for the Medicaidpatients with asthma. The data are collected electroni-cally and includes health services utilization (hospitaliza-tions, ED and clinic visits) and data on medicationcompliance (prescriptions filled). These data are sent toCHS on a monthly basis and will continue to be sharedfor this study. Data for Medicaid will be matched based


Page 5 of 10

77

on patient name and birth date and entered into theACER database for analysis.

Patient Level DataPatient level data will be collected using direct patientsurveys, focus groups, and information exchange duringthe community forums.SurveysDepending on patient age, the patient survey will includethe Mini Asthma Quality of Life Questionnaire (Mini-AQLQ) or Mini Pediatric Asthma Quality of Life Ques-tionnaire (Mini-PAQLQ), the Asthma Therapy Assess-ment Questionnaire (ATAQ) and 2 additional five-pointscale questions added by the research team. The MiniAQLQ (5 items) will licensed from QOLteck (West Sus-sex, UK) and used to collected data from all patients thatare 17 years and older where this survey has been vali-dated [36] The Mini PAQLQ will also be licensed from

QOLteck, and this 13 question instrument will be usedto assess quality of life in children between the ages of 7and 17. The 5 point ATAQ questionnaire will be used toevaluate the effectiveness of asthma management for allpatients. The pediatric version will be administered forchildren age 5-17 and adult version for patients age 17and older [37]. The final questions are identical to thoseused by Wilson and colleagues to measure the patient’sperceptions of a shared decision making approach andasthma self-efficacy [37-39] These questions will askabout the patients’ perceptions of the quality of care theyreceived and the influence they had versus the medicalteam in the development of their treatment plan. The fin-ished survey will include 12 questions for adults and 20questions for children under the age of 17 which will beprinted on a single page for mailing.Surveys will be mailed to a randomly selected group of

asthma patients who receive care within the 95 clinics

Figure 3 Development of the Asthma Comparative Effectiveness Research Database (ACER). Data will be collected from 8 sources intothe database including: Billing data, clinical data drawn electronically from the EMR, School data, Information collected by the school-basedintervention, Medicaid data, clinic data via chart abstraction from control practices without an EMR, Patient level data from surveys andcommunity-level data from the Center for Metropolitan Studies.


Page 6 of 10

78

followed for this study every 6 months with the aim ofcollecting 100 surveys from each of the 5 groups at eachtime point. Patients will be incentivized to complete thesurvey with a $10 gift card for participation. Five of theclinics in this study are part of a network that providescare to disadvantaged patients including the uninsured,low-income minority populations, the indigent, disabledMedicare, and Medicaid patients. These populationstend to have frequent address changes and lower ratesof literacy, resulting in lower rates of participation withmailed surveys. This group is of key importance for thisresearch study and 4 of these 5 clinics have been tar-geted to receive additional interventions to improveasthma outcomes. Patients that complete the survey on-site will be provided with a $10 gift card to reimbursethem for the time required to participate.All survey data will be added to the ACER database

on a weekly basis. Online survey results will be collectedusing Survey Monkey (http://Surveymonkey.com) andthis data with the patient identifier and date of comple-tion will be downloaded and transferred into ACER.Mailed surveys will be compiled by the research teamand manually added to the database. In addition, thedate that the survey was mailed and returned will beincluded for surveys returned by mail. If a survey iscompleted both online and via the mail, the mailed ver-sion will be discarded.Focus GroupsQualitative data will be collected through focus groupsperformed with patients (or their parents for childrenunder age 17) and providers from each of the 5 groupsthroughout the study. Each focus group will havebetween 8-10 participants and will occur at baseline andevery 6 months to collect qualitative information aboutthe study. There will also be a focus group once every1-3 months to evaluate the process and the monthlyasthma meetings. There will be no recruitment forthese, as the participants will be the regular attendeesfrom the monthly asthma meetings (i.e. providers, keypersonnel from the clinic). A focus group guide forthese sessions has been developed by the research team.Groups B-E will be asked to provide feedback abouttheir perceptions of the study and its impact on theirability to receive or provide high quality asthma care.Particular interest will be focused in these groups onsoliciting critical feedback about the project to be usedfor process improvement. Clinics in Group A will alsobe asked to participate, but these groups will be askedmore general questions about their management ofasthma patients and how they hope to improve care forasthma patients in the future. Focus group data will beanalyzed and provided back to all participating practicesand research team. Data from focus groups will beindentified only by practice and no individual data from

the focus groups will be collected. Focus group data willnot be added to the centralized ACER database.Community Level DataThrough MAPPR’s affiliation with the University ofNorth Carolina at Charlotte, UNCC, and Center forMetropolitan Studies, geocoded data will be made avail-able for this study showing: housing density, neighbor-hood quality of life, pollution sources, transportationand built environment elements, race/ethnicity, andhousehold income. These data have been developed bythe center or purchased from Claritas (New York, NY).We will geocode patient addresses for the ACER data-base which will allow the research team to examine andmap asthma outcomes compared with other communitylevel variables. One example of this technique can beseen in Figure 4, where patients with an asthma diagno-sis within the clinic system for underserved or disadvan-taged patients are mapped across Mecklenburg County.These maps can quickly be compared to other commu-nity-level data that will be shared by the Center forMetropolitan Studies.

AnalysisIntervention Effectiveness AnalysisComparisons between each group (A-E) and withinGroup E will be performed by statistical tests of regres-sion model parameters. Specifically, we will generate aseparate regression model for each of the six clinicalmeasures and both quality measures (Table 2). Compar-ison groups will be defined as factors in the model. Pro-pensity score methods will be used to control fordifferences in pharmaceutical management, patientdemographics, co-morbidities, and community levelvariables between groups. Since Group E is not indepen-dent of Groups A-D, we will test for significant interac-tions between Group E and the other groups. Asecondary analysis of the data will be performed toexamine other important variables that impact asthmaoutcomes. This includes pharmaceutical management,co-morbidities (Tobacco Use/Exposure, GERD, AllergicRhinitis, Obesity, Sleep Apnea), patient demographics(age, sex, race/ethnicity, insurance status), and commu-nity level variables (neighborhood quality of life, buildenvironment and transportation elements, pollutionsources, and housing density). In this analysis, we willbuild a regression model for each of the 6 clinical out-comes and use pharmaceutical management, patientdemographics, co-morbidities, and community levelvariables as predictors. Regression tree methods will beused to help understand the potentially complicatedinteractions between these predictor variables. In boththe primary and secondary analyses, linear mixed effectsmodels will be used to account for practice-level effectswithin intervention group.


Page 7 of 10

79

http://Surveymonkey.com

Cost AnalysisThe basic form of our analysis is that of a follow upstudy with three years of analysis. We will constructmeasures of service use and costs, by category of serviceas well as for all services combined, expressing all thesemeasures in per-time-period terms for comparabilityacross groups. This will permit comparison amonggroups to identify the effects of the intervention onhealth disparities, for example. Each of these dependentvariables (use and cost, by service category) is specific toa calendar year. Thus the basic unit of analysis is theperson-year of use or cost. We will use regression analy-sis to estimate the mean difference in outcome variablesbetween groups, where the groups are identified by adummy variable to examine effects of the integratedapproach to asthma management based on the chroniccare model (CCM), while holding constant other factors,and, in a more detailed analysis, by a set of dummy vari-ables to examine the effects of the group interventions Bthrough E, as compared with the control group Areceiving usual care. Interactions between the dummyvariable(s) and dummy variables indicating race/ethni-city will identify the effect of the CCM on health dispa-rities. Cost savings to be achieved have the potential tobe substantial, especially for the costs of hospitalization,

Patients per Square Mile

0 - 23

24 - 46

47 - 108

Hospital

Figure 4 Map of Asthma Diagnosis in Mecklenburg County.Map of Mecklenburg County, NC showing Geographic InformationSystems (GIS) analysis of the 6,800 patients with a primary diagnosisof asthma seen within the community clinic network over 1 year.The map shows patient density per square mile.

Table 2 Asthma Related Quality Improvement Goals and Outcome

Variable Measure/Instrument Data Source Frequency

Quality of Life Mini AQLQ, Mini PAQLQ, ATAQ Individual PatientSurveys and FocusGroups

6 months

HospitalizationRates

Inpatient Admission CHS Hospital Data,Medicaid Data

Monthly

Acute CareUtilization

ED/Urgent Care Visits CHS Data, MedicaidData

Monthly

Primary CareUtilization

Primary Care Visits(Fam. Med, Peds, IM)

CHS Data, MedicaidData

Monthly

SchoolAttendance

Days Missed from School School System (CMS)Data

6 months

SchoolPerformance

Test Scores End of Grade (EOG)Test Scores

12 months

Demographics Race/Ethnicity, Gender, Age, Insurance Type, Address CHS System Data Monthly

Income Per capital household income, Home Value Claritas, Census Data,UNC Urban InstituteData

6 Months

HousingDensity/Location

Homes per sq. mile Claritas, Census Data,UNC Urban InstituteData

6 Months

Other MedicalConditions

Obesity, Growth Delay, Developmental Delay, Upper Respirator Infections, Allergic Rhinitis,GERD, Diabetes, COPD, Cancer diagnosis, Hypertension, Congestive Heart Failure

CHS Hospital andclinical data

Monthly

Clinical OutcomeMeasures

FEV1, Peak Flow, BMI, Height, Tobacco Use, Tobacco Exposure, Vaccination Rates CHS Electronic MedicalRecord/Chart Abstraction

Monthly/Quarterly

MedicationCompliance

Asthma Medications Prescribed, Prescriptions Filled, Refill requests CHS ElectronicPrescribing Database,Medicaid Data

Monthly


Page 8 of 10

80

which is a major cost driver. For example, using nation-ally representative data [40] we found that, among per-sons ages 19-64, the adjusted rate of preventablehospitalization for asthma for African American womenwas 2.2 times as great as that for non-Hispanic whitewomen, while the comparable adjusted difference formen was 4.2 times as great. In this same age range, theadjusted rate for women was 2.2 times as great for His-panics as for non-Hispanics, while among men the ratewas 3.3 times as great for Hispanics.Differences were much larger among those ages 65

and older, with rates 6 times higher than whites amongmen for African Americans, and over 8 times higher forHispanics. Thus, even a modest reduction in the dispari-ties of preventable hospitalization attributable to asthmaaffecting African Americans and Hispanics will substan-tially reduce total costs. Of course, in addition to redu-cing disparities in preventable hospitalization for AfricanAmericans and Hispanics, the intervention is likely toreduce preventable hospitalization related to asthma forall persons in the intervention groups.

DiscussionIdentifying new mechanisms that improve the delivery ofasthma care is an important step towards advancingpatient outcomes, avoiding preventable EmergencyDepartment visits and hospitalizations, while simulta-neously reducing overall healthcare costs. The wide rangeof patient types with asthma (eg. pediatrics vs. adults) aswell as the varying degree of severity of the disease makesit difficult for a single approach to work universally. Inaddition, variation in primary care clinics and providersmakes it challenging to implement new practices and con-cepts around asthma management such as shared decisionmaking. To overcome these limitations, this study wasdesigned with multiple interventions and two controlgroups. Perhaps most importantly, the shared decisionmaking component of the intervention was designed to bedeveloped and implemented using participatory methodsto insure broad uptake and dissemination.

AcknowledgementsWe would like to gratefully acknowledge Drs John Carew, Cheryl Courtlandand Maria Bonaiuto and the member organizations of the MAPPR networkfor their assistance with this work. The project described is funded by AwardNumber 1R18HS019946-01 from the Agency for Healthcare Research andQuality.

Author details1Department of Family Medicine, Carolinas HealthCare System, 2001 VailAvenue, Charlotte, NC 28207. USA. 2Carolinas Physicians Network, CarolinasHealthCare System, PO Box 32861, Charlotte, NC 28232, USA.

Authors’ contributionsAll authors made significant contributions to the conception and design ofthis study and read and approved the final manuscript. HT and MD draftedthe manuscript.

Competing interestsThe authors declare that they have no competing interests.

Received: 28 June 2011 Accepted: 16 August 2011Published: 16 August 2011

References1. What is asthma?. [http://www.nhlbi.nih.gov/health/dci/Diseases/Asthma/

Asthma_WhatIs.html].2. Current asthma population estimates–in thousands by age, United

States: National Health Interview Survey. 2008 [http://www.cdc.gov/asthma/nhis/08/table3-1.htm].

3. Moorman J, Rudd R, Johnson C, King M, Minor P, Bailey C, Scalia M,Akinbami L: National surveillance for asthma–United States, 1980-2004.MMWR Surveill Summ 2007, 56:1-54.

4. Trends in asthma morbidity and mortality. [http://www.lungusa.org/finding-cures/our-research/trend-reports/asthma-trend-report.pdf].

5. Skrepnek G, Skrepnek S: Epidemiology, clinical and economic burden, andnatural history of chronic obstructive pulmonary disease and asthma.Am J Manag Care 2004, 10:S129-S138.

6. Behavioral Risk Factor Surveillance Survey (BRFSS). [http://www.cdc.gov/brfss/index.htm].

7. Jensen K: The Burden of Asthma in North Carolina, 2006. Book The Burdenof Asthma in North Carolina, 2006 City: NC Division of Public Health; 2006,(Editor ed.êds.).

8. Sox HC, Greenfield S: Comparative effectiveness research: a report fromthe Institute of Medicine. Ann Intern Med 2009, 151:203-205.

9. Gillissen A: Patients’ adherence in asthma. J Physiol Pharmacol 2007,58:205-222.

10. National Asthma Education and Prevention Program. Expert panelreport-3: guidelines for the diagnosis and management of asthma. .

11. Haynes RB, Ackloo E, Sahota N, McDonald HP, Yao X: Interventions forenhancing medication adherence. Cochrane Database Syst Rev 2008,CD000011.

12. Bosley CM, Parry DT, Cochrane GM: Patient compliance with inhaledmedication: does combining beta-agonists with corticosteroids improvecompliance? Eur Respir J 1994, 7:504-509.

13. Sackett D, Snow J: The magnitude of compliance and non-complianceBaltimore: Johns Hopkins University Press; 1979.

14. Gong H, Simmons MS, Clark VA, Tashkin DP: Metered-dose inhaler usagein subjects with asthma: comparison of Nebulizer Chronolog and dailydiary recordings. J Allergy Clin Immunol 1988, 82:5-10.

15. Coutts JA, Gibson NA, Paton JY: Measuring compliance with inhaledmedication in asthma. Arch Dis Child 1992, 67:332-333.

16. Milgrom H, Bender B, Ackerson L, Bowry P, Smith B, Rand C:Noncompliance and treatment failure in children with asthma. J AllergyClin Immunol 1996, 98:1051-1057.

17. Gamble J, Stevenson M, McClean E, Heaney LG: The prevalence ofnonadherence in difficult asthma. Am J Respir Crit Care Med 2009,180:817-822.

18. Cates CJ, Lasserson TJ: Regular treatment with formoterol and an inhaledcorticosteroid versus regular treatment with salmeterol and an inhaledcorticosteroid for chronic asthma: serious adverse events. CochraneDatabase Syst Rev 2010, CD007694.

19. Cates CJ, Lasserson TJ: Combination formoterol and inhaled steroidversus beta2-agonist as relief medication for chronic asthma in adultsand children. Cochrane Database Syst Rev 2009, CD007085.

20. Ducharme FM, Lasserson TJ, Cates CJ: Long-acting beta2-agonists versusanti-leukotrienes as add-on therapy to inhaled corticosteroids forchronic asthma. Cochrane Database Syst Rev 2006, CD003137.

21. Cates CJ, Jefferson TO, Rowe BH: Vaccines for preventing influenza inpeople with asthma. Cochrane Database Syst Rev 2008, CD000364.

22. Castro-Rodriguez JA, Rodrigo GJ: The role of inhaled corticosteriods andmontelukast in children with mild-moderate asthma: results of asystematic review with meta-analysis. Arch Dis Child 2009.

23. Rodrigo GJ, Moral VP, Marcos LG, Castro-Rodriguez JA: Safety of regularuse of long-acting beta agonists as monotherapy or added to inhaledcorticosteroids in asthma. A systematic review. Pulm Pharmacol Ther2009, 22:9-19.


Page 9 of 10

81

http://www.nhlbi.nih.gov/health/dci/Diseases/Asthma/Asthma_WhatIs.html

http://www.nhlbi.nih.gov/health/dci/Diseases/Asthma/Asthma_WhatIs.html

http://www.cdc.gov/asthma/nhis/08/table3-1.htm

http://www.cdc.gov/asthma/nhis/08/table3-1.htm

http://www.ncbi.nlm.nih.gov/pubmed/18075488?dopt=Abstract

http://www.lungusa.org/finding-cures/our-research/trend-reports/asthma-trend-report.pdf

http://www.lungusa.org/finding-cures/our-research/trend-reports/asthma-trend-report.pdf



http://www.cdc.gov/brfss/index.htm

http://www.cdc.gov/brfss/index.htm


















24. Rodrigo GJ, Castro-Rodriguez JA: Anticholinergics in the treatment ofchildren and adults with acute asthma: a systematic review with meta-analysis. Thorax 2005, 60:740-746.

25. National Heart Lung and Blood Institute: National Asthma Education andPrevention Program, Expert panel report 3: Guidelines for the diagnosisand management of asthma. Book National Asthma Education andPrevention Program, Expert panel report 3: Guidelines for the diagnosis andmanagement of asthma City: National Asthma Education and PreventionProgram (NAEPP) Coordinating Committee; 2007, (Editor ed.êds.).

26. Navarro R, Rice G, Schaecher K: Asthma management guidelines: updates,advances, and new options. J Manag Care Pharm 2007, 13:S3-S11.

27. Joint Task Force on Practice Parameters, American Academy of Allergy,Asthma and Immunology, American College of Allergy, Asthma andImmunology and Joint Council of Allergy, Asthma and Immunology:Attaining optimal asthma control: a practice parameter. J Allergy ClinImmunol 2005, 116:S3-S11.

28. Chronic care for low-income children with asthma: strategies forimprovement. Research in Action Issue 18. AHRQ Publication No. 05-0073. [http://www.ahrq.gov/research/chasthria/chasthria.htm].

29. Kolbe J: Asthma education, action plans, psychosocial issues andadherence. Can Respir J 1999, 6:273-280.

30. Rank M, Volcheck G, Li J, Patel A, Lim K: Formulating an effective andefficient written asthma action plan. Mayo Clin Proc 2008, 83:1263-1270.

31. Gibson P, Powell H: Written action plans for asthma: an evidence-basedreview of the key components. Thorax 2004, 59:94-99.

32. Mayberry RM, Daniels P, Akintobi TH, Yancey EM, Berry J, Clark N:Community-based organizations’ capacity to plan, implement, andevaluate success. J Community Health 2008, 33:285-292.

33. Morgan D: Focus groups. Annual Review of Sociology 1996, 22:129-152.34. Wegner SE, Dobson LA, Lawrence WW, Ciesco E: Role and structure of the

North Carolina Physician Advisory Group: a collaborative effort betweenproviders and Medicaid. North Carolina medical journal 2009, 70:270-273.

35. St Martin E, Harris-Stevens L: One size does not fit all: tailoring casemanagement to a community. North Carolina medical journal 2009,70:241-244.

36. Olajos-Clow J, Minard J, Szpiro K, Juniper EF, Turcotte S, Jiang X, Jenkins B,Lougheed MD: Validation of an electronic version of the Mini AsthmaQuality of Life Questionnaire. Respir Med 2010.

37. Wilson SR, Strub P, Buist AS, Knowles SB, Lavori PW, Lapidus J, Vollmer WM:Shared Treatment Decision Making Improves Adherence and Outcomesin Poorly Controlled Asthma. Am J Respir Crit Care Med 2010, 181:566-577.

38. Schlosser M, Havermans G: A self-efficacy scale for children andadolescents with asthma: construction and validation. J Asthma 1992,29:99-108.

39. Tobin DL, Wigal JK, Winder JA, Holroyd KA, Creer TL: The “Asthma Self-Efficacy Scale”. Ann Allergy 1987, 59:273-277.

40. Laditka JN, Laditka SB: Race, ethnicity and hospitalization for six chronicambulatory care sensitive conditions in the USA. Ethn Health 2006,11:247-263.

Pre-publication historyThe pre-publication history for this paper can be accessed here:http://www.biomedcentral.com/1472-6963/11/188/prepub

doi:10.1186/1472-6963-11-188Cite this article as: Tapp et al.: Comparative effectiveness of asthmainterventions within a practice based research network. BMC HealthServices Research 2011 11:188.

Submit your next manuscript to BioMed Centraland take full advantage of:

• Convenient online submission

• Thorough peer review

• No space constraints or color figure charges

• Immediate publication on acceptance

• Inclusion in PubMed, CAS, Scopus and Google Scholar

• Research which is freely available for redistribution

Submit your manuscript at www.biomedcentral.com/submit


Page 10 of 10

82







http://www.ahrq.gov/research/chasthria/chasthria.htm






















http://www.biomedcentral.com/1472-6963/11/188/prepub

Documents

Bibliography - Confex · Bibliography Charles C, ... paper attempts to provide greater conceptual clarity about shared treatment ... mation to be collected as relevant examples