JAGS 51:961–969, 2003© 2003 by the American Geriatrics Society 0002-8614/03/$15.00

Life-Sustaining Treatments: What Do PhysiciansWant and Do They Express Their Wishes to Others?

Joseph J. Gallo, MD, MPH,* Joseph B. Straton, MD, MS CE,* Michael J. Klag, MD, MPH,

†‡§

Lucy A. Meoni, ScM,

†

�

Daniel P. Sulmasy, OFM, MD, PhD,

¶#

Nae-yuh Wang, PhD,

†

and Daniel E. Ford, MD, MPH

†‡§

OBJECTIVES:

To assess whether older physicians havediscussed their preferences for medical care at the end oflife with their physicians, whether they have established anadvance directive, and what life-sustaining treatment theywish in the event of incapacity to make these decisions forthemselves.

DESIGN:

Mailed survey to a cohort of physicians.

SETTING:

Physicians who were medical students at theJohns Hopkins University in graduating classes from 1946to 1964.

PARTICIPANTS:

Physicians who completed the advancedirective questionnaire (mean age 68).

MEASUREMENTS:

Questionnaires were sent out toknown surviving physicians of the Precursors Study, anon-going study that began in 1946, asking physiciansabout their preferences for life-sustaining treatments.

RESULTS:

Of 999 physicians who were sent the survey,765 (77%) responded. Forty-six percent of the physiciansfelt that their own doctors were unaware of their treat-ment preferences or were not sure, and of these respon-dents, 59% had no intention of discussing their wisheswith their doctors within the next year. In contrast, 89%thought their families were probably or definitely aware oftheir preferences. Sixty-four percent reported that they hadestablished an advance directive. Compared with physi-cians without advance directives, physicians who estab-

lished an advance directive were more likely to believe thattheir doctors (odds ratio (OR)

�

3.42, 95% confidence in-terval (CI)

�

2.49–4.69) or family members (OR

�

9.58,95% CI

�

5.33–17.23) were aware of their preferences forend-of-life care and were more likely to refuse treatmentsthan those without advance directives.

CONCLUSION:

This survey of physicians calls attentionto the gap between preferences for medical care at the endof life and expressing wishes to others through discussionand advance directives, even among physicians.

J AmGeriatr Soc 51:961–969, 2003.Key words: advance directives; aged; ethics; decision mak-

ing; physicians.

ew individuals in our society are more familiar thanphysicians with the technology and factors related to

treatment decisions at the end of life. One of the key vari-ables in discussions about the validity and stability ofchoices related to end-of-life care is the extent to which theindividual has experience and knowledge with the healthconditions that might be encountered in the future. Physi-cians have a clearer understanding of the health statesothers may only imagine and may have considered whatthey would want for themselves in such circumstances.Older physicians are likely to have had not only clinicalexperience, but also personal experience with decision-making regarding their own illness or the illness of familymembers such as elderly parents. It is likely that physicians,consistent with assessments of other highly educatedgroups, would place a great value on autonomy and wouldtake measures to protect their autonomy. For these rea-sons, in the event of catastrophic illness or permanent in-capacity, if physicians do not discuss preferences or exe-cute advance directives, others without direct experienceor training related to end-of-life care will be unlikely to doso. Surveys of physicians, published between 1991 and1999, reveal that between 13% and 32% report having

From the *Department of Family Practice and Community Medicine, University of Pennsylvania School of Medicine, Philadelphia, Pennsylvania;

†

Department of Medicine, School of Medicine, and Departments of

‡

Epidemiology,

§

Health Policy and Management, and

�

Biostatistics, Bloomberg School of Public Health, The Johns Hopkins University,Baltimore, Maryland;

¶

Department of Ethics, Saint Vincent’s Hospital, Manhattan, New York, New York; and

#

The Bioethics Institute of New York Medical College, Valhalla, New York.

Presented at the Annual Meeting of the American Geriatrics Society, Nashville, Tennessee, May 2000.

Supported by National Institute on Aging Grant AG01760 and National Research Service Award AG00253-04 (Dr. Straton).

Address correspondence to Joseph J. Gallo, MD, MPH, Department of Family Practice and Community Medicine, University of Pennsylvania,3400 Spruce Street/2 Gates, Philadelphia, PA 19104.E-mail: jjgallo@mail.med.upenn.edu

F

962

GALLO ET AL.

JULY 2003–VOL. 51, NO. 7 JAGS

established an advance directive,

1,2

but studies have beenlimited by small sample sizes,

3,4

poor response rates,

2,5

andrestriction to a specific specialty or geographic region.

1,3,5–8

Although the study of what physicians want regardingend-of-life care may be important in its own right, under-standing physicians’ choices has relevance for patient care.Physicians’ attitudes appear to influence those of their pa-tients.

9

Physician treatment preference scores have beenshown to be highly positively correlated with those oftheir patients.

3

Patient willingness to withdraw life sup-port appears to be related to physician attitudes aboutend-of-life planning.

10

This study was based on responses to a questionnairefrom physicians who have participated in an ongoing ob-servational study since they were medical students at theJohns Hopkins University School of Medicine between1946 and 1964 (the Precursors Study).

11,12

The study wasinitiated by Caroline Bedell Thomas, MD, in 1946 and re-mains one of the longest continually functioning longitudi-nal studies of aging in the world. From its inception, thegoal of the Precursors Study has been to identify the psy-chological, physiological, social, and behavioral character-istics that predict morbidity and mortality. The presentinvestigation describes responses from a questionnaire mailedto the study cohort in 1998. The questionnaire includeditems about whether the study physicians thought thattheir own physicians were aware of their preferences forcare in the event of incapacity, whether their spouses or fam-ily members were aware of their preferences for care, whomtheir doctors should ask about medical care if they were tobecome incapacitated, whether the participating physicianhad or was considering obtaining a living will (LW) or du-rable power of attorney for health care (DPAHC), andwhat the physician’s preferences for care would be in theevent of brain injury resulting in incapacity.

The present investigation had four goals. First, the au-thors wished to assess the extent to which physicians haddiscussed with others end-of-life issues related to theirown personal wishes. Second, whether they had taken anyaction in the form of a formal advance directive was as-sessed. Third, the authors sought to compare the characteris-tics of physicians who reported having executed an advancedirective with those who did not. Finally, the preferencesof participating physicians for life-sustaining treatmentwere assessed, paying special attention to variation ac-cording to the age of the physician.

METHODS

The Precursors Study

Beginning with the Johns Hopkins University School ofMedicine class of 1948 and continuing through the classof 1964, baseline assessments were obtained on all 1,337students. Annually thereafter, participants were contactedwith a mailed questionnaire to detect incident disease, riskfactors, and health behaviors. The incidence of cardiovascu-lar disease and other conditions has been ascertained usingstandardized methods and endpoints.

11

In 1988, functionalmeasures were modernized with the addition of the 36-item Medical Outcomes Study Short-Form health survey(SF-36).

13,14

For any 5-year period, 87% to 94% of the co-hort has responded, with 70% to 80% of participants

completing any particular annual survey. Mortality follow-up is over 99%. If participants do not respond to the an-nual questionnaire, intensive follow-up activities (post-cards, telephone calls, National Death Index Search) areinstituted. Nonrespondents are sent a reminder postcard 2months after their questionnaire is mailed. One monthlater, telephone contact is initiated. Participants are called,reminded to return their study questionnaire, and offereda telephone interview immediately or to be scheduled for alater time. The Johns Hopkins University School of Medi-cine Joint Committee on Clinical Investigation approvedand continually reviews the Precursors Study.

Measurement Strategy: The AdvanceDirectives Questionnaire

The advance directives questionnaire (see Appendix 1) wasmailed to the Precursors participants in the fall of 1998.The first set of questions asks whether the physician re-spondent has ever had a discussion about the kind of med-ical care desired in the event of illness or incapacity, assessesthe extent to which the physician respondent believes thatpersons such as children are aware of preferences, and asksthe physician to identify whom they would want their doc-tor to ask in the event of illness or incapacity.

In the second set of questions, physician respondentswere asked to indicate whether they had established anyform of advance directive for themselves, such as a LW orDPAHC. In assessing physician respondents for decisionsregarding advance directives, the transtheoretical model ofbehavioral change was taken as a framework (see reference

15

for a review). The study cast the decision in terms ofwhether the physician had considered obtaining specificadvance directives even if they currently did not have one.

The third set of questions, on preferences in the eventof irreversible illness, were modeled after a medical direc-tive document previously published.

16

The questions inthis section asked physician-respondents to indicate thetype of intervention they would wish in the event of irre-versible brain damage resulting in inability to recognizepeople or to speak understandably (the brain injury sce-nario): “If you had brain damage or some brain diseasewhich cannot be reversed and makes you unable to recog-nize people, or to speak understandably, but you have noterminal illness and you live in this condition for a longtime, indicate your wishes regarding the use of each of thefollowing medical procedures by placing a check mark inthe appropriate column.” The 11 intervention options werecardiopulmonary resuscitation (CPR), mechanical ventila-tion, intravenous hydration, feeding tube placement, ma-jor surgery, dialysis, chemotherapy, invasive diagnostictests, blood or blood products, antibiotics, and pain medi-cations. Possible responses were “Yes, I would want,”“No, I would not want,” “Undecided,” and “Trial, butstop if no improvement.” For CPR, major surgery, and in-vasive diagnostic tests, the response category of “Trial, butstop if no improvement” was not applicable and was notoffered, consistent with the published medical directive.

16

To minimize respondent burden while still elicitingpreferences for health care in the event of incapacity, itwas necessary to select only one scenario. The brain injuryscenario was selected over others in the medical directivebecause previous research had shown that the variability

JAGS JULY 2003–VOL. 51, NO. 7

LIFE-SUSTAINING TREATMENT CHOICES OF PHYSICIANS

963

of responses was greater in the brain injury scenario thanin vignettes portraying terminal illnesses with little chanceof survival.

6,8

In other words, the brain injury scenario waschosen rather than scenarios describing a persistent vege-tative state because it was felt that the brain injury sce-nario would elicit the widest range of treatment responses.

Covariates

The association between self-reported behavior and advancedirectives was examined according to age and functional sta-tus as measured by the SF-36. The SF-36 represents eighthealth concepts, or dimensions: physical functioning, roledisability due to physical health problems, bodily pain,general health perceptions, vitality, social functioning, roledisability due to emotional problems, and general mentalhealth.

17

This instrument has been employed in studies ofoutcomes of patient care

14,17–20

and appears to be reliableand valid even in frail elders.

21

The SF-36 was scored usingpreviously described techniques.

22

The scale for each di-mension ranges from 0 to 100, with higher numbers repre-senting better health.

Analytic Strategy

Physicians who reported having an advance directive werecompared with physicians who did not, using simple chi-square or

t

tests depending on the form of the dependentvariable (proportions or continuous measures). Adjust-ments were made to estimates of association through theuse of linear multiple regression or multivariate logistic re-gression as appropriate.

23

SPSS version 10 (SPSS Inc., Chi-cago, IL) was used to perform all analyses. In these analy-ses, a level of statistical significance of

�

�

.05 was used,recognizing that tests of statistical significance are approx-imations that serve as aids to interpretation and inference.

RESULTS

Study Sample

Of the 999 questionnaires sent out to known surviving phy-sicians in the cohort, 765 returned questionnaires (77%response rate). Although questionnaire respondents wereslightly older than nonrespondents (68.0 vs 66.4), the re-spondents and nonrespondents were similar with regard tosex, and prevalence of medical comorbidity (such as can-cer, cardiovascular disease, and diabetes mellitus).

Knowledge of Others About Wishesfor End-of-Life Care

As shown in Table 1, about half of the physicians re-ported that they thought their personal doctor was defi-nitely or probably aware of preferences in the event of in-capacity to make healthcare decisions. Older respondentswere more likely to think that their own doctors wereaware of their preferences. A substantial proportion ofthe physicians reported that they have not had a discus-sion with a doctor and did not intend to do so any timesoon, with younger physicians endorsing this responsemore frequently than older physicians. Of the participantswho stated that they had had a discussion with their doc-tor, 34% reported that the initial discussion took placemore than 5 years previously, and 85% stated that they,not others, initiated the discussion. Although only 53% ofphysicians thought their own doctors were aware of theirpreferences, almost 90% thought their spouses or otherswere aware of their preferences for care at the end of life.Most of the physicians stated that in the event they wereunable to speak for themselves they would prefer thetreating doctor to ask their spouse (62% of mentions) ora child (29% of mentions).

Table 1. Physician Ratings of Others’ Knowledge of Wishes for Treatment at the End of Life

Age of Physician

58–64 65–69

�

70

TotalKnowledge of Wishes n (%)

Doctors aware of preferences* n

�

231 n

�

209 n

�

310 n

�

750Definitely aware 49 (21.2) 49 (22.9) 116 (37.4) 214 (28.5)Probably aware 64 (26.8) 51 (24.4) 77 (24.8) 190 (25.3)Not sure 50 (21.6) 46 (22.0) 54 (17.4) 150 (20.0)Not aware 70 (30.3) 63 (30.1) 63 (20.3) 196 (26.1)

Had a discussion with any doctors

†

n

�

228 n

�

202 n

�

303 n

�

733 No, and I do not intend to do so soon 107 (46.9) 83 (41.1) 92 (30.4) 282 (38.5)No, but I considered doing so 46 (20.2) 42 (20.8) 68 (22.4) 156 (21.3)No, but I intend to do so next visit 19 (8.3) 27 (13.4) 29 (9.6) 76 (10.2)Yes, I have discussed these matters 56 (25.5) 50 (24.8) 114 (37.6) 220 (30.0)

Spouse or others aware of preferences

‡

n

�

234 n

�

208 n

�

315 n

�

757Definitely aware 146 (62.4) 147 (70.7) 255 (81.0) 548 (72.4)Probably aware 49 (20.9) 39 (18.8) 40 (12.7) 128 (16.9)Not sure 22 (9.4) 12 (5.8) 8 (2.5) 42 (5.5)Not aware 17 (7.3) 10 (4.8) 12 (3.8) 39 (5.2)

*

�

2

�

23.7,

df

�

6,

P

�

.001.

†

�

2

�

23.0,

df

�

6,

P

�

.001.

‡

�

2

�

27.1,

df

�

6,

P

�

.001.

964

GALLO ET AL.

JULY 2003–VOL. 51, NO. 7 JAGS

Establishment of Advance Directives

Responses to the questions about advance directives areshown in Figure 1. In all, 60% of physicians had establisheda LW, 59% established a DPAHC, and 64% had establisheda LW or a DPAHC. There was substantial overlap in that82% of the physicians who reported they had a LW also hada DPAHC. Three physicians indicated that they had an ad-vance directive that was not a LW or DPAHC. Examinationof these responses for specification of directives other thanLW or DPAHC revealed that in one case the physician par-ticipant had written “healthcare proxy” which was coded asa DPAHC. Two other responses were recorded as no direc-tive present. In one case, the physician stated that wisheswere delineated in a “note not shown to anyone” and in theother case it was described as a “letter to son or attorney,”which was deemed too vague to count as a directive for thepurposes of this study. Older physicians were much morelikely to report having prepared an advance directive. Mostphysicians reported that advance directives were establishedafter 1990 regardless of the respondent’s age (the PatientSelf-Determination Act was passed by Congress in 1990).

Characteristics of Physicians Who HaveEstablished an Advance Directive

Table 2 compares physicians who reported they have aLW or DPAHC with physicians who reported they have

Figure 1. Physicians who reported that they did not have an ad-vance directive and had decided not to obtain one, who did nothave an advance directive but were considering obtaining one,or who reported that they had an advance directive. Percentageof physicians shown on the vertical axis. Data from the JohnsHopkins Precursors Study, 1998. LW � living will; DPAHC �durable power of attorney for health care.

Table 2. Characteristics of Physicians Who Have Established a Living Will or a Durable Power of Attorney for Health Care(DPAHC)

Reportedto Have a

Living Will orDPAHCn

�

492

No AdvanceDirectiveReported n

�

273

Age-Adjusted Odds Ratio forEstablishing an

Advance DirectiveOdds Ratio

(95% Confidence Interval)Characteristic n (%)

Divorced 106 (22.9) 63 (24.4) 0.94 (0.65–1.34)Percentage with specified advance directive who reported that:

Doctor definitely or probably knows preferences(reference group

�

not sure or doctor not aware)* 312 (64.9) 92 (34.2) 3.42 (2.49–4.69)Had a discussion with doctor about wishes (reference

group

�

did not have a discussion)* 190 (40.2) 30 (11.5) 4.93 (3.23–7.54)Spouse or others definitely or probably know preferences

(reference group

�

not sure or spouse or others not aware)* 472 (96.9) 204 (75.6) 9.58 (5.33–17.23)Percentage with specified advance directive who definitely

desired intervention or desired a trial of the interventionin the event of incapacity

Cardiopulmonary resuscitation* 16 (3.3) 22 (8.2) 0.36 (0.19–0.71)Mechanical ventilation* 33 (6.8) 36 (13.1) 0.46 (0.28–0.76)IV hydration* 126 (26.9) 93 (34.8) 0.64 (0.46–0.88)Feeding tube* 40 (8.2) 51 (19.1) 0.36 (0.23–0.59)Major surgery* 34 (7.0) 33 (12.4) 0.50 (0.30–0.83)Dialysis* 23 (4.7) 33 (12.4) 0.33 (0.19–0.57)Chemotherapy for cancer* 33 (6.8) 29 (10.9) 0.54 (0.32–0.92)Invasive diagnostic testing* 41 (8.4) 36 (13.5) 0.56 (0.35–0.91)Blood or blood products* 60 (12.3) 58 (21.7) 0.48 (0.32–0.71)Antibiotics* 118 (24.3) 89 (33.3) 0.62 (0.45–0.87)Pain medications 422 (86.7) 217 (81.6) 1.44 (0.96–2.17)

Note:

Percentages represent the proportion of persons within categories of advance directives.

*

P

�

.05.IV

�

intravenous.

JAGS JULY 2003–VOL. 51, NO. 7

LIFE-SUSTAINING TREATMENT CHOICES OF PHYSICIANS

965

no advance directives. Physicians who reported having anadvance directive were about 2 years older than other phy-sicians (mean age 68.6 vs 67.0,

P

�

.001). Age-adjustedodds ratios (ORs) are provided in Table 2 for assessingwhether the characteristics of physicians who do or do notreport having an advance directive differ significantly.Mean SF-36 scale scores were not statistically different forphysicians who had established an advance directive thanfor physicians who had not (data not shown in table), ex-cept that physicians who reported having obtained an ad-vance directive rated their mental health as significantlybetter than physicians who reported not having an ad-vance directive. Specifically, the mean mental health scorewas 86.0 for physicians who reported having established aLW or DPAHC and 83.4 for physicians who did not (

P

�

.01). Physicians who reported having an advance directivewere more likely than those without an advance directiveto say that their doctors and spouses definitely or proba-bly knew what they would want if they were unable tospeak for themselves and were more likely to report hav-ing had a discussion with their doctor. Physicians with anadvance directive were more likely than others to refuselife-sustaining treatments. For example, although intrave-nous fluids might be considered relatively noninvasivecompared with other interventions assessed, physicianswith an advance directive were less likely to want intrave-nous fluids in the given scenario (OR

�

0.66, 95% confi-dence interval (CI)

�

0.46–0.88). Although doctors whohad an advance directive were more likely to refuse life-sustaining treatments, they tended to be more likely to ac-cept the use of pain medications even if it shortened life(OR

�

1.44, 95% CI

�

0.96–2.17), although this did notreach standard levels of statistical significance.

Preferences for End-of-Life Treatment

Figure 2 shows the responses corresponding to the physi-cians’ preferences for end-of-life care in the event of irre-versible brain disease without terminal illness. In general,the physicians declined life-prolonging treatments in thesituation of brain injury presented. For example, morethan 85% of the physicians would forego cardiopulmo-nary resuscitation or mechanical ventilation. A substantialproportion of the physicians (about 10%) appeared to beundecided about what they would want. In general, theage of the participants was not associated with acceptanceor refusal of individual treatments in a consistent manner.Preferences for mechanical ventilation, intravenous hydra-tion, and pain medicines showed statistically significant dif-ferences according to the age of the physician-participant.For the three interventions, younger participants weremore likely than older participants to be undecided aboutthe interventions, whereas older participants were morelikely to definitely decline the interventions. However, forthe other eight interventions, there were no statisticallysignificant differences in preferences by age.

DISCUSSION

This study found that only 30% of the physicians hadplanned for future decision-making incapacity by havingdiscussions with their doctors about their wishes for futuremedical care. Twice as many of the physicians (64%) re-ported that they had executed a formal advance directive.

Physicians with an advance directive tended to be some-what older and were much more likely to report havingdiscussed their wishes with their doctors or with a spouseor family member. Most physicians responding to thequestionnaire indicated that they would choose to foregolife-sustaining medical treatments. In general, physicianswith an advance directive were more likely to refuse life-sustaining treatments in the event of a brain injury result-ing in chronic inability to recognize others or to speak.

Before discussing these findings in detail, the potentiallimitations of the study should be pointed out. First, thissurvey was limited to graduates from one medical schoolparticipating in a longitudinal study of aging and thusmight not be representative of all physicians. Second, thestudy did not ascertain the reasons for the physicians’ de-cisions to have a discussion related to end-of-life matterswith their family, to execute an advance directive, or forwanting to have specific interventions in the event of braindamage without hope of recovery. Third, the study did notverify the reports of the physicians regarding execution ofadvance directives (e.g., by obtaining and examining thedocuments).

Nevertheless, these findings deserve attention becausethey differ in substantial ways from other questionnairesthat have assessed the wishes of healthcare professionals inthe event of their own inability to make medical decisions.First, this study’s respondents were in the age group whenend-of-life and quality-of-life concerns become particu-larly salient, in contrast to samples of younger physicians(generally of mean age around 45 years

1,2,4,5,7,8

). It seemslikely that whether the scenario is one that is viewed as inthe far distant future rather than a more near-term possi-bility due to age would affect one’s perspectives on thetreatment one would want in the event of a serious illness.However, this study did not find strong differences in pref-

Figure 2. Preferences of physician-participants for treatmentgiven a scenario of irreversible brain injury without terminal ill-ness (see text for details). Percentage of physicians shown onthe vertical axis. For cardiopulmonary resuscitation (CPR), sur-gery, and invasive diagnostic testing, no choice for a trial oftreatment was given. Data from the Johns Hopkins PrecursorsStudy, 1998.

966

GALLO ET AL.

JULY 2003–VOL. 51, NO. 7 JAGS

erences according to age. Second, this study was nationalin scope and included a sample size larger than any studyof physicians reported previously. Third, the response rateof 77% was higher than that reported by others.1,2,4,5,7,8

Lastly, unlike other studies, attempts were made to exam-ine personal characteristics associated with having ob-tained an advance directive and whether the physician wasconsidering obtaining one.

The results of this investigation of older physicians’planning for their own end-of-life medical care are partic-ularly noteworthy in relation to the findings of the Studyto Understand Prognoses and Preferences for Outcomesand Risks of Treatment (SUPPORT).24,25 SUPPORT, astudy involving 4,301 seriously ill hospitalized patients infive academic medical centers, was unsuccessful in its mainobjective of improving care for patients through enhancedcommunication and medical decision-making. Neverthe-less, SUPPORT shed light on the effectiveness of differentmethods of advance care planning. SUPPORT demon-strated that reliance on advance directives or surrogatedecision-makers did not ensure that healthcare providersunderstood the resuscitation preferences of seriously ill pa-tients.26 Additionally, the SUPPORT investigators foundthat the strongest predictor of physician understanding ofpatients’ preferences was direct doctor-patient communi-cation about resuscitation preferences.27 In this study, itwas found that the large majority of physicians state thatthey potentially would refuse many life-sustaining treat-ments. However, most physicians report that they havenot discussed their treatment preferences with their doc-tors. Instead, the majority of physicians appeared to relyon advance directives and surrogate decision-makers, themeans of communication that were shown to be less effec-tive in SUPPORT.

Knowledge of Others About Wishes for End-of-Life CareFewer than a third of the physicians (30%) in this studyreported having discussed their wishes for future medicalcare with their own doctors, and only 29% stated thattheir doctors were definitely aware of their wishes in theevent of incapacity. Given physicians’ experience with andknowledge of the consequences of life-sustaining treat-ments, it was anticipated that most physicians in this studywould have discussed their preferences for future treatmentwith their own doctors, but the physician-participants inthis study were no more likely to report having discussedtheir wishes for future medical treatment with their owndoctors than were the participants in SUPPORT.28,29

Concerning surrogate decision-makers, the physician-participants in this investigation were likely to report thattheir spouses or family members definitely or probablyknew their wishes in the event of personal decision-makingincapacity. Specifically, 72% thought that their spouses orothers were definitely aware. In addition, spouses and chil-dren were most likely to be mentioned as the person theparticipants’ personal doctor should ask in case of inabil-ity to decide for oneself. Unfortunately, many studies,28,30–35

though not all,36,37 have found poor agreement betweensurrogate decision-makers and patients. This report doesnot have any information on the ability of surrogates topredict the preferences for life-sustaining treatment of thesurvey respondents. Yet, there is no evidence to indicate

that the surrogate decision-makers for physicians more ac-curately predict treatment preferences than surrogates fornonphysicians. The physicians in this study who had es-tablished an advance directive were much more likely thanthose who did not to indicate that their doctors or someoneelse definitely or probably knew their medical care wishes inthe event of incapacity. This finding suggests that comple-tion of an advance directive may have provided the oppor-tunity for physicians and their surrogates or physicians todiscuss treatment preferences. However, another explana-tion is that the physicians assume that completion of anadvance directive means that their surrogate decision-maker understands their treatment preferences.

Establishment of Advance DirectivesThe reported rate of having obtained an advance directivein this study (64%) was substantially higher than rates re-ported from other recent surveys of physicians (13%,1

24%,5 27%,7 and 32%2). Although the physicians in thisstudy reported a relatively high rate of advance directivecompletion, studies have shown that simple completion ofan advance directive does not effectively communicateone’s treatment preferences. SUPPORT demonstrated thatthe majority of doctors were unaware of existing advancedirectives26 and that most previously completed advancedirectives provided little useful guidance for the actualtreatment of the participating patients.38 Additionally, incases where advance directives do provide specific infor-mation, many patients reported wanting their doctors toallow surrogate decision-makers to override treatmentpreferences stated in an advance directive.39 In this presentstudy, older physicians were more likely to report havingcompleted an advance directive than younger physicians.However, regardless of age, most physicians who reportedthe year during which the advance directive was obtainedindicated they had done so during or after 1990, the yearthe Patient Self-Determination Act was passed. It cannotbe certain the extent to which the legislation has influ-enced the physician-participants to establish an advancedirective, for example, whether consideration of establish-ing an advance directive occurred in the context of hospi-talization (when the LW or DPAHC might be offered) oras a component of estate planning.

Preferences for End-of-Life TreatmentThe physicians in this study stated that they would not ac-cept most life-sustaining interventions in the event of irre-versible brain injury. The results of this study extend thefindings from previous investigations of physicians’ prefer-ences for life-sustaining treatment that also demonstratedhigh rates of treatment refusal.1,2,4–6,8,40 Physicians who re-ported having established an advance directive (either aLW or DPAHC) were more likely to refuse life-sustainingtreatments and were more likely to want to receive painmedications than those without an advance directive. Thesefindings suggest that the physicians use advance directivesas a means to limit life-sustaining treatments and to ensureaggressive pain management rather than as a means to en-sure that life-sustaining treatments are provided.

This report did not find marked differences in treat-ment refusal rates in response to the brain injury scenarioin the 1998 survey according to age. The lack of association

JAGS JULY 2003–VOL. 51, NO. 7 LIFE-SUSTAINING TREATMENT CHOICES OF PHYSICIANS 967

between older age and treatment refusal differs substan-tially from the results of SUPPORT. In analyses adjusted forpotentially influential covariates such as diagnosis and se-verity of illness, the SUPPORT investigators found that ad-vancing age was independently associated with increasedodds of refusing CPR.28 Important differences between thispresent study sample and the SUPPORT sample may ex-plain the contrast between the results of the two studies.First, the physicians in this study were relatively healthycompared with the patients in SUPPORT, who were all hos-pitalized for serious illnesses. Second, the preferences elic-ited in SUPPORT reflected the patients’ actual desire for re-suscitation in their current state of health, whereas thispresent study elicited treatment preference in response to ahypothetical scenario adapted from the medical directive.

CONCLUSIONPrevious studies have shown that the attitudes of physi-cians influence how patients plan for end-of-life medicalcare. In this present study of older physicians, it was foundthat nearly two-thirds of physicians had completed an ad-vance directive and that most physicians believed thattheir families were aware of preferences for medical treat-ment in the event of future incapacity. Most physicians re-ported that they would not accept CPR or other poten-tially life-sustaining interventions in the scenario provided,yet fewer than one-third of physicians believed that theirdoctors were aware of their treatment preferences. Thephysicians in this study did not report engaging in whatmay be the most-effective way to communicate preferencesfor future treatment, namely having a discussion withone’s doctor. This survey of physicians calls attention tothe gap between preferences for limiting medical care atthe end of life and expressing wishes to others through dis-cussion and advance directives, even by physicians.

REFERENCES1. Brunetti LL, Carperos SD, Westlund RE. Physicians’ attitudes towards liv-

ing wills and cardiopulmonary resuscitation. J Gen Intern Med 1991;6:323–329.

2. Mebane EW, Oman RF, Kroonen LT et al. The influence of physican race, age,and gender on physician attitudes toward advance care directives and prefer-ences for end-of-life decision-making. J Am Geriatr Soc 1999;47:579–591.

3. Gramelspracher GP, Zhou XH, Hanna MP et al. Preferences of physiciansand their patients for end-of-life care. J Gen Intern Med 1997;12:346–351.

4. Hillier TA, Patterson JR, Hodges MO et al. Physicians as patients. Choicesregarding their own resuscitation. Arch Intern Med 1995;155:1289–1293.

5. Blondeau D, Valois P, Keyserlingk EW et al. Comparison of patients’ andhealth care professionals’ attitudes towards advance directives. J Med Ethics1998;24:328–335.

6. Fischer GS, Alpert HR, Stoeckle JD et al. Can goals of care be used to predictintervention preferences in an advance directive? Arch Intern Med 1997;157:801–807.

7. Markson L, Clark J, Glantz L et al. The doctor’s role in discussing advancepreferences for end-of-life care. Perceptions of physicians practicing in theVA. J Am Geriatr Soc 1997;45:399–406.

8. Gillick MR, Hesse K, Mazzapica N. Medical technology at the end of life:What would physicians and nurses want for themselves? Arch Intern Med1993;153:2542–2547.

9. Schneiderman LJ, Kaplan RM, Rosenberg E et al. Do physicians’ own prefer-ences for life-sustaining treatment influence their perceptions of patients’preferences? A second look. Camb Q Healthcare Ethics 1997;6:131–137.

10. Christakis NA, Asch DA. Physician characteristics associated with decisionsto withdraw life support. Am J Public Health 1995;85:367–372.

11. Klag MJ, He J, Mead LA et al. Validity of physicians’ self-reports of cardio-vascular disease risk factors. Ann Epidemiol 1993;3:442–447.

12. Klag MJ, Ford DE, Mead LA et al. Serum cholesterol in young men and sub-sequent cardiovascular disease. N Engl J Med 1993;328:313–318.

13. Ware JE, Sherbourne CD. The MOS 36-Item Short-Form Health Survey(SF-36). Med Care 1992;30:473–480.

14. McHorney CA. Measuring and monitoring general health status in elderlypersons: Practical and methodological issues in using the SF-36 health sur-vey. Gerontologist 1996;36:571–583.

15. Prochaska JO, Redding CA, Evers KE. The transtheoretical model and stagesof change. In: Glanz K, Lewis FM, Rimer BK, eds. Health Behavior andHealth Education: Theory, Research, and Practice, 2nd Ed. San Francisco:Jossey-Bass Publishers, 1997, pp 60–84.

16. Emanuel LL, Emanuel EJ. The medical directive: A new comprehensive ad-vance care document. JAMA 1989;261:3288–3293.

17. Stewart AL, Ware JE, eds. Measuring Functioning and Well-Being. Durham,NC: Duke University Press, 1993.

18. Stewart AL, Hays RD, Ware JE, The MOS Short-form General Health Survey.Reliability and validity in a patient population. Med Care 1988;26:724–735.

19. Stewart AL, Greenfield S, Hays RD et al. Functional status and well-being ofpatients with chronic conditions: Results from the Medical Outcomes Study.JAMA 1989;262:907–913.

20. Wells K, Stewart A, Hays R et al. The functioning and well-being of de-pressed patients: Results from the Medical Outcomes Study. JAMA 1989;262:914–919.

21. Stadnyk K, Calder J, Rockwood K. Testing the measurement properties ofthe short form-36 health survey in a frail elderly population. J Clin Epide-miol 1998;51:827–835.

22. International Resource Center for Health Assessment. How to Score theMOS 36-Item (Short Form Health Survey SF-36). Boston: New EnglandMedical Center, 1991.

23. Hosmer D, Lemeshow S. Applied Logistic Regression, 2nd Ed. New York:John Wiley & Sons, Inc., 2000.

24. The SUPPORT Principal Investigators. A controlled trial to improve care forseriously ill hospitalized patients. The study to understand prognoses andpreferences for outcomes and risks of treatments (SUPPORT). JAMA 1995;274:1591–1598.

25. Lynn J, De Vries KO, Arkes HR et al. Ineffectiveness of the SUPPORT interven-tion: Review of explanations. J Am Geriatr Soc 2000;48 (5 Suppl.):S206–S213.

26. Covinsky KE, Fuller JD, Yaffe K et al. Communication and decision-makingin seriously ill patients: Findings of the SUPPORT project. The Study to Un-derstand Prognoses and Preferences for Outcomes and Risks of Treatments. JAm Geriatr Soc 2000;48 (5 Suppl.):S187–S193.

27. Wenger NS, Phillips RS, Teno JM et al. Physician understanding of patientresuscitation preferences: Insights and clinical implications. J Am Geriatr Soc2000;48 (5 Suppl.):S44–S51.

28. Hofmann JC, Wenger NS, Davis RB et al. Patient preferences for communi-cation with physicians about end-of-life decisions. SUPPORT Investigators.Study to Understand Prognoses and Preference for Outcomes and Risks ofTreatment. Ann Intern Med 1997;127:1–12.

29. Golin CE, Wenger NS, Liu H et al. A prospective study of patient-physician com-munication about resuscitation. J Am Geriatr Soc 2000;48 (5 Suppl.):S52–S60.

30. Uhlmann RF, Pearlman RA, Cain KC. Physicians’ and spouses’ predictionsof elderly patients resuscitation preferences. J Gerontol 1988;43:115–121.

31. Zweibel NR, Cassel CK. Treatment choices at the end of life: A comparisonof decisions by older patients and their physician-selected proxies. Gerontol-ogist 1989;29:615–621.

32. Seckler AB, Meier DE, Mulvihill M. Substituted judgment: How accurate areproxy decisions. Ann Intern Med 1991;115:92–98.

33. Miles SH, Koepp R, Weber EP. Advance end-of-life treatment planning: A re-search review. Arch Intern Med 1996;156:1062–1068.

34. Marbella AM, Desbiens NA, Mueller-Rizner N et al. Surrogates’ agreementwith patients’ resuscitation preferences. Effect of age, relationship, and SUP-PORT intervention. Study to Understand Prognoses and Preferences for Out-comes and Risks of Treatment. J Crit Care 1998;13:140–145.

35. Layde PM, Beam CA, Broste SK et al. Surrogates’ predictions of seriously illpatients’ resuscitation preferences. Arch Fam Med 1995;4:518–523.

36. Sulmasy DP, Terry PB, Weisman CS et al. The accuracy of substituted judg-ments in patients with terminal diagnoses. Ann Intern Med 1998;128:621–629.

37. Sulmasy DP, Haller K, Terry PB. More talk, less paper. Predicting the accu-racy of substituted judgments. Am J Med 1994;96:32–38.

38. Teno J, Lynn J, Wenger N et al. Advance directives for seriously ill hospital-ized patients. Effectiveness with the patient self-determination act and theSUPPORT intervention. SUPPORT Investigators. Study to Understand Prog-noses and Preferences for Outcomes and Risks of Treatment. J Am GeriatrSoc 1997;45:500–507.

39. Sehgal A, Galbraith A, Chesney M et al. How strictly do dialysis patientswant their advance directives followed? JAMA 1992;267:59–63.

40. Payne K, Taylor RM, Stocking C et al. Physicians’ attitudes about the care ofpatients in the persistent vegetative state: A national survey. Ann Intern Med1996;125:104–110.

968 GALLO ET AL. JULY 2003–VOL. 51, NO. 7 JAGS

Appendix 1. The Johns Hopkins Precursors Study 1998 Questionnaire On End-of-life Issues

The following questions concern healthcare decision-making if you were unable to make such decisions for yourself due toillness or incapacity. Answer these questions from the perspective of being a patient and not a physician.

1. Do you think any of your doctors are aware of your preferences about what kind of medical care you would want ifyou could not decide for yourself?� Definitely aware � probably aware � not sure � not aware

2. Have you ever had a discussion with any of your doctors about the kind of medical care you would want if you wereso ill or incapacitated that you could not make such decisions for yourself?� No, and I do not intend to do so anytime soon (not within the next year)� No, but I have considered doing so (within the past year)� No, but I intend to do so at my next visit� Yes, I have discussed these matters with my doctor

2a. If yes, when was the first time you discussed these matters with your doctor?� In the past year � 1–5 years ago � More than 5 years ago

2b. Who initiated the discussion?� Me � My physician� Someone else, specify their relationship to you

____________2c. What prompted the discussion? (please check as many boxes as applicable)

� Serious physical illness/hospitalization� Routine examination� Death or serious illness of a friend or family member� Something else, please specify: ____________

3. Do you think your spouse, child, or other significant person is aware of your preferences about what kind of medicalcare you would want if you were so ill or incapacitated that you could not make such decisions for yourself?� Definitely aware � probably aware � not sure � not aware

4. If you could not make decisions about your medical care, whom would you want your doctor to ask on your behalf?� Spouse � Son � Daughter � My attorney� Other family member, specify their relationship to you: ____________� Someone else, specify their relationship to you: ____________� My doctor should make these decisions without asking others

The next set of questions concerns advance directives. An advance directive is a written statement that tells your doctorwhat form of medical care you would accept or would refuse in specific medical circumstances. A living will is a particularkind of advance directive that allows life-sustaining therapy to be withdrawn in case one is terminally ill and cannot makehealthcare decisions for oneself. A durable power of attorney for health care (or Health Care Proxy) gives someone youdesignate the legal authority to make decisions if you are unable to do so for yourself.

5. Have you established any form of advance directive for yourself?

6. If you have an advance directive that is not a living will or DPAHC/proxy, please specify the type of advance directive.

7. Is there any additional information you can give us about plans you have made about serious illness?

8. How were your plans communicated to your family? to your physician? to others?

Whether or not you have an advance directive, we are interested in assessing your willingness to undergo certain proce-dures that are sometimes used to prolong life in the case of severe illness. We realize that answering such questions can bedifficult; however, we are trying to gauge how aggressive you feel your doctor should be.

9. Here is the scenario: If you had brain damage or some brain disease that cannot be reversed and makes you unable torecognize people or to speak understandably, but you have no terminal illness, and you live in this condition for a longtime, indicate your wishes regarding the use of each of the following medical procedures by placing a check mark in theappropriate column.

Type of advance directiveNo, I have decided

not to get oneNo, but I am

considering getting oneYes, I have this type of

advance directiveIf yes, indicate the date the advance

directive was obtained

Living will � � � ___________Durable power of attorney for

health care (DPAHC)/proxy � � � ___________

JAGS JULY 2003–VOL. 51, NO. 7 LIFE-SUSTAINING TREATMENT CHOICES OF PHYSICIANS 969

ProcedureYes, I

would wantNo, I would

not want Undecided

I would want a trial of treatment,

but stop if no clear improvement

Cardiopulmonaryresuscitation � � � N/A

Mechanical ventilation � � � �Intravenous hydration � � � �Feeding tube (gastrostomy

or jejunostomy) to provide nutrition � � � �

Major surgery such ascholecystectomy � � � N/A

Dialysis � � � �Chemotherapy for cancer � � � �Invasive diagnostic testing,

such as endoscopy � � � N/ABlood or blood products � � � �Antibiotics � � � �Pain medications, even if

they dull consciousness and indirectly shortenmy life � � � �