JAGS 50:300–306, 2002© 2002 by the American Geriatrics Society 0002-8614/02/$15.00

Completion of Advance Directives by Older Health Maintenance Organization Members: The Role of Attitudes and Beliefs Regarding Life-Sustaining Treatment

Arne Beck, PhD,* Jonathan Brown, PhD, MPP,

†

Myde Boles, PhD,

†

andPaul Barrett, MD, MSPH*

OBJECTIVES:

This study examined whether older healthmaintenance organization (HMO) members’ attitudes andbeliefs regarding life-sustaining treatment were associatedwith their completion of advance directives (ADs).

DESIGN:

A mailed survey of 1,247 subjects, aged 75 andolder, participating in a randomized trial comparing twoeducational interventions designed to increase AD comple-tion: mailed written materials versus written materials plusa videotape. AD completion was documented by medicalrecord review.

SETTING:

The initial trial took place in October 1994 ata large group-model HMO in the Denver metropolitanarea. The survey was mailed to participants 3 months afterthe trial.

PARTICIPANTS:

All HMO members aged 75 and older

who received medical care at one of the medical offices (n

�

1247) were mailed the survey. Participants were the 735members who returned the survey.

MEASUREMENTS:

AD completers were compared withnoncompleters on survey items pertaining to attitudes andbeliefs regarding life-sustaining treatment.

RESULTS:

Of survey respondents, AD completers (n

�

313) differed from noncompleters (n

�

422) in their agree-ment with the following attitudinal/belief statements: (1)“Putting my wishes for life-sustaining treatment in writingis too binding” (completers

�

12%, noncompleters

�

35%); (2) “I prefer that my family decide what kind ofmedical care is best for me if I should become unable tocommunicate my wishes” (completers

�

69%, noncom-

pleters

�

88%); and (3) “My physician clearly understandsmy wishes for life-sustaining treatment” (completers

�

76%, noncompleters

�

43%; all values,

P

< .0001).

CONCLUSIONS:

Completers were more likely to believethat their physicians understood their wishes and less

likely to think that ADs are “too binding.” More noncom-pleters wanted their “family to decide,” even though anAD would increase their families’ ability to do so. ADcompletion rates might increase if they were characterizedas a way to preserve flexibility in a complex medical sys-tem, help families reach amicable decisions on behalf oftheir loved ones, and increase patients’ confidence thattheir physician understands their wishes for life-sustainingtreatment.

J Am Geriatr Soc 50:300–306, 2002.Key words: advance directives completion; end of life care;

patient attitudes and beliefs

T

he merits of completing advance directives (ADs), in-cluding durable powers of attorney for health care

(DPAHCs) and living wills (LWs), have received significantemphasis over the past several years. Proposed advantagesinclude increased patient autonomy in end-of-life care deci-sions, clarification of patients’ preferences for end-of-lifecare to ensure care consistent with patients’ goals and val-ues, and the possibility of reduced costs of end-of-life carebecause patients often prefer limits on aggressive care.

Despite survey results from older persons endorsingthe merits of completing ADs, low rates of completionand/or placement of ADs in medical records are thenorm.

1,2

Barriers to completion include patients’ lack ofeducation or knowledge regarding ADs; difficulties withpaperwork completion; discordance between patient, fam-ily members or proxy, and provider; deferring to others(especially to one’s physician); procrastination; and lack oftime.

3–10

These findings have stimulated research on physi-cian- and patient-targeted interventions to increase com-pletion of ADs and their placement in patients’ medicalrecords.

10,11–15

Both physician- and patient-focused inter-ventions have shown modest success. The latter approachhas the additional appeal of reaching large numbers of pa-tients with relatively inexpensive mailed educational mate-rials. For example, in a randomized trial of two methodsto stimulate AD completion (mailed educational materialsvs educational materials plus a videotape on ADs), each

From the *Clinical Research Unit, Kaiser Permanente, Colorado Region; and

†

Center for Health Research, Kaiser Permanente, Northwest Division, Denver, Colorado.

This research was funded by a grant from the Garfield Memorial Fund of the Kaiser Permanente Medical Care Program.

Address correspondence to Arne Beck, PhD, Kaiser Permanente, Research and Development Department, 10350 East Dakota Avenue, Denver, CO 80231. E-mail: arne.beck@kp.org

JAGS FEBRUARY 2002–VOL. 50, NO. 2

ADVANCE DIRECTIVE COMPLETION

301

method led to a 14% increase in completion and place-ment of ADs, resulting in a final overall completion rate of33.8%.

15

Despite this moderate success, there is ample op-portunity to develop, refine, and evaluate additional strat-egies to stimulate AD completion.

Further development of strategies for increasing ADcompletion might be informed by an understanding of therelationship between patients’ attitudes and beliefs regard-ing end-of-life care and their completion of ADs. Some re-search has examined the role of knowledge, attitudes, andbeliefs regarding ADs and end-of-life care.

1,16,17

However,little work has been done to examine the relationship be-tween these factors and objectively documented AD com-pletion. The goal of this study was to explore the relation-ship between attitudes and beliefs regarding end-of-lifecare and the presence or absence of an AD in patients’medical records. Specifically, we sought to determine whetherpatient’s attitudes and beliefs regarding life-sustaining treat-ment differentiate those who do and do not complete ADs.If such attitudinal factors are associated with AD comple-tion or noncompletion, it may be possible to design educa-tional interventions more precisely tailored to reduce orremove perceived barriers to AD completion.

METHODS

Study Design

The present study was part of a randomized controlledtrial comparing two educational interventions designed toincrease AD completion: mailed written materials versuswritten materials plus a videotape on ADs.

15

In January1995, 3 months after the mailing of the educational mate-rials about ADs, we mailed a questionnaire to participat-ing patients to better understand their reasons for comple-tion or noncompletion of ADs and their attitudes andbeliefs about life-sustaining treatment. The principal ob-jective of the questionnaire was to determine whether atti-tudes and beliefs regarding end-of-life care would differen-tiate those patients who had completed and placed ADs intheir medical records from those who had not.

Advance Medical DirectivesFollow-Up Questionnaire

The questionnaire was designed to assess reasons for non-completion of ADs, attitudes and beliefs about life-sustain-ing medical treatment, length of time recently spent discuss-ing ADs with one’s healthcare provider, and demographicvariables.

If patients reported on the questionnaire that they hadnot completed an AD, they were asked to check one ormore of 12 reasons for not completing ADs that applied tothem (e.g., “Haven’t gotten around to it,” “Don’t feelcomfortable discussing it with my family or friends”). Pa-tients also had an opportunity to write an open-ended re-sponse to this question if they chose. Items were based ona review of literature on barriers to completion of ADs.

Patients were presented with an additional 12 itemsthat represented attitudes and beliefs about life-sustainingmedical treatment and were asked whether they agreed ordisagreed with the statements. Items reflected several di-mensions we thought to be important in decisions aboutlife-sustaining treatment. Item development was based on

a review of literature on ADs, review of educational mate-rials (

You and Your Choices

, published in 1994 by Choicein Dying, Inc.), and our experiences collaborating with theFoundation for Informed Medical Decision Making whilecreating the AD videotape (

Peace of Mind: Advance Direc-tives Videotape

, based on discussions of actual experiencesof patients and physicians with end-of-life treatment deci-sions). Items included the following dimensions: (1) the lo-cus of decision-making for life-sustaining treatment (e.g.,“It is up to my doctor to decide if I should receive life-sus-taining treatment if I cannot speak for myself,” “I preferthat my family decide what kind of medical care is best forme if I should become unable to communicate my wishes”),(2) perceived lack of flexibility of ADs (e.g., “Putting mywishes for life-sustaining treatment in writing is too bind-ing”), and (3) communication with family and physician(e.g., “Discussing my wishes for life-sustaining treatmentwith my family would only lead to disagreement and con-flict,” “My physician clearly understands my wishes forlife-sustaining treatment”).

The questionnaire also contained an item asking re-spondents how much time in the last 3 months “have youspent discussing advance medical directives or your wishesfor life-sustaining treatment with a health care provider atKaiser Permanente?” Four response categories were pro-vided: none, 5 minutes or less, more than 5 minutes butless than half an hour, and more than half an hour.

The last section of the questionnaire included generalhealth status and the demographic items race, education,and living arrangement.

Study Population

Eligible patients were an unselected, representative sampleof people aged 75 and older who were receiving care atone medical office of Kaiser Permanente’s Colorado Re-gion, a group-model health maintenance organization (HMO)in metropolitan Denver, Colorado, serving 338,000 mem-bers. We used an administrative database to identify all1,302 members aged 75 and older attending the HMO’sFranklin Medical Office. Physicians were asked to excludeany patients who were blind or cognitively impaired, yield-ing a sample of 1,268. Further assessment of membershipstatus led to exclusion of 34 persons who had recentlymoved out of the area, terminated Kaiser Permanente mem-bership, or died. An additional 21 patients were excludedafter randomization because their medical records couldnot be located. The final study sample consisted of 1,247patients. Patients were prerandomized to one of two inter-vention study arms to receive either mailed written materi-als or written materials plus a videotape on ADs. After theintervention, but before mailing of the AD questionnaire,patients’ medical records were reviewed to identify thepresence of completed directives. Directives included LWsand DPAHCs.

The Committee for the Protection of Human Patientsat Kaiser Permanente Northwest Division reviewed andapproved the study design, intervention materials, and im-plementation plan. Because our interventions were exclu-sively educational and voluntary, advanced informed con-sent was not required. Patients were free to ignore thematerials that we mailed to them and to refuse our requestto complete a follow-up questionnaire.

302

BECK ET AL.

FEBRUARY 2002–VOL. 50, NO. 2 JAGS

Data Analysis

Questionnaire items and AD completion status were en-tered into a microcomputer database and analyzed withSPSSpc for Windows, Version 6.1, and with PC SAS.Analyses were performed to examine differences in itemresponses for AD completers and noncompleters. Re-sponse differences for dichotomous items were tested withthe chi-square statistic, and those for ordinal categoricalitems were tested with the Mann-Whitney

U

statistic.To determine the independent contribution of atti-

tudes and beliefs to AD completion while controlling fordemographic variables possibly related to AD completion,multiple logistic regression models were constructed usingthe 12 attitude and belief items as explanatory variables.However, because the multiple logistic regression modelsexcluded all cases with any missing independent variables,the sample size was reduced to 266 (36.2%) individualswith complete data on all items, raising the possibility ofbias in the data. Missing values for these variables weretherefore imputed using the multiple imputation (MI)method.

18

This procedure was based on an explicit condi-tional multivariate model in which the 12 items were in-cluded as dependent variables and selected demographiccharacteristics were included as independent variables.These explanatory demographic variables for the imputa-tion model included predictors of missing data (completionof AD, age, highest grade), demographic characteristicspredictive of attitudes (age, gender, race), and predictors ofAD completion (white race, college education). The miss-ing values were imputed as a function of the demographiccharacteristics and the observed responses on the attitudes.

Eighty-three (11.3%) of the patients had missing datafor the demographic covariates and could not be includedin the MI procedure, but only 69 (9.4%) also had missingattitude responses. Thus, although all of the missing datawere not imputed, we were able to include the majority ofpatients (

�

90%) in the analysis.

RESULTS

Seven hundred thirty-five patients, 58.9% of the originalsample receiving the intervention, returned 3-month fol-low-up questionnaires. Table 1 displays the demographiccharacteristics of the sample. Of patients returning ques-tionnaires, 313 (42.6%) had a documented AD (a LW,DPAHC, or both) in their medical record.

Comparison of Questionnaire Respondentswith Nonrespondents

Questionnaire respondents did not differ significantly fromnonrespondents in gender (63.7% female vs 65.6%) butwere on average 0.75 years younger (

t

�

2.62,

df

(unequalvariances)

�

1,042.69,

P

�

.009). In addition, the propor-tions of patients from each arm of the randomized trial didnot differ significantly among questionnaire respondents(48.2% videotape and 51.8% written materials groups),suggesting that group assignment did not bias question-naire return. However, questionnaire respondents andnonrespondents had different AD completion rates. Non-respondents were significantly less likely to have a docu-

mented AD in their medical record than respondents(21.3% vs 42.6%, odds ratio

�

0.500, 95% CI

�

0.415–0.602,

P

�

.0001).

Reasons for Noncompletion

Table 2 shows the frequency and percentage of patientsendorsing different reasons for not completing an AD.The most frequently cited reasons for noncompletionincluded “haven’t gotten around to it” (37.9%), “notsure of my own wishes” (17.1%), “haven’t thoughtabout it” (16.8%), and “future situations are too com-plex” (13.3%).

Comparison of Advanced Directive Completersand Noncompleters

Table 3 shows frequencies of demographic and health sta-tus variables for AD completers and their percentage ofthe total for completers and noncompleters combined. Nosignificant gender, age, or self-reported health status dif-ferences were found between completers and noncompl-eters, but race and educational status were associated withcompletion. A higher proportion of whites (47.3%) thannon-whites (28.0%) were completers (

�

2

�

17.94,

df

�

1,

P

�

.0001), and a higher proportion of respondents withsome college education or who were college graduates(49.1%) than of those with less education (39.4%) werecompleters (

�

2

�

6.42,

df

�

1,

P

�

.011).In addition, completers had spent more time discuss-

ing ADs with their healthcare provider in the prior 3months (Mann-Whitney

U

test, z

�

�

3.26,

P

�

.001).

Table 1. Demographics of Population Returning Questionnaires

Characteristic Number* Percent

Age (mean

�

standard deviation) 80.5

�

4.9Gender

Male 254 36.2Female 447 63.8

RaceHispanic 54 7.6African American 56 7.9White 563 79.4Asian/Pacific Islander 9 1.3American Indian 5 0.7Other 22 3.1

Education8th grade or less 112 16.0Some high school 98 14.0High school graduate 201 28.7Some college 143 20.4College graduate 76 10.9Some graduate work 70 10.0

Health statusExcellent 39 5.5Very good 147 20.9Good 298 42.3Fair 182 25.8Poor 39 5.5

*

Numbers of subjects who responded to each question vary.

JAGS FEBRUARY 2002–VOL. 50, NO. 2

ADVANCE DIRECTIVE COMPLETION

303

Attitudes and Beliefs Associated with AdvancedDirective Noncompletion

Table 4 shows the frequency and percentages of respon-dents endorsing various attitudes and beliefs regardinglife-sustaining care, by AD completion status. The follow-ing individual items significantly differentiated AD compl-eters and noncompleters: (1) “Putting my wishes for life-sustaining treatment in writing is too binding” (completers

�

12%, noncompleters

�

34.6%), (2) “Discussing my wishesfor life-sustaining treatment with my family would onlylead to disagreement and conflict” (completers

�

6.9%,noncompleters

�

16.3%), (3) “My family understands mywishes for life-sustaining treatment” (completers

�

95.4%,noncompleters

�

89.2%), (4) “I prefer that my family de-cide what kind of medical care is best for me if I should be-come unable to communicate my wishes” (completers

�

69.2%, noncompleters

�

87.5%), and (5) “My physicianclearly understands my wishes for life-sustaining treat-ment” (completers

�

76.2%, noncompleters

�

42.5%).

Predictors of Survey Completion

Of those individuals who completed at least part of thesurvey, 469 (63.8%) failed to complete at least one of the12 questions concerning attitudes toward life-sustainingtreatment. These individuals who skipped questions weremore likely to be younger, more educated, and less likelyto have completed an AD.

Multivariate Results

Of the demographic variables, only race (white vs other)and education (high school or less vs some college ormore) were statistically significant predictors of comple-

tion of ADs in a multiple logistic regression model. Thesetwo demographic characteristics were retained in the re-gression model that was performed using the imputed dataset. All 12 attitude/belief questions were included in the re-gression model. Results are presented in Table 5. The re-sults from the MI analysis were generally similar to thoseobtained on the more limited set of patients with completedata. Race remained a significant predictor in the regres-sion model, with whites more than twice as likely to com-plete ADs as non-whites (

P

�

.025). However, college ed-ucation became nonsignificant in the regression model.“Putting my wishes for life-sustaining treatment in writingis too binding” became a stronger predictor in the MIanalysis (

P

�

.002). Survey respondents who agreed withthis statement were less than half as likely to have com-pleted an advance directive as those who disagreed. Othersignificant predictors in the regression model included “Iprefer that my family decide what kind of medical care isbest for me if I should become unable to communicate mywishes” (

P

< .001) and “My physician clearly understandsmy wishes for life-sustaining treatment” (

P

�

.001). Sur-vey respondents who agreed with the former statementwere 74% less likely to have completed an AD as thosewho disagreed with it. Respondents who agreed with thelatter statement were almost four times as likely to havecompleted an AD as those who disagreed.

DISCUSSION

Consistent with previous research, the findings from thisstudy confirm common reasons for noncompletion of ADs,including procrastination, uncertainty, lack of awareness,and appreciation of the complexity of predicting future sit-uations in which ADs might be important.

4,6

,

8–10

In addi-tion, non-white race (in particular, black) was associatedwith AD noncompletion, a finding also supported by prior

Table 2. Reasons for Not Completing an Advance Directive(N

�

422)

Reason

Numberendorsingreason* Percent

Have not gotten around to it 160 37.9Not sure what my own wishes would be 72 17.1Have not thought about it 71 16.8Future situations are too complex 56 13.3Do not think I will need it 55 13.0Did not know about it 49 11.6Do not feel comfortable discussing it with

my family or friends (or they don’t feel comfortable discussing it with me) 42 10.0

Do not want to think about death or dying 34 8.1Do not have a friend or relative that I’m

sure I can rely on 18 4.3Do not feel comfortable discussing it with

my physician 12 2.8Do not think doctors would follow my

wishes or my proxy’s decisions 10 2.4One or more of my family members

disagrees with my wishes 7 1.7

*

Subjects were allowed multiple responses; 51 did not provide any reason for notcompleting an advance directive.

Table 3. Characteristics of Advance Directive CompletersVersus Noncompleters

Characteristic

Yes (n

�

313*) No (n

�

422)

Number Percent Number Percent

GenderMale 104 41.6 146 58.4Female 198 44.9 243 55.1

Race

†

White 263 47.3 293 52.7Non-white 40 28.0 103 72.0

Education

‡

High school graduate or less 160 39.4 246 60.6

Some college or more 140 49.1 145 50.9Health status

Excellent 17 44.7 21 55.3Very good 67 46.9 76 53.1Good 122 41.2 174 58.8Fair 78 42.9 104 57.1

Poor 16 44.4 20 55.6

*

Numbers vary because of missing questionnaire data.

†

chi square,

P � .0001;‡chi square, P � .011.

304 BECK ET AL. FEBRUARY 2002–VOL. 50, NO. 2 JAGS

research.19,20 Although educational level below college hada univariate association with AD noncompletion, this vari-able became nonsignificant in the logistic regression model.

Several attitudes and beliefs differentiated AD non-completers from completers, after adjusting for race andeducation. Most notably, completers were much morelikely to believe that their physicians understood theirwishes, which was perhaps a consequence of having statedtheir wishes in writing with an AD. Moreover, completershad spent more time discussing ADs with their healthcareprovider in the prior 3 months, and the ADs may have re-sulted from providers’ suggestions that patients expresstheir wishes for end-of-life care in writing. Additionalfindings were that AD completers were much less likely tothink that ADs are “too binding” and that more noncom-pleters wanted their “family to decide,” even though anAD would presumably increase their families’ ability to doso. However, in other respects, noncompleters appearedmore realistic about the medical care system, given pub-lished research documenting lack of knowledge of patientpreferences and adherence to ADs by physicians and thedifficulty of crafting AD preference statements that areflexible enough to encompass the range of medical choicesthat might be encountered.

Several themes emerge from these findings. First, ADnoncompleters were more likely to adopt a passive role in

life-sustaining treatment decisions, preferring that theirfamily members make these decisions. Paradoxically, thepresence of an AD would aid family members in makingdecisions about life-sustaining treatment that would beconsistent with the patients’ preferences. Educational in-terventions that stress the benefit of completing an AD toprovide better tools for rather than guidance to familymembers faced with treatment decisions would seem afruitful approach. Second, noncompleters were less likelyto believe that their physicians understood their wishes forlife-sustaining treatment. In light of findings from other re-search suggesting that patients want their physicians to bemore involved in these decisions,21 AD completion couldbe marketed as a vehicle to increase physician-patient dia-logue regarding life-sustaining-treatment decision-making.Third, noncompleters were more likely to believe that exe-cuting an AD would be too binding. Clearly, educationalefforts aimed at demonstrating the flexibility of ADs andthe opportunity to change or rescind them are warranted.Studies examining change in end-of-life treatment prefer-ences over time suggest that these preferences are fairly sta-ble.22,23 Thus, reassurance to patients that their preferencesare likely to be relatively stable over time may be useful.

This study has several limitations. A baseline assess-ment of attitudes and beliefs about life-sustaining treat-ment was not made to avoid sensitizing patients to the

Table 4. Advance directive (AD) Completion by Attitude Toward Life-Sustaining Treatment

Attitude

No AD Completed AD

% Agree % Agree OR 95% CI P-Value*

It is up to my doctor to decide if I should receive life-sustaining treatment if I cannot speak for myself. 40.1 34.2 1.12 0.96–1.31 .167

I do not trust the medical system to make decisions about my medical care if I should become unable to communicate my wishes. 38.8 33.2 1.12 0.93–1.35 .231

I believe that my wishes should determine what life-sustaining treatment I receive. 94.7 94.2 1.04 0.724–1.49 .833

Putting my wishes for life-sustaining treatment in writing is too binding. 34.6 12.0 1.67 1.43–1.96 �.0001

Having a durable power of attorney for health care means I can never get certain life-sustaining treatments. 19.0 13.7 1.19 0.96–1.47 .137

Discussing my wishes for life-sustaining treatments with my family would only lead to disagreement and conflict. 16.3 6.9 1.44 1.19–1.74 .001

My family understands my wishes for life-sustaining treatment. 89.2 95.4 0.707 0.576–0.868 .007

I prefer that my family decide what kind of medical care is best for me if I should become unable to communicate my wishes. 87.5 69.2 1.79 1.38–2.33 �.0001

I can rely on my family to communicate my wishes to my doctors, even if my family does not all agree with my wishes. 92.9 93.5 0.956 0.709–1.29 .77

My physician clearly understands my wishes for life-sustaining treatment. 42.5 76.2 0.504 0.414–0.614 �.0001

I can rely on my physician to carry out my wishes for life-sustaining treatment, even if he or she disagrees with my wishes. 74.3 81.7 0.819 0.671–1.00 .064

OR � odds ratio; CI � confidence interval.

JAGS FEBRUARY 2002–VOL. 50, NO. 2 ADVANCE DIRECTIVE COMPLETION 305

outcomes being addressed in the randomized trial of theeffects of different educational materials on AD comple-tion. Lack of a baseline assessment prevented us from de-termining whether observed attitudes and beliefs werelong held or were a function of the more-recent educa-tional interventions. A related limitation is that it is impos-sible to determine whether attitudes and beliefs regardinglife-sustaining treatment preceded AD completion or werean outcome of AD completion, especially because patientsmay have responded to the questionnaire items in waysthat were consistent with their AD completion status.

Another limitation is that, although questionnaire re-sponse rates were acceptable, responders were more likelyto have completed an AD. The generalizability of these re-sults to patients who did not return questionnaires istherefore limited. Because our HMO does not routinelycapture such demographic data as race or education, wewere unable to determine whether questionnaire respon-dents were different from nonrespondents on these otherdemographic factors. Finally, although we asked physi-cians to exclude patients from the study if they had cogni-tive impairment, we were unable to measure cognitive sta-tus, because the sample was selected from administrativedatabases. Some measure of cognitive status could have re-duced the sample size beyond the reduction that resultedfrom the exclusion criteria, which might have improvedthe survey response rate.

Educational materials often emphasize the use of ADsas a means to achieve individual “remote control” of med-ical events. Medical ethicists tend to see ADs as an expres-sion of individual rights to choice and control. AD promo-tion might achieve greater success among noncompleters ifADs were characterized as a way to preserve flexibility ina complex medical system, help families reach amicabledecisions on behalf of their loved ones, and increase physi-

cians’ understanding of their patients’ wishes for life-sus-taining treatment.

REFERENCES1. Gamble ER, McDonald PJ, Lichstein PR. Knowledge, attitudes, and behavior

of elderly persons regarding living wills. Arch Intern Med 1991;151:277–280.2. Stelter KL, Elliott BA, Bruno CA. Living will completion in older adults.

Arch Intern Med 1992;152:954–959.3. Llovera I, Mandel FS, Ryan JG et al. Are emergency department patients

thinking about advance directives? Acad Emerg Med 1997;4:976–980.4. Sulmasy DP, Song KY, Marx ES et al. Strategies to promote the use of advance di-

rectives in a residency outpatient practice. J Gen Intern Med 1996;11:657–663.5. O’Brien LA, Grisso JA, Maislin G et al. Nursing home residents’ preferences

for life-sustaining treatments. JAMA 1995;274:1775–1779.6. Elpern EH, Yellen SB, Burton LA. A preliminary investigation of opinions

and behaviors regarding advance directives for medical care. Am J Crit Care1993;2:161–167.

7. Sam M, Singer PA. Canadian outpatients and advance directives: Poorknowledge and little experience but positive attitudes. Can Med Assoc J1993;148:1497–1502.

8. Glick KL, Mackay KM, Balasingam S et al. Advance directives: Barriers tocompletion. J N Y State Nurses Assoc 1998;29:4–8.

9. Palker NB, Nettles-Carson B. The prevalence of advance directives: Lessonsfrom a nursing home. Nurse Pract 1995;20:7–8.

10. High DM. Advance directives and the elderly: A study of intervention strate-gies to increase use. Gerontologist 1993;33:342–349.

11. Landry FJ, Kroenke K, Lucas C et al. Increasing the use of advance directivesin medical outpatients. J Gen Intern Med 1997;12:412–415.

12. Sachs GA, Stocking CB, Miles SH. Empowerment of the older patient: A ran-domized, controlled trial to increase discussion and use of advance directives.J Am Geriatr Soc 1992;40:269–273.

13. Rubin SM, Strull WM, Fialkow MF et al. Increasing the completion of thedurable power of attorney for health care: A randomized, controlled trial.JAMA 1994;271:209–212.

14. Hare J, Nelson C. Will outpatients complete living wills: A comparison oftwo interventions. J Gen Intern Med 1991;6:41–46.

15. Brown JB, Beck A, Boles M et al. Practical methods to increase use of ad-vance directives. J Gen Intern Med 1999;14:21–26.

16. Lo B, McLeod G, Saika G. Patient attitudes to discussing life-sustaining treat-ment. Arch Intern Med 1986;146:1613–1615.

17. Eisemann M, Richter J. Relationships between various attitudes towards self-determination in health care with special reference to an advance directive.J Med Ethics 1999;25:37–41.

Table 5. Multiple Logistic Regression Model Predicting Advance Directive Completion

Characteristic/Attitude OR 95% CI P-Value

White 2.12 1.10–4.1 .025College education 1.37 0.93–2.01 .106It is up to my doctor to decide if I should receive life-sustaining treatment if I cannot speak for

myself. 1.02 0.59–1.75 .947I do not trust the medical system to make decisions about my medical care if I should become

unable to communicate my wishes. 0.7 0.44–1.12 .139I believe that my wishes should determine what life-sustaining treatment I receive. 0.67 0.26–1.77 .422Putting my wishes for life-sustaining treatment in writing is too binding. 0.41 0.23–0.72 .002Having a durable power of attorney for health care means I can never get certain life-

sustaining treatments. 0.78 0.42–1.44 .422Discussing my wishes for life-sustaining treatment with my family would only lead to

disagreement and conflict. 0.78 0.39–1.57 .493My family understands my wishes for life-sustaining treatment. 2.33 0.96–5.63 .060I prefer that my family decide what kind of medical care is best for me if I should become

unable to communicate my wishes. 0.26 0.14–0.48 �.001I can rely on my family to communicate my wishes to my doctors, even if my family does not

all agree with my wishes. 2.19 0.84–5.72 .110My physician clearly understands my wishes for life-sustaining treatment. 3.83 0.02–6.34 �.001I can rely on my physician to carry out my wishes for life-sustaining treatment, even if he or

she disagrees with my wishes. 0.64 0.34–1.19 .161

OR � odds ratio; CI � confidence interval.

306 BECK ET AL. FEBRUARY 2002–VOL. 50, NO. 2 JAGS

18. Mesbah M, Cole BF, Lee M-L., eds. Statistical Methods for Quality of LifeStudies: Design, Measurement and Analysis. Boston, MA: Kluwer AcademicPublishers, in press.

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

20. Eleazer GP, Hornung CA, Egbert CB et al. The relationship between ethnic-

ity and advance directives in a frail older population. J Am Geriatr Soc 1996;44:938–943.

21. Johnston SC, Pfeifer MP. Patient and physician roles in end-of-life decisionmaking. End-of-life study group. J Gen Intern Med 1998;13:43–45.

22. Carmel S, Mutran EJ. Stability of elderly persons’ expressed preferences re-garding the use of life-sustaining treatments. Soc Sci Med 1999;49:303–311.

23. Emanuel LL, Emanuel EJ, Stoeckle JD et al. Advance directives: Stability ofpatients’ treatment choices. Arch Intern Med 1994;154:209–217.