Acta Anaesthesiol Scand 1999; 43: 167–172 Copyright C Acta Anaesthesiol Scand 1999Printed in Denmark. All rights reserved

ACTA ANAESTHESIOLOGICA SCANDINAVICA

ISSN 0001-5172

Attitudes of Swedish physicians and nurses towards theuse of life-sustaining treatment

P. SJOKVIST1, L. BERGGREN1 and D. J. COOK2

1Department of Anesthesia and Intensive Care, Orebro Medical Center Hospital, Orebro, Sweden and 2Departments of Medicine and ClinicalEpidemiology, McMaster University, Hamilton, Ontario, Canada

Background: Withdrawal or withholding of life-sustaining treat-ment have become accepted clinical practice within the intensivecare unit (ICU). One important factor influencing these decisionsis the attitudes of physicians and nurses.Method: Questionnaire survey of physicians and nurses in ICUsin 12 Swedish university-affiliated and/or tertiary referral hos-pitals.Results: The response rate was 850 of 1081 (79%) potentiallyeligible health care workers. Respondents first rated the import-ance of 16 factors considered in the decision to withdraw lifesupport. The most important factors were the patient’s likeli-hood of surviving the current episode, patient advance direc-tives, patient age and likelihood of long-time survival. Respon-dents also chose between five levels of care, ranging from com-fort measures to full intensive care, in two of 12 differentscenarios. Respondent characteristics affecting the level of care

THE INTENSIVE CARE UNIT (ICU) is not only a placefor life support, but also the place where some of

us will meet death. Helping our patients to die withdignity and without unnecessary suffering is one ofthe duties of intensive care. The decision to changefocus from life-sustaining treatment to palliative careis often described as the withdrawal of life support.

Several guidelines on withholding and withdraw-ing life-sustaining therapy have been published (1–5).The frequency of decisions to withdraw or withholdtherapy in the ICU has varied in different countriesfrom 6 to 11% (6–10). To understand the nature ofsuch decisions it is important to consider not only theprior and current health status, illness severity andwishes regarding the terminal care of critically ill pa-tients, but also the attitudes among health careworkers regarding withholding or withdrawing of lifesupport. Such attitudes vary both across cultures (11–13) and within countries (14, 15). However, the atti-tudes of health care personnel in most countries to-wards these end-of-life decisions are just beginning tobe studied.

The purpose of this study was to examine the atti-

167

chosen were the number of years of ICU experience and theparticular ICU in which the respondent worked.Conclusion: Advance directives are believed by Swedish inten-sive care personnel to be very important in the decision to with-draw life support, contrary to several descriptive studies sug-gesting modest patient and family influence on these decisions.Attitudes towards the intensity of care vary between differentcenters, raising the possibility that levels of care for similar pa-tients may differ across the country.

Received 12 march, accepted for publication 24 August 1998

Key words: Life support; critical care; ethics; forgoing life-sus-taining treatment; questionnaire; attitudes.

c Acta Anaesthesiologica Scandinavica 43 (1999)

tudes of intensive care physicians and nurses inSweden regarding decisions to limit life support incritically ill patients.

Methods

I. Development and testingIn this survey we used an existing thoroughly testedquestionnaire specifically aimed at examing attitudesof intensive care professionals. This instrument wasoriginally developed by one of the authors.

The development and testing of this questionnairehas previously been reported in detail (14). Potentialdeterminants of the decision to limit life support werederived both from a literature search and through in-terviews with 30 intensive care physicians and nurses.The English version of the questionnaire was testedfor reliability, showing good intraclass correlation be-tween answers when administered to 25 health careworkers twice two weeks apart. The clinical sensibil-ity was examined by asking 75 ICU workers andmethodologists to rate the instrument’s potential abil-ity to discriminate among respondents, and its clarity,

P. Sjokvist et al.

face validity, and comprehensiveness. The proportionof respondents rating the instrument positively inthese domains ranged from 75% to 92%.

The original English language questionnaire wastranslated into Swedish by the two Swedish authors,retaining the meaning of the original questions. A fewchanges were made for cultural adaptation. A bi-lingual clinician who did not participate in the initialtranslation translated the questionnaire back to Eng-lish. This back-translated version proved true to theEnglish original.

The clinical sensibility of the Swedish version wasassessed again by five intensive care physicians andfive critical care nurses. A Likert scale, ranging from1 (agrees completely) to 4 (disagrees), was used to ass-es face validity, clarity and utility (16).– Face validity: all 10 respondents agreed completelythat the questions were directed at important ele-ments of the decision to limit life support.– Clarity: all 10 respondents agreed completely ormainly with the language being easy to understandwith no risk of misunderstanding.– Utility: 9 of 10 respondents agreed completely ormainly that the questionnaire was likely to give a truepicture of the respondent’s opinions.

II. The questionnaire formatThe questionnaire consisted of three parts. The firstpart contained questions about the respondent’s back-ground. In the second part, the respondent was askedto rate the importance of 16 different factors for thedecision to limit life support, using a Likert scaleranging from 1 (completely irrelevant) to 7 (extremelyimportant).

In the third part, two clinical scenarios were pre-sented in each questionnaire. These scenarios weredrawn from a pool of 12 scenarios. Four patient fac-tors were varied in the scenarios: age (45 vs. 75 years);premorbid cognitive function (highly functional vs.encephalitis or Alzheimer’s disease); likelihood ofsurviving current episode (50% mortality [AcutePhysiology and Chronic Health Evaluation II(APACHE II) scoreΩ24] vs. 90% mortality [APACHEII scoreΩ38]); and likelihood of long-term survival(90% 1-year mortality [breast cancer with vertebralmetastasis] vs. no underlying co-morbidity affectinglong-term survival). An example scenario follows:

A 75-year-old woman was admitted to the ICU ten daysago with urosepsis. She requires one inotrope to maintaina mean arterial pressure of 80 mm Hg, is starting to beweaned from mechanical ventilation, but has acute non-

168

oliguric renal failure. Her APACHE score is 24, indicatinga 50% chance of survival.

Her past history includes long-standing depression, re-sponsive to treatment, and breast cancer diagnosed fiveyears ago. Three months prior to admission, she was foundto have recurrent local disease with vertebral metastasesand is now receiving palliative hormonal therapy. Her can-cer-related one-year mortality rate is 90%. She has beenwalking with a cane due to a chronic deformity from polio.

She used to run the family manufacturing business,which involved supervision of ten people. However, her pre-morbid cognitive function was limited by Alzheimer’s dis-ease. She now lives in a chronic care facility because of aninability to look after herself. She is able to feed and dressherself and carry on a simple conversation.

The patient is single. There are no known written or ver-bal advance directives. She has an older brother living inNorway, with whom she has not spoken for years, and whois impossible to contact. There are no other living relatives.A few friends visit her in the ICU, but none want to beinvolved in decisions regarding her medical care.

Table 1

Principal demographic data of the respondents.

Physicians Nurses(nΩ352) (nΩ498)

Male, no. (%) 246 (69.9) 59 (11.9)Age, years (mean∫SD) 42.6∫7.3 38.3∫7.1Time since graduation, years (mean∫SD) 14.5∫7.3 13.4∫6.3ICU experience, years (mean∫SD) 10.6∫7.1 9.5∫7.7

Table 2

Ratings of determinants of the limiting of life support.

Physician NurseFactor score score

Likelihood of surviving current episode 6.1 5.8Patient advance directives 6.0 6.2Patient age 4.9 5.2Likelihood of long-term survival 4.9 4.8Premorbid physical function 4.5 4.2Family directives 4.2 4.5Premorbid intellectual function 4.0 3.8Risk of legal complication 3.2 4.4Alcohol abuse 3.1 3.2Compliance with medical care 3.0 4.1Drug abuse 2.9 3.0Premorbid emotional function 2.6 2.8Religious conviction 1.6 2.3Religious affiliation 1.3 1.9Socio-economic status/occupation 1.3 1.4Ethnic background 1.2 1.7

Ratings: 7 indicates that a factor is extremely important and 1 com-pletely irrelevant.

Survey on life-sustaining treatment

The respondents were asked to choose a manage-ment strategy for each of two clinical scenarios. Therespondents were offered the following options: (1)discontinue inotropes and mechanical ventilation butcontinue comfort measures; (2) discontinue inotropesand other maintenance therapy but continue mechan-ical ventilation; (3) continue with current manage-ment but add no new therapeutic intervention; (4)continue with current management, add further in-otropes, change antibiotics as needed, but do not startdialysis; or (5) continue with full aggressive manage-ment and plan for dialysis if necessary. These fivemanagement strategies will henceforth be referred toas ‘‘levels of care’’.

First, respondents were asked to choose one of thefive management strategies for each of the two scen-arios based solely on what they considered to be inthe best interest of the patient. Next, respondentswere asked to consider all other factors which mayalso influence the decision (i.e. peer opinion, hospital/department policy, medico-legal issues) and, with thisin mind, to choose the level of care closest to whatthey really would do facing the same two cases.

III. Questionnaire administrationWe approached 13 departments in 12 university-af-filiated and/or tertiary referral hospitals. This repre-sents all hospitals of this type in Sweden. All depart-

Table 3

Distribution (%) of responses to scenarios.

Scenario number

1 2 3 4 5 6

a. When the question is ‘‘what is in the patient’s best interest?’’Full aggressive treatment 0 13 16 9 29 70Don’t start dialysis 8 13 14 9 11 13No new intervention 30 35 30 38 30 15Mechanical ventilation 38 28 24 32 21 1Comfort measures only 24 12 16 13 9 1

b. When the question is ‘‘what would you do if you were faced with a case as in the scenario?’’Full aggressive treatment 6 11 14 14 31 71Don’t start dialysis 7 18 19 14 16 17No new intervention 39 42 46 46 31 12Mechanical ventilation 39 24 14 22 18 0Comfort measures only 10 5 7 3 4 0

Characteristics of scenario 1–6 presented in Table 3a, bPatient age (years) 75 75 75 45 45 45Chance of short-term survival (%) 10 10 10 10 10 50Metastatic cancer/No cancer cancer cancer no cancer cancer cancer no cancerCognitive function dementia normal dementia dementia normal dementia

Each scenario is characterised by these four factors. For example, the first scenario described a 75-year-old woman with urosepsis and a 10%probability of short-term survival, with prior breast cancer and prior cognitive impairment; the sixth scenario described a 45-year-old womanwith urosepsis and a 50% probability of short-term survival, with no prior cancer, but with prior cognitive impairment.

169

ments agreed to participate and provided us with listsof their employees. Twelve of the 13 ICUs were runby the department of anesthesia while one was runby an internal medicine service. Eleven of the ICUswere mixed medical-surgical, one was medical andone a surgical ICU.

The questionnaires were mailed directly to all 1081physicians and critical care nurses working in the par-ticipating departments during the autumn of 1994.However, in one ICU the head nurse instead handedout the questionnaires to the nurses. Two reminderletters were mailed to non-responders. Participationin the survey was voluntary and all responses werekept confidential. The research ethics committee ofOrebro Medical Centre Hospital was approached butwaived the need for formal consent for this survey.

Statistical methodsContinuous demographic variables are expressed asmean and standard deviation (mean∫SD).

Respondent characteristics (age, sex, professionalstatus, years since graduation, years of ICU experi-ence, country of graduation, religion, ICU, and city inwhich the respondents worked) were analysed fortheir ability to predict respondents’ decisions aboutthe level of care chosen for each scenario. We first con-ducted univariate regressions for each variable usingthe level of care chosen for each scenario as the de-

P. Sjokvist et al.

pendent variable. We then conducted multivariate re-gression, including each variable that was significant(P∞0.01) in the univariate analysis.

The correlation between the answers for the ques-tions ‘‘what is in the best interest of the patient?’’ andthe question ‘‘what would you really do?’’ was ana-lysed by Cohen’s Kappa-test which quantifies agree-ment beyond chance (17).

Results

The response rate for all respondents was 850/1081(78.6%), for physicians 352/457 (77.0%) and for nurses498/624 (79.8%). The number of respondents mightbe lower than 850 for individual questions becausenot all respondents answered every question.

Table 4

Respondent characteristics that influenced level of care chosen inthe scenarios.

Discrete levels LevelVariable of variable of care P-value

a. Univariate regression analysisAge (years) ∞28 3.4 0.0006

28–32 3.632–37 3.7±37 3.4

No. years since graduation ∞6 3.6 0.00336–20 3.5±20 3.3

No. years of ICU experience ∞6 3.6 ∞0.00016–15 3.5∞15 3.3

ICU Most aggressive 3.8 0.0002Intermediate 3.6

Least aggressive 3.5

City Most aggressive 3.8 0.0148Intermediate 3.7

Least aggressive 3.4

b. Multivariate regression analysisNo. years of ICU experience ∞6 3.5 ∞0.0027

6–15 3.4∞15 3.3

ICU Most aggressive 3.6 0.0002Intermediate 3.4

Least aggressive 3.3

a. The respondent variables that were predictive of the level of carechosen in the scenarios using univariate linear regression. b. Themultivariate regression analysis in which the five variables found sig-nificant in the univariate analysis were analysed for independent pre-dictive value. Level of care 1–5 (1Ωcontinue comfort measures; 2Ωdiscontinue therapy except mechanical ventilation; 3Ωno new thera-peutic intervention; 4Ωcontinue with life support but no dialysis; 5Ωfull aggressive management).

170

Demographic dataThe demographics of the respondents are presentedin Table 1. Of the physicians, 79% were born andtrained in Sweden, while 90% of the nurses were bornin Sweden. Almost all physicians were trained inanesthesia, reflecting the organisation of critical carein Sweden. A majority of respondents answering thequestion about their religious belief, 646/834 (77.5%),described themselves as not being religious. Of theremaining 188 (22.5%) who described themselves asbeing religious and stated an affiliation, the great ma-jority described themselves as being Protestants or de-nominational Christians.

Determinants of life support limitationThe rating of importance of the 16 potential determi-nants of limiting life support is presented in Table 2.The four factors considered to be most important byboth the physicians and the nurses were: patient ad-vance directives; chances of long- and short-term sur-vival; and the age of the patient. The two factors withthe greatest difference in rating between nurses andphysicians were ‘‘risk of legal complication’’ and‘‘compliance with medical care’’, which both were rat-ed higher by the nurses. The differences betweennurses and physicians were due to professional statusrather than gender differences within each group.

ScenariosThe distribution of responses to the question of ‘‘whatwas in the best interest of the patient?’’ is presentedfor 6 of the 12 scenarios in Table 3a, while responsesto the question of ‘‘what would you actually do iffaced with such a case?’’ for the same 6 scenarios ispresented in Table 3b. The greatest variability inanswers was to the question of what was in the bestinterest of the patient. In answering this question,more than 10% of the respondents chose the extremealternatives in two scenarios. In answering the ques-tion of what they actually would do, there was noscenario in which more than 10% of the respondentspicked the extreme alternatives.

In the univariate regression analysis, when respon-dents were asked what they actually would do in acase as described in the scenario, factors that pre-dicted the level of care chosen included age of therespondent, years since graduation, years of ICU ex-perience, city and in which ICU the respondentworked (Table 4a). Professional status (nurse or physi-cian) was not associated with the level of care chosen.In the multivariate regression analysis, years of ICUexperience and ICU of the respondent remained sig-nificant (Table 4b). The two ICUs with the most ag-

Survey on life-sustaining treatment

Table 5

Agreement between questions of ‘‘Best interest of patient’’ and ‘‘Whatyou really would do’’.

Correlation coefficient* (n)

Physicians 0.69 (699)Nurses 0.58 (983)Full-time intensivists 0.86 (54)

Total 0.63 (1679)

These values represent the agreement between what respondentsbelieved was ‘‘in the best interest of the patient’’ and what they actu-ally would do for the patient described in the scenario.* According to Cohen’s kappa test where ∞0.40 represents poor0.40–0.75 fair to good and ±0.75 excellent agreement (17).

gressive approach had a mean of 3.9 while the leastaggressive ICU had 3.4 on the Likert scale rangingfrom 1 to 5.

The agreement between what the respondent reallywould do when facing the patient described in thescenario, and the level of care chosen when answeringthe question of what is in the best interest of the pa-tient is presented in Table 5. The data are reportedseparately for all physicians, all nurses and for thesubgroup of physicians who spend at least 90% oftheir clinical time in intensive care (labelled as full-time intensivists). The full-time intensivists had thegreatest agreement, while the nurses displayed thelowest rate of agreement. The overall agreement be-tween the answers to these questions was fair to good.

Discussion

This study includes utilisation of a rigorously testedinstrument for exploring attitudes about withdrawingand withholding ICU-therapy. The survey had a highresponse rate (79%) and included both nurses andphysicians. It is limited in that surveys measure whatintensive care personnel state they would do in re-sponse to scenarios, not what they actually do in prac-tice. This survey was also restricted to university-af-filiated and tertiary referral hospitals, and the extentto which these findings can be generalised to com-munity settings is unclear.

The patient-related decision determinants con-sidered to be most important when deciding to limitlife support were: likelihood of surviving current epi-sode; patient advance directives; patient age; and like-lihood of long-time survival. Age has previously beendescribed to be an important factor by Swedish healthcare workers in deciding whether life-sustaining treat-ment should be limited (18), while North Americanintensive care personnel appear to emphasise age less

171

strongly (14, 15). Acute and chronic illness or poorprognosis have also been shown to be important inother surveys of attitudes toward withdrawing of lifesupport (14, 15) and in descriptive clinical studies (6–10).

There seems to be a discrepancy between the resultsof our survey, which suggest that Swedish intensivecare personnel place a high value on advance direc-tives, and other descriptive studies of clinical practice.In two Swedish ICU studies none of the patients wasknown to have given an advance directive, and thedecision to limit life support was made without aprior documented discussion with the patient or thefamily in about half of the cases (8, 10).

If advance directives of the patient are consideredto be more important than for example the premorbidphysical function, as suggested by this survey, onewould expect that the patient or the family would beconsulted more often than 50% of the time when lifesupport decisions are made. Why are the wishes ofthe patient sought so infrequently if they are import-ant? One explanation might be that the patient’s willis considered primarily when it is spontaneously ex-pressed. Perhaps such directives are considered unre-liable when the patient’s competence is questionableand family members’ interpretation is debatable. Onemay also speculate that patriarchal attitudes not ex-pressed in this survey still influence much of clinicalpractice.

We explored differences between the levels of carechosen when the best interest of the patient was con-sidered and what the respondents stated they actuallywould have done facing a patient such as in the scen-ario. Overall there was a fair to good agreement be-tween these two approaches. However, the level ofagreement was highest among full-time intensivists,who probably also have the most powerful influenceon these decisions. In contrast, agreement betweenthese two approaches was lowest among nurses. Al-though most of the responding nurses think that thelife support decisions made in their ICUs are in thebest interest of the patients, a substantial minority donot. This suggests that many nurses disagree with life-sustaining therapy or withdrawal of life support or-dered by some physicians. Situations where nursesfeel that they have to act contrary to their own beliefshave also been described in previous studies frommedical and surgical wards (19, 20) and in the ICU(21). This might lead to substantial frustration andeven potential conflicts within the ICU-team. How-ever, the more positive finding in this study of anoverall good agreement between nurses and phys-icians in response to the question ‘‘what would you

P. Sjokvist et al.

really do?’’ is also supported by similar agreementfound in a study from another Swedish ICU (22).Further investigations are needed to elucidate if andto what extent these potential conflicts are a signifi-cant problem in caring for dying patients.

We found that the hospital in which the respondentworked was a predictor of the level of care chosen forthe scenarios. Previous studies have shown that theattitudes towards the withdrawal of life support dif-fer among physicians from different countries (11–13).In a Canadian survey, the level of care chosen in pa-tient scenarios varied between different cities andprovinces (14). In the American SUPPORT study, asubstantial variation among the five participatinghospitals in the use of do-not-resuscitate orders wasfound, indicating that the differences in attitudes be-tween centers found in this study also exist elsewhere(23).

Guidelines in Sweden and Great Britain on with-holding and withdrawal of life support describe thephysician as the decision-maker, but they also empha-sise the autonomy of the competent patient and theneed to consult the family of the unconscious patient(2, 3, 24). One way to minimise both the problem ofvariability between centers and the problem of inad-equate patient and family participation might be tomore rigorously elicit patient preferences and to en-gage in more dialogue with family members and thehealth care team. This would also bring clinical prac-tice in better agreement with existing guidelines.However, a universally acceptable approach abouthow to withdraw and withhold life support maynever be achieved and may not be an appropriate goalsince these decisions should be individualised to eachpatient.

References

1. American Thoracic Society. Withholding and withdrawinglife-sustaining therapy. Am Rev Respir Dis 1991: 144: 726–731.

2. The Swedish Medical Association. [When are physiciansallowed to refrain from treatment?] Lakartidningen 1991: 88:2165. (In Swedish).

3. The Swedish Board of Health and Welfare. [When shouldphysicians refrain from treatment?]. Nord Med 1991: 106:239–240. (In Swedish).

4. The Society of Critical Care Medicine Task Force on Ethics.Consensus report on the ethics of foregoing life-sustainingtreatments in the critically ill. Crit Care Med 1990: 18: 1435–1439.

5. Stanley JM. The Appleton Consensus: suggested interna-tional guidelines for decisions to forego medical treatment.J Med Ethics 1989: 15: 129–136.

6. Prendergast TJ, Luce JM. Increasing incidence of withhold-ing and withdrawal of life support from the critically ill. AmJ Respir Crit Care Med 1997: 155: 15–20.

172

7. Koch KA, Rodeffer HD, Wears RL. Changing patterns of ter-minal care management in an intensive care unit. Crit CareMed 1994: 22: 233–243.

8. Melltorp G, Nilstun T. Decisions to forego life-sustainingtreatment and the duty of documentation. Intensive Care Med1996: 22: 1015–1019.

9. Turner JS, Michell WL, Morgan CJ, Benatar SR. Limitationof life support: frequency and practice in a London and aCape Town intensive care unit. Intensive Care Med 1996: 22:1020–1025.

10. Sjokvist P, Sundin PO, Berggren L. Limiting life support. Ex-periences with a special protocol. Acta Anaesthesiol Scand1998: 42: 232–237.

11. Asai A, Fukuhara S, Lo B. Attitudes of Japanese and Ja-panese-American physicians towards life-sustaining treat-ment. Lancet 1995: 346: 356–359.

12. Vincent JL. European attitudes towards ethical problems inintensive care medicine: results of an ethical questionnaire.Intensive Care Med 1990: 16: 256–264.

13. Sjokvist P, Cook DJ, Berggren L, Guyatt GH. A cross culturalcomparison of attitudes toward life support limitation inSweden and Canada. Clinical Intensive Care 1998: 9: 81–85.

14. Cook DJ, Guyatt GH, Jaeschke R et al. Determinants in Can-adian health care workers of the decision to withdraw lifesupport from the critically ill. Canadian Critical Care TrialsGroup. JAMA 1995: 273: 703–708.

15. The Society of Critical Care Medicine Ethics Committe. Atti-tudes of critical care medicine professionals concerning for-going life-sustaining treatments. Crit Care Med 1992: 20: 320–326.

16. Feinstein A. Clinical sensibility. In: Feinstein A, ed. Clini-metrics. New Haven, Conn.: Yale University Press, 1987:141–166.

17. Hunt RJ. Percent agreement, Pearson’s correlation, andkappa as measures of inter-examiner reliability. J Dent Res1986: 65: 128–130.

18. Melltorp G, Nilstun T. Age and life-sustaining treatment: At-titudes of intensive care unit personnel. Acta AnaesthesiolScand 1996: 40: 904–908.

19. Uden G, Norberg A, Lindseth A, Marhaug V. Ethical reason-ing in nurses’ and physicians’ stories about care episodes. JAdv Nurs 1992: 17: 1028–1034.

20. Uden G, Norberg A, Norberg S. The stories of physicians,registered nurses and enrolled nurses about ethically diffi-cult care episodes in surgical care. Scand J Caring Sci 1995: 9:245–253.

21. Asch DA. The role of critical care nurses in euthanasia andassisted suicide. N Engl J Med 1996: 334: 1374–1379.

22. Soderberg A, Norberg A. Intensive care: situations of ethicaldifficulty. J Adv Nurs 1993: 18: 2008–2014.

23. The SUPPORT Principal Investigators. A controlled trial toimprove care for seriously ill hospitalized patients. Thestudy to understand prognoses and preferences for out-comes and risks of treatments (SUPPORT). JAMA 1995: 274:1591–1598.

24. Williams R. The ‘do not resuscitate’ decision: guidelines forpolicy in the adult. J R Coll Physicians Lond 1993: 27: 139–140.

Address:Dr. Peter SjokvistDepartment of Anesthesia & Intensive CareOrebro Medical Center HospitalS-701 85 OrebroSweden