Quality in Health Care 1994;3:107-113

Qualitative methods for assessing health care

Ray Fitzpatrick, Mary Boulton

The evaluation of health care and efforts to

maintain and improve quality in health care

have very largely drawn on quantitativemethods. Quantification has made possibleprecise expression of the extent to whichinterventions are efficient, effective, or appro-

priate and has allowed the use of statisticaltechniques to assess the significance offindings. For many questions, however,quantitative methods may be neither feasiblenor desirable. Qualitative methods may bemore appropriate when investigators are

"opening up" a new field of study or are

primarily concerned to identify and con-

ceptualise salient issues. Various qualitativemethods have been developed which poten-

tially have an enormous role in assessinghealth care. This paper examines some of themore important forms that have been used inthat assessment and outlines principles ofgood practice in the application of qualitativemethodology. It is intended to encourage a

wider use of qualitative methods in assessinghealth care and a greater appreciation of howmuch such methods have to offer.The term "qualitative" is sometimes used

quite loosely. We will review some of themethods to which the term is properly applied,but first what is not a qualitative study shouldbe emphasised. Research based on a smallnumber of patients or respondents should notbe considered qualitative just because thesample size is too small to assess statisticalsignificance. This is more likely to prove to bean inadequate quantitative study. Similarly, a

study is not qualitative because it is based on

answers to a questionnaire about subjectivematters nor because data are collected bypersonal interview. If such data are analysedand reported largely in terms of frequenciesand proportions of respondents expressingparticular views, that is also a quantitativestudy. Qualitative research depends upon not

numerical but conceptual analysis andpresentation. It is used where it is important to

understand the meaning and interpretation ofhuman social arrangements such as hospitals,clinics, forms of management, or decisionmaking. Qualitative methods are intended to

convey to policy makers the experiences ofindividuals, groups, and organizations whomay be affected by policies.

Qualitative methods of data collectionQualitative methods of collecting dataencompass a range of approaches (box).

IN DEPTH INTERVIEWS

Perhaps the main method of obtaining data forqualitative analysis is by interview. Interviewsare a particularly flexible method of gatheringdata, allowing the investigator to respond to

the individual way in which respondentsinterpret and answer questions. There are

various interview formats, but qualitativeanalysis requires "in depth" interviews so theinterviewer can obtain more detailedinformation than is possible, for example,from an interviewer administered question-naire, particularly regarding the perceptionsand reasons behind respondents' statements.Interviews may be "semistructured," wherethe interviewer has a fixed set of topics todiscuss, or "unstructured," where the inter-viewer has only very broad objectives inrelation to the interview and is largely led bythe respondent's priorities and concerns.

Unstructured interviews are most appropriatewhen subject matter is particularly complex or

when investigators want to understand reasons

for views or are exploring aeras which have notpreviously been extensively described. Aninterview format increasingly used in healthservices research is the "critical incidenttechnique," in which respondents are asked torecall the details of a particular experiencesuch as hospitalization.'

Because in depth interviews require theactive participation and judgement of the inter-viewer it is important that they be conductedby interviewers who have been appropriatelytrained. There is a range of generalinterviewing skills that are relevant to per-

formance, such as an ability to demonstrateinterest in the respondent without excessiveinvolvement that may result in bias and an

ability to ask both open ended facilitatingquestions and specific probes when relevant.The interviewer also needs to understand thegeneral objectives of a study as well as theintentions behind specific questions. Anotheressential requirement is that interviewers beperceived by respondents as neutral withregard to the subject matter of an interview.For example, hospital staff are inappropriate toconduct interviews about patient satisfaction,and individuals perceived as judgemental are

inappropriate for more sensitive topics.The analysis of in depth interviews normally

entails some degree of formal content analysisof what the respondent has said. For thisreason investigators usually tape recordinterviews which can then be transcribed for

Nuffield College,Oxford OXI 1NFRay Fitzpatrick, fellow

Academic Departmentof Public Health, StMary's HospitalMedical School,London W2 lPGMary Boulton, seniorlecturerCorrespondence to:Dr Fitzpatrick

Qualitative methods of data collection* In depth interviews* Focus groups* Nominal group techniques* Observational studies* Case studies

107 on A

ugust 28, 2020 by guest. Protected by copyright.

http://qualitysafety.bmj.com

/Q

ual Health C

are: first published as 10.1136/qshc.3.2.107 on 1 June 1994. Dow

nloaded from

108 b citprlfilk, BIoltoi

detailed content analysis. Broad principles ofcontent analysis for qualitative research arediscussed below as, to some extent, they arecommon to all of the methods discussed inthis paper.An illustration of the use of in depth

interviews is a study of the neurologicalmanagement of chronic headache. Neurolo-gists were unsure about the appropriateness oftheir role in this area. The initial purpose ofthe study was to identify the expectations ofpatients with chronic headache regarding theirneurological referral, and to establish whetherthose expectations were satisfied. Wheninterviewed before their outpatient appoint-ments, patients proved quite unsure what toexpect. Content analysis of these interviewsinstead disclosed a small number of differentconcerns felt by patients about their headaches- for example, a need to be reassured aboutserious disease or a desire to receive adviceabout how to change lifestyle or diet to avoidheadaches. After their neurological consul-tations a minority of patients did experienceserious disappointment with their clinicattendance but for various reasons were veryuncomfortable with the language of"satisfaction" or "dissatisfaction" to describetheir experiences. Overall the study suggestedparticular failures of communication in clinicsthat led to some patients' concerns not beingfully addressed.

In this example the primary advantage of indepth interviews was that they allowed theinvestigator to focus maximally on patients'perceptions of their health problems andresponses to health care rather than imposinghis own categories. This led to a new andmore meaningful typology of patients'concerns. Examples of other qualitativeanalyses of in depth interviews include studiesof dimensions of patients' experiences of carefor various chronic health problems,' of therole of clinical audit in British medicine,' ofinformal reasons for clinicians' use ofechocardiography," and of the concerns ofsinglehanded GPs in an inner London area.'

FOC US GROUPS

Groups rather than individuals may beinterviewed bv means of a techniquecommonly referred to as a focus group.Typically, eight to ten individuals are recruitedto a group discussion about specified topics.The discussion is led by a trained moderatoror facilitator and normally lasts for one and ahalf to two hours. The discussion is taperecorded, transcribed, and subjected tocontent analysis. Rather like the interviewerconducting in depth interviews, the moderatorhas a clear agenda of issues on which he or shestimulates discussion but aims to be fairlynon-directive in encouraging discussion.Membership of a focus group needs to bereasonably homogeneous; too much hetero-geneity in terms of social background orperspective on a topic tends to result inparticipants feeling inhibited from revealingviews. Therefore if the purpose is to identifythe range of views of individuals of different

social backgrounds the normal practice is toconduct a series of focus groups.Focus groups are particularly useful where

investigators wish to establish quickly therange of perspectives on an issue ofimportance among different groups - forexample, when it is felt necessary to "bridge agulf' in understanding between providers of aservice and the intended users. The dynamicsof a well conducted focus group are such thatindividuals' revelations of their views can"spark off" other participants to reveal broaderinsights than are possible from individualinterviews. However, focus group methodshave been compared to "pulling teeth" tostimulate discussion in a group that has noexperience or interest in a topic.' It is lessappropriate in a focus group to probe forelaboration of individuals' statements than ispossible in an in depth interview, so that focusgroups may be said to be stronger on breadththan depth. The facilitator is less in control ofa focus group than is an interviewer with singlerespondents so it is a method that maximisesthe expression of perspectives not imposed bythe researchers. There are some topics that aresufficiently sensitive that group discussion isinhibiting whereas individual interviewsprovide a more confidential context for selfrevelation.Focus groups are commonly regarded as an

exploratory method, so that investigators mayconduct them in order to design the question-naire for a more definitive quantitative survey.Some would argue, however, that focus groupscan provide valuable evidence in their ownright, provided that investigators areconcerned with conceptual rather thannumerical analysis.'There are questions still to be addressed

with focus group methodology. For example,little is known about how the effects of socialdesirability and conformity influence expres-sion of views and to what extent heterogeneityof participants influences results. Similarlythough facilitators clearly require substantialinterpersonal skills such as ability to listen andto facilitate without becoming so involved asto bias discussion, facilitator effects on thequality of focus groups are not wellunderstood.( However, health is a particularlyappropriate application for this method, "'which is being increasingly advocated for useby purchasers in obtaining local views abouthealth and health care.''

NO.\AINAI GROUP 1 T( IHNIQUFSFor some purposes a more impersonal and lessthreatening form of group dynamics may beneeded. A somewhat more structured form ofgathering data from groups has beendeveloped, called "nominal groups" becauseexchange and interaction between groupmembers is more controlled than in focusgroups. It is considered a technique less proneto the bias arising from vocal individualsinfluencing a group's views in opendiscussion.A nominal group normally is composed of

eight members who meet together with a

108 on A

ugust 28, 2020 by guest. Protected by copyright.

http://qualitysafety.bmj.com

/Q

ual Health C

are: first published as 10.1136/qshc.3.2.107 on 1 June 1994. Dow

nloaded from

Qualitative methods for assessing health care

leader who introduces the group's tasks. Theleader explains the question or problem onwhich the group's views are sought.Individuals are asked to list on a paper formtheir different feelings or experiences inrelation to the question (for example, differentkinds of disappointments experienced by agroup of users of a clinic), without discussionwith other group members. Participants arethen asked to declare their written commentswhich are recorded, as closely as possible tothe participants' own words, on a blackboardor flipchart. The complete list is thendiscussed by the group and a preliminaryranking made by the group from most to leastimportant item of the total list. Afterdiscussion of this preliminary ranking asecond, private ranking of items is performedby individuals on paper. This informationconstitutes the researcher's core data. As withfocus groups, investigators usually set upseveral different groups that are needed torepresent the different perspectives withinterests involved with an issue.The primary advantages of this form of

group research are that individuals may givemore carefully considered expressions of theirviews compared with focus groups and areconstrained by the tasks to produce morestructured and explicit priorities. The struc-tured nature of data gathering means that thisis a method particularly appropriate tocontexts where technical or complex issuesneed to be assessed. The primary disadvantageis that individuals are discouraged from"sparking off' each other, so that it is possiblethat the full range of views and observationsare not elicited. Overall, like focus groups,nominal group techniques are a relatively lowcost and quick method of exploring theparameters of an issue.A variant of this methodology has been

developed for use in developing professionalconsensus about appropriate indications forhealth care interventions. Panels representingrelevant clinical expertise are provided withpatient vignettes with varying symptoms andother indications. Vingettes are privately ratedin terms of degree of appropriateness for theintervention. The group is then actuallyassembled, it discusses the ratings, and it isgiven the opportunity to revise prior decisions.A recent example indicates that substantialagreement could be obtained regarding appro-priateness for prostatectomy. 14 Most of thesteps in data gathering and interpretation ofthis application of nominal group techniquesare actually quantitative, and readers arereferred elsewhere for methodological prob-lems associated with such techniques.'5

OBSERVATIONAL STUDIES

When the research question concerns whatactually happens in health care settings, ratherthan participants' perceptions of andresponses to it, a more appropriate method ofdata collection may be direct observation. Forthis method the investigators attend the eventsthey are concerned with, pay close attention towhat goes on, and make a careful record of it

for future analysis, using analytical techniquessimilar to those for other qualitative data.

Careful recording of what occurs is thecornerstone of observational research. Fieldnotes, taken during or immediately afterperiods of observation, have been thetraditional method of recording. The value offield notes depends on the skill and disciplineof the investigator in recording as much detailas accurately as possible. Field notes areinevitably selective and so may be subject toobserver bias. However, they are a relativelyefficient way of recording observations,particularly in sensitive situations, and providedata in a form that is immediately accessiblefor analysis. Audio recordings and videorecordings provide a more detailed andaccurate record of events, although thedemands of the technology may limit therange of events that can be observed and thetapes themselves present considerableadditional demands in terms in transcriptionbefore analysis. Tape recordings are attractivein that they allow the investigator to go backand "reobserve" events, which may allowmore detailed analysis than is possible withfield notes. They also afford a record of theevidence on which interpretations are madewhich may be inspected by independentobservers in establishing the validity of theanalysis. However, audio recorders andparticularly video recorders have beencriticised as being so obtrusive as to changethe nature of the events they are recording,thereby invalidating the observation itself.With smaller machines and consulting roomsspecifically adapted for recording this isbecoming less of a problem, and, in any case,those who have been the subject of audio orvideo recording have almost always found thatthe need to deal with the immediate demandsof the situation constrain them to carry on insuch the same way as they would if they werenot being observed.

Investigators may collect data as "partici-pant" or as "non-participant observers."Participant observers participate in the dailylife of the organisation over an extendedperiod of time, watching what happens,listening to what is said, and asking questions.It is an exceptionally demanding way ofgathering data. The researcher may encounterdifficulties in being accepted by the groupinitially and then in sustaining the role longenough to observe the full range of events.The justification for such efforts is in the wayparticipant observation enables the researcherto see and experience the institutional culturefrom the point of view of an "insider."Goffman's classic study of asylums illustrateswell the unique insights that can be gained bythis method of data collection.'6 Goffmanjoined an American mental hospital as theassistant to the athletic director in order toobserve the culture of the institution withoutdrawing attention to himself. The resultingstudy develops the concept of the "totalinstitution," identifying its core elements andusing it as a framework to describe and makesense of the culture of both the patients and

109

on August 28, 2020 by guest. P

rotected by copyright.http://qualitysafety.bm

j.com/

Qual H

ealth Care: first published as 10.1136/qshc.3.2.107 on 1 June 1994. D

ownloaded from

Fitzpatrick, Boulton

the staff. By closely examining what actuallywent on in the hospital, he was able to getbehind official accounts of what the institutiondid to show the process and strategies thatmade it work in practice, and theconsequences they had for patients' lives.

Participant observation has become lesscommon as a research method, partly becauseof questions raised about the ethics of covertobservation. Non-participant observationallows the researcher to remain as an acceptedoutsider, watching and recording the inter-actions as a "fly on the wall." Non-participantobservation is particularly useful when theresearcher is concerned to describe andconceptualise the "taken for granted" prac-tices of everyday medical life: the routines andstrategies that those they are studying developin carrying out their work which may be socommon and familiar as to be outside theirconscious awareness. An example is Strong'sstudy described in The Ceremonial Order of theClinic in which he set out to elucidate the ruleswhich govern interactions in medicalconsultations.'7 On the basis of observations of1 120 paediatric consultations in severalmedical settings in Scotland and the UnitedStates he documented in detail the existence,nature, and scope of the rules which make upthe limited number of institutionalized rolesuniversally adopted by patients and doctors.The main advantage of observation as a

method of data collection is in allowing theinvestigators to "see for themselves," thusavoiding the biases inherent in participants'reports, such as selective perception, poorrecall, and the desire to present themselveswell. It is particularly appropriate when theinvestigator is concerned with describing andconceptualising how health services operate -that is, with practices and processes which aredirectly accessible to an outside observer. Itsmain drawback is that it limits the number andrange of situations that can be studied to thoseat which the investigator is present. It is themost labour intensive form of data collectionand perhaps for this reason is more often usedin conjunction with other methods rather thanas the main source of data.

CASE STUDIES

Case studies are not really a distinctivemethod of data collection or analysis butwarrant separate discussion because they areincreasingly used in the study of health caresystems. They provide interpretations andanalyses that are ultimately qualitative innature. The most obvious respect in whichcase studies are qualitative is that typicallythey involve the study of a single or smallnumber of units where quantitative manipu-lation of variables associated with units wouldbe inappropriate. However, to a greater extentthan the methods discussed so far, case studiesoften draw on mixture of quantitative andqualitative data. An example illustrates thepoint. In 1986 a resource management (RM)initiative was announced for the NHS, thepurpose of which was to improve patient careby providing systems whereby managers and

clinicians made better informed and moreeffective use of resources. The initiative was tobe piloted in six acute hospitals and the costsand benefits were to be evaluated by a researchteam.'8 Much quantitative data were gatheredby investigators in relation to hospital activitiesand their costs in the pilot sites over time.However, as a number of other institutionalchanges to the NHS were simultaneouslyhaving effects in both the study hospitals andany potential "control" hospitals outside thepilot it was decided that it would beimpossible to isolate quantitative benefits dueto the resource management initiativespecifically. Moreover, as with most humanorganisations, NHS defined objectives of theresource management inititative drifted overtime. The authors concluded that qualitativeevaluation of the initiative base on qualitativeand quantitative evidence available was moreinformative. Qualitative inferences were drawnfrom the study such as that it was possible toinvolve clinicians in hospital resource manage-ment but that it was not clear that the highadministrative and other costs associated withsuch involvement were matched by benefits topatient care.

Case studies such as the resource manage-ment study usually entail a combination ofmethods. Investigators carry out interviews,conduct participant observation on relevantmeetings, and inspect written documents suchas minutes or records. A systematic approachto interviews in an organisation can beadopted. For example, in a study of the impactof general management in the NHS,investigators, if confronted with discrepantaccounts between different actors, wouldprobe in interviews to obtain possible explan-ations. '9 Case studies attempt to get anaccurate picture by means of "triangulation,"whereby the degree of convergence betweendifferent sources (for example, interview anddocuments) is carefully considered.20 Aparticular form of checking the plausibility ofinvestigators' explanations is that of'respondent validation' in which the analysisthat has emerged of a setting is presented toparticipants for their reactions. By means ofthis technique analysis can be refined andimproved by respondents' feedback.'9

Qualitative analysis of dataIf the distinguishing feature of quantitativeevidence is the manipulation of numerical datathen qualitative analysis is characterised by thedevelopment and manipulation of concepts.The investigator's primary tasks are theinspection and coding of his or her data in

. . .qualitative analysis ischaracterised by the developmentand manipulation of concepts.

terms of concepts and then the manipulationof such concepts into analyses of underlyingpatterns. Although there are significant

110 on A

ugust 28, 2020 by guest. Protected by copyright.

http://qualitysafety.bmj.com

/Q

ual Health C

are: first published as 10.1136/qshc.3.2.107 on 1 June 1994. Dow

nloaded from

Qualitative methods for assessing health care

differences of form or emphasis, this crucialrole of conceptual analysis is common to all ofthe qualitative methods outlined above. It isprobably the phase of work that is most arcaneand mysterious for audiences of qualitativereports and requires some account of howconceptual analysis of qualitative data isactually done.Whatever method of data collection is used

in qualitative methodology, it results in rawmaterial in the form of "text" - written words.The first stage of processing text is to code andclassify. The investigator therefore reviews hisor her textual material for coding in terms ofconcepts and categories that may emerge fromthe material or that inform the investigator'sstudy to begin with. This transformed, codedmaterial can then be regrouped or indexed tofacilitate further analysis. This stage ofgrouping and indexing coded material hasbeen considerably eased by the developmentof computer packages such as Ethnograph,although such technology cannot replace thefar more important human interpretation oftext before mechanical manipulation.2'The next and most difficult stage is the

analysis of both the original textual data andthe transformed conceptual material. Thisphase is the most difficult to prescribe orindeed describe since it entails a large amountof creative interpretation of evidence. Thoughtprocesses include constant comparison ofevidence regarding different settings or view-points represented in the data and the searchfor deviant or contrasting observations. Onevery common feature of this central analyticalphase is the development of typologies thatconvey the range of views, responses, orarrangements under study. Indeed a typologymay be one of the central results of aqualitative analysis. Thus in the example citedearlier of a study of patients presentingheadaches to neurological clinics a key insightinto the strengths and weaknesses of suchclinics came from developing a typology of thedifferent major concerns that motivatedpatients to seek medical help for headache.Patients' views of the benefits of such clinicscould be largely understood in terms of thesedifferent concerns. A quite contrasting studyanalysing the role of medical audit advisorygroups (MAAGs) identified a typology ofthree different major models of MAAG.22Often it is simply that certain unanticipatedthemes emerge from content analysis.Elsewhere in this issue (p 69) is described animportant study of stroke survivors' views ofthe benefits of physiotherapy.23 Severalpatients regarded physiotherapy as a source offaith and hope for the future. The authorsargue that such insights would not emergefrom the current batteries of available patientsatisfaction instruments used in quantitativeanalysis, which concentrate on more familiardimensions of process and outcome.A variety of intimidating terms have been

developed to describe the logic of qualitativeanalysis including "analytic induction" and"grounded theory".24 In different ways suchaccounts emphasise one common feature -

namely, that qualitative analysis is iterative.The investigator goes back and forth betweenhis developing concepts and ideas and the rawdata of texts or, ideally, fresh observations ofthe field.

Good practice for qualitative researchThe accusation is sometimes made thatqualitative research is an "easy option." Thisusually takes the form of an invidious contrastbetween informally acquired impressions andthe systematic rigour of quantitative method-ology. In reality the difficulties of obtainingand conveying insights that are convincing torelevant audiences are at least as great forqualitative research. This section outlinessome of the general principles of good practicefor qualitative methods that have beendeveloped to facilitate analyses that areplausible and relevant to policy (box). Theseprinciples are further developed in a growingnumber of more detailed guides to qualitativemethodology.25-27

Principles of good practice forqualitative research* Theoretical sampling* Validation* Conceptual analysis

SAMPLINGIn quantitative studies considerable attentionis given to obtaining a sample which isstatistically representative of the population ofinterest so that generalisations may be madefrom the study. This is generally achievedthrough random sampling. In qualitativestudies numerical generalisations are lessimportant than conceptual generalisations. Inthis case what it is important in drawing asample is to ensure that it contains the fullrange of possible observations so that theconcepts and categories developed provide acomprehensive conceptualization of thesubject. The way in which this is done isthrough "theoretical sampling."28 On the basisof his or her theoretical understanding theinvestigator determines what factors mightaffect variability in the observations and thenendeavours to draw the sample in a way whichmaximises the variability. This may be decidedin advance so that, for example, in a study ofa clinic the investigator may be careful tointerview patients of varying age, sex, socialclass, and ethnic background or patients seenby different doctors at different times of day.Theoretical sampling may also requireadjusting sources of observation in the courseof a study to respond to unanticipated patternsor subgroups of experience which may needincreased representation in the sample.

Because qualitative data are morecumbersome to manipulate and analyse most

qualitative studies are restricted to a smallsample size which is unlikely to be statisticallyrepresentative of the population. Nevertheless,accounts of qualitative research should alwaysprovide a clear account of sampling strategies

III on A

ugust 28, 2020 by guest. Protected by copyright.

http://qualitysafety.bmj.com

/Q

ual Health C

are: first published as 10.1136/qshc.3.2.107 on 1 June 1994. Dow

nloaded from

F'itzpatrick, BI0oilto112

to allow readers to judge the generalisability ofthe conceptual analyses.

VAL 1)ID'IY

In the same way every effort is required toestablish the validity of analyses based onqualitative material. The two most commonlycited methods of validating qualitative analysishave already been mentioned in relation tocase studies - that is, triangulation andrespondent validation. In the first methodevery effort is made to obtain evidence from asdiverse and independent a range of sources aspossible. This approach is not very differentfrom the process of establishing constructvalidity for quantitative measures in that one islooking for patterns of convergence betweendata sources that together corroborate anoverall interpretation. Respondent validationrequires that investigators obtain subjects'reactions to their analysis and incorporatesuch reactions into a more complete analysis.However, neither method provides a perfectsolution to the problem of validation.Unfortunately, establishing the plausibility ofanalyses of social settings and organizationssuch as health care cannot be done bymechanical use of procedures such astriangulation or respondent validation. Theworld of health care is particularly complexand interpretations of decisions, behaviour,and arrangements will generally vary amongparticipants. Participants may have variousreasons for not agreeing with analyses of theirbehaviour, and, indeed, such disagreementsmay provide further revealing evidence of howan organisation works.

It is therefore important that users ofqualitative methods adopt a number ofprinciples to convince audiences of the validityof their analyses. Have investigators sampledthe diverse range of individuals and settingsrelevant to their question? How much havethey drawn on and collected evidence in termsof interviews, records, field notes, that are inprinciple capable of independent inspection byothers? How much have they drawn on,whenever available or appropriate, quan-titative evidence to check or test qualitativestatements? To what extent do investigatorsseem to have sought out observations thatmight contradict or modify their analyses? 9It remains a matter of judgement for audiencesof qualitative studies to determine howsystematically investigators have approachedsuch questions in assembling and interpretingtheir evidence. It is essential that qualitativeresearchers make more explicit the methodswhereby their analyses have emerged so thataudiences can make informed judgementsabout plausibility.

UNDERSTANDING ISSUES 01' QUALIlTY OF CARE

What is the role of qualitative research inunderstanding issues of quality of care? On theone hand, qualitative methods have enormouspotential to illuminate how health carecurrently operates and the impact of care onpatients. Thus puzzling issues such as thepersistence of variations in clinical practice

should be addressed by qualitative methods.How patients experience and benefit fromhealth services also need examination by suchmethods. On the other hand, as more syste-matic evidence accumulates of effectiveinterventions and appropriate forms of healthcare, another kind of application of qualitativemethods is required. The changes in healthservices needed to promote quality are organ-isational and cultural and involve differingperspectives of a diverse range of healthprofessionals, managers, and patients.Mechanisms of change such as clinical audit,quality assurance, and the adoption of clinicalguidelines are social processes with meaningsto participants that we need to understand. Insuch a complex environment qualitativemethodology will be essential to give usmodels of how organizations change andinnovate to adopt quality in health care.There is an exciting range of qualitative

methods capable of providing basic under-standing of the processes and outcomes ofhealth care. Rigorous and transparentattention to methodology is needed toconvince audiences of the value of insightsinto the "black box" of health care. An under-standing of these methods will in turn promotea fuller appreciation of the insights they canprovide.

Pryce-Jones M. Critical incident technique as a method ofassessing patient satisfaction. In: Fitzpatrick R, HopkinsA, eds. AlWcaslfc ll.t o/ pantiten9 S'tiactllti)1 i/ithicit .tLondon: Royal C(ollege of Phs sicians of t ondon,1993:87 98.

2 Fitzpatrick R, Hopkins A. Illness behaviour and headache,and the sociology of consultations for headache. In:Hopkins A, ed. Heache./c: probl/.tts i'll 11tilololiCt11.London: 1988:351 835.

3 Fitzpatrick R, Hopkins A. Patient satisfaction in relation toclinical care: a neglected contribution. In: Fitzpatrick R,Hopkins A, eds. .XIaso-utciii of pat'ioctso S'ti /ff "t,2ait/ith1/it Ctare. Londc)n: Royal College of Phs sicians of'London, 1993:77 86.

4 Anderson R, Busr\1, eds. Lib iik,, oh ./l, i//i.London: Unwsin Hvrman, 1988.

5 Black N, Thompson E. Obstracles to medical audit: Britishdoctors speak. Soc. Sci MXIcd 1993;36:849 ',6.

6 Daly J. Innocent murmurs: echocardiography and thediagnosis of cardiac normality. .Sc'()w/ogv of HI-I.lth andIllness 1989;11:99-116.

7 Green J. The 'iews of singlehanded general practitioners: aqualitative study. BA'I 1993;307:607 10.

8 Morgan D, Krueger R. When to use focus groups and sw hs.In: Morgan D, ed. oSicd.x/id foot ipoitp,. I ondon: Sage,1993:3-19.

9 Miorgan D. Future directions for focus groups. In': MorganD, ed. Suiccss.fi/d f/it 'tOUPs. London: Sage,1993:225-44.

10 de Vries H, Weijts W, Dijkstra M, Kok G. 'IThe utilisationiof qualitative and quantitative data for health educationprogram planning, implementation and evaluation, aspiral approach. Hcalth I/itdc Q 1 992;19:10(1 15.

11 NHS Management Executive. ILoal zo-i.. ILondon:Department of Health, 1992.

12 Murray S, Tapson J, Turnbull L, McCallum J, little A.listening to local voices: adapting rapid appraisal toassess health and social needs in general practice. IAIY'71994;308:698-700.

13 Van de Ven A, Delbecq A. 'IThe nominal group techniqueas a research instrument for exploratory health studies.Aii Y Pub/ic Health 1972;62:337 -42.

14 Hunter D, McKee C, Sanderson C, Black N. Appropriateindications for prostatectomy in the UK results of aconsensus panel. 7 Epidetinil (t11111tt1ittii' , Health1994;48:58-64.

15 Scott E, Black N. When does consensus exist in expertpanels? ] Public Hcalt/i 1991;13:35 9.

16 Goffman E. Asyuitts: ssastan'ttes/icilsi/ll ttia))l af tic (tIlpaticts atid ot/het ittttiatcs. Ness York: Anchor, 1961.

17 Strong P. The ctrtittiia/itrdcto ifd/ic c/f iti London:Routledge, 1979.

18 Packwood T, Keen J, Buxton M. Hospitals iii traisitiOi: thi/iesoiocc ttiatiagectcewt c spetiewtt. Milton Keynes: OpenUniversity Press, 1991.

19 Pollitt C, Harrison S, Hunter D, Mlarnoch Gi. Generalmanagement in the NHS: the initial impact 1')83 88.I-'P bl/c A(iittiiiisticitt ti 1991;69:61 83.

on August 28, 2020 by guest. P

rotected by copyright.http://qualitysafety.bm

j.com/

Qual H

ealth Care: first published as 10.1136/qshc.3.2.107 on 1 June 1994. D

ownloaded from

Qualitative methods for assessing health care

20 Bloor M. On the analysis of observational data: a discussionof the worth and use of inductive techniques andrespondent validation. Sociology 1978;12:545-52.

21 Fielding N, Lee R, eds. Using computers in qualitativeresearch. London: Sage, 1991.

22 Humphrey C, Berrow D. Developing role of medical auditadvisory groups. Quality in Health Care 1993;2:232-38.

23 Pound P, Bury M, Gompertz P, Ebrahim S. Views ofsurvivors of stroke on the benefits of physiotherapy.Quality in Health Care 1994;3:69-74.

24 Bryman A, Burgess R. Developments in qualitative dataanalysis: an introduction. In: Bryman A, Burgess R, eds.Analysing qualitative data. London: Routledge,1993:1-15.

25

26

27

28

29

Crabtree B, Miller W, eds. Doizg qualitative research:multiple strategies. London: Sage, 1993.

Silverman D. Interpretiug qualitative data. London: Sage,1993.

Bryman A, Burgess R, eds. Aualvsing qualitative data.London: Routledge, 1993.

Strauss A. Qualitative ayiah'sis for social scientists.Cambridge: Cambridge University Press, 1987.

Dingwall R. "Don't mind him - he's from Barcelona:"qualitative methods in health studies. In: Daly J,McDonald I, Willis E, eds.Researchinlg health care.London: Routledge, 1992:161-75.

113 on A

ugust 28, 2020 by guest. Protected by copyright.

http://qualitysafety.bmj.com

/Q

ual Health C

are: first published as 10.1136/qshc.3.2.107 on 1 June 1994. Dow

nloaded from