The Johns Hopkins Precursors Study was supported by Grant AG01760 from the National Institutes of Health, Bethesda, Maryland. Data analysis manuscript preparation was supported by Grant 1-D14-HP-000084 from the Health Resources and Services Administration and by a National Research Service Award AG00253-04 granted to Dr. Straton. Presented at the American Geriatrics Society annual meeting, Washington DC, May 2002.
Address correspondence to Joseph B. Straton, MD, MSCE, Department of Family Practice and Community Medicine, University of Pennsylvania, 3400 Spruce Street/2 Gates, Philadelphia, PA 19104. E-mail: email@example.com
(See editorial comments by Dr. Linda Emanuel on pp 641–642.)
Objectives: To examine the relationship between worsening physical function and depression and preferences for life-sustaining treatment.
Design: Mailed survey of older physicians.
Setting: Longitudinal cohort study of medical students in the graduating classes from 1948 to 1964 at Johns Hopkins University.
Participants: Physicians who completed the life-sustaining treatment questionnaire in 1998 and provided information about health status in 1992 and 1998 (n=645, 83% of respondents to the 1998 questionnaire; mean age 68).
Measurements: Preferences for life-sustaining treatment, assessed using a checklist questionnaire in response to a standard vignette.
Results: Of 645 physicians, 11% experienced clinically significant decline in physical functioning, and 18% experienced worsening depression over the 6-year period. Physicians with clinically significant functional decline were more likely (adjusted odds ratio (AOR)=2.14, 95% confidence interval (CI)=1.18–3.88) to prefer high-burden life-sustaining treatment. Worsening depression substantially modified the association between declining functioning and treatment preferences. Physicians with declining functioning and worsening depression were more likely (AOR=5.33, 95% CI=1.60–17.8) to prefer high-burden treatment than respondents without declining function or worsening depression.
Conclusion: This study calls attention to the need for clinical reassessment of preferences for potentially life-sustaining treatment when health has declined to prevent underestimating the preferences of older patients.
When facing life-threatening illness, patients and surrogate decision-makers must make a variety of choices about treatment options, particularly choices to forgo or withdraw life-sustaining treatment. To understand patient wishes in the event of incapacity and to provide guidance for surrogate decision-makers, physicians are encouraged to elicit their patients' preferences for life-sustaining treatments in advance of potentially life-threatening situations.1
Clinical discussions of preferences for treatment desired in the event of incapacity assume that treatment preferences remain stable over time,2,3 but the results of several studies are mixed concerning the relationship between changing health and life-sustaining treatment preferences.4–8 Nevertheless, clinicians and surrogate decision-makers must make life-sustaining treatment decisions for patients unable to make decisions for themselves. To make decisions that accurately reflect the wishes of patients, clinicians need information on the relationship between worsening health and life-sustaining treatment preferences.
The main objective of this study was to determine whether older persons who have experienced deteriorating physical function and worsening mental health stated preferences for life-sustaining treatment that were different from the preferences of persons who experienced only declining physical functioning or only worsening mental health or whose health was not declining. To address this goal, physical health and depression were prospectively assessed over a 6-year period in a large cohort of older physicians, and preferences for life-sustaining treatment were assessed using a scenario adapted from the medical directive.9
The Johns Hopkins Precursors Study
The late Caroline Bedell Thomas designed The Johns Hopkins Precursors Study in 1946 to identify characteristics associated with premature cardiovascular disease and death.10 All 1,337 students who matriculated into the graduating classes of 1948 to 1964 of the Johns Hopkins University School of Medicine were eligible for the study. Ninety-five percent of the eligible students were enrolled during medical school. Through annual mailed questionnaires, the cohort has been followed since graduation, with a response rate of more than 70% each year. The Johns Hopkins School of Medicine institutional review board approved and continually reviews the Precursors Study.
Assessment of Changes in Physical Functioning and Depression
Physical functioning and depression were assessed in 1992 and 1998. Changes in self-reported physical functioning and depression were measured using the physical functioning and mental health subscales of the Medical Outcomes Study 36-item Short Form (SF-36).11 The SF-36 was administered to the Precursors Study cohort as part of the annual questionnaires in 1992 and 1998 and was scored using standard techniques.11 The SF-36 assesses eight health domains (physical functioning, role disability due to physical health problems, bodily pain, general health perceptions, vitality, social functioning, role disability due to emotional problems, and general mental health),11 has been employed in studies of patient care outcomes,11,12 and appears to be reliable and valid even in frail elders.13 The mental health subscale, known as the five-item version of the Mental Health Inventory, has been found to be an accurate screening instrument for the detection of major depression and other affective disorders.14
Changes in physical functioning and depression were calculated by determining the difference between the 1992 and 1998 scores in the physical functioning subscale and the mental health subscale for each participant. Although changes in these subscales could be analyzed as continuous variables, it was believed that the results would be more clinically relevant if each change score was dichotomized at a threshold representing a clinically meaningful decline. Because there are no standardized thresholds for dichotomized change scores on the SF-36 subscales, thresholds for physical functioning and depression were established a priori based on clinically significant thresholds reported in a psychometric analysis of the SF-36.15 Based upon the psychometric analysis, the physical functioning change score was dichotomized at 20 (i.e., a decline of 20 or more points on the physical functioning subscale between the 1992 and 1998 questionnaires represented a clinically meaningful decline in physical functioning) and the depression change score at five.
Assessment of Treatment Preferences
In the 1998 questionnaire, the participants were asked to consider what types of treatment they would want if they suffered irreversible brain damage that left them unable to speak understandably or to recognize people (instrument published in a previous paper16). In the scenario, based on the medical directive,9 the participants were told that they had no terminal illness and might remain in this brain-injured state for a long period of time (the brain injury scenario). To reduce respondent burden, the assessment of treatment preferences was limited to one scenario. The scenario of irreversible brain damage without terminal illness was selected because previous research demonstrated that this scenario provided a greater degree of variability of responses than other vignettes in the medical directive.9,17 The participants were asked to state their wishes regarding the use of 10 medical interventions. For each intervention, the response categories were: “Yes, I would want,”“No, I would not want,” Undecided,” and “Trial, but stop if no improvement.” The responses for each intervention were coded into a dichotomous variable as reject (No, I would not want) or accept (all other choices). Other investigators have used this dichotomization, which reflects common clinical practice in which Trial, but stop if no improvement and Undecided would translate into providing life-sustaining treatment to incompetent patients, at least initially.17,18 Participant responses to individual treatments and the rationale for focusing on the brain injury scenario were described in a previous paper.16
The dependent variable, treatment preferences, was operationalized based upon the interventions accepted by each participant in response to the brain injury scenario. Life-sustaining treatments can be characterized as more or less burdensome based upon the invasiveness or discomfort caused by the treatment.19 Consistent with investigators who found that acceptance of more-invasive treatments predicted acceptance of less-invasive treatments,18 each participant was categorized based upon the most-burdensome intervention accepted by the participant.17,18 The algorithm employed to code the dependent variable is depicted in Figure 1.
To account for other factors that might affect changes in physical functioning or depression and confound the association between changing health and life-sustaining treatment preferences, the influence of all SF-36 subscales and medical comorbidity on the association between changing health and treatment preferences was investigated using multivariate models. The SF-36 subscales were included in models as continuous variables.20 Baseline medical comorbidity was measured by summing the presence in 1992 of 11 chronic diseases or conditions. Information about these diseases or conditions was gathered at study enrollment and during follow-up from annual questionnaires and medical records.
The analysis proceeded in three phases. The first phase consisted of calculating descriptive statistics for the participants according to life-sustaining treatment preferences. The associations were analyzed using tests for trend.21 The second phase consisted of conducting multivariate analyses modeling the association between treatment preferences and clinically significant physical functioning decline. All multivariate models were adjusted for age and sex, and for physical functioning in 1992.
The multinomial logistic regression model22 was employed to estimate the relative odds for preferring higher burden treatment to no treatment for persons who experienced clinically meaningful physical functioning decline compared with persons without such decline. The multinomial logistic regression model is an extension of the logistic regression model for binary dependent variables and is well suited to the situation in which the dependent variable consists of three levels (in this case, three levels of treatment preference).
In the third phase of the analysis, a term representing the interaction between worsening depression and declining physical functioning was included in the models. The final data analysis model was a simplification of other models in which terms were dropped from the analysis when it was found that their inclusion had little effect on the odds ratios (ORs) of interest. For these analyses, a level of statistical significance set at α=0.05 was employed, recognizing that tests of statistical significance are approximations that serve as aids to interpretation and inference. Data analysis was performed using STATA 7.0 for Windows (STATA Corp., College Station, TX).
Of the 999 participants to whom the 1998 questionnaires were sent, 769 responded with information regarding medical interventions for the brain injury scenario. Of these, comorbidity data and complete SF-36 data from both the 1992 and 1998 questionnaires were available for 645 physicians. These 645 physicians constituted the study sample for the investigation. This sample represents 84% of the respondents to the brain injury scenario in the 1998 questionnaire and 65% of all eligible participants. The mean age±standard deviation of the study sample was 68±5.3 (range 58–85). Reflecting medical school enrollment between 1948 and 1964, the study sample is predominantly male and nonminority. Compared with the 645 participants in the study sample, the 124 participants omitted from the sample for missing SF-36 or comorbidity data were younger (mean age of 68 for group with complete data vs 66 for group with missing data, P<.001) but otherwise were similar to the study sample in terms of sex, baseline health status, and treatment preferences.
Association Between Personal Characteristics and Treatment Preference
Characteristics of the participants according to life-sustaining treatment preferences are shown in Table 1. In response to the scenario on severe brain injury, 47% of the physicians refused all interventions except for pain medication and were classified as desiring no treatment. Other participants were classified as desiring low-burden treatment (17% of all participants) and high-burden treatment (36% of all participants). Respondents in all three categories of treatment preferences were similar according to age, sex, and medical comorbidity. Preferences for higher-burden treatment were associated with worse concomitant physical functioning and vitality scores as reported on the SF-36.
Table 1. Characteristics by Life-Sustaining Treatment Preference in 1998
High-Burden Treatment n=234 (36%)
Low-Burden Treatment n=108 (17%)
No Treatment n=303 (47%)
Note: Test statistics were calculated using tests for trend. Data from the Johns Hopkins Precursors Study, 1992 to 1998.
Men, n (%)
White, n (%)
Health status, mean±SD
Comorbid conditions, 1992
Physical functioning, 1998
Bodily pain, 1998
Role physical, 1998
Role emotional, 1998
General health perception, 1998
Health status decline, 1992–1998, n (%)
Declining physical functioning
Association Between Declining Function and Treatment Preference
The association between declining physical function and life-sustaining treatment preferences was evaluated using multinomial logistic regression. The estimated ORs from multinomial logistic regression analysis, adjusting for baseline physical functioning, age, sex, and bodily pain, indicated that decline in physical functioning was independently associated with preferences for higher-burden treatment (data not shown). Specifically, physicians who experienced clinically significant physical functioning decline were more likely (adjusted odds ratio (AOR)=2.14, 95% confidence interval (CI)=1.18–3.88) than physicians who did not experience clinically significant physical functioning decline to prefer high-burden treatment rather than no treatment. Similarly, participants who experienced clinically significant physical functioning decline were more likely (AOR=2.05, 95% CI=0.97–4.36) to prefer low-burden than no treatment, although these results were not statistically significant.
Effect Modification by Depression
There was substantial effect modification involving worsening depression scores and declining function (Table 2) even after adjustment for age, sex, medical comorbidity, and physical functioning and depression at baseline. Compared with physician-participants without declining function or worsening depression, the estimated OR of preferring high-burden treatment compared with no treatment was 5.33 for participants reporting both declining function and worsening depression (95% CI=1.60–17.8). In the regression model that includes a term for effect modification of worsening depression and declining function, preferences for high-burden treatment compared with no treatment were not significantly predicted by only declining physical functioning (OR=1.71, 95% CI=0.87–3.32) or only worsening depression (OR=1.32, 95% CI=0.80–2.18). Similarly, the estimated odds of preferring low-burden treatments compared with no treatment were 6.68 for physicians who experienced both declining function and worsening depression (95% CI=1.72–25.9), whereas the OR was less than 1.5 for those experiencing decline in only one domain or the other.
Table 2. Modification of the Association Between Declining Physical Functioning and Preferences for Life-Sustaining Treatment by Worsening Depression
Physical Functioning Decline/Worsened Depression: OR (95% CI)
Note: Estimates of odds of preferring higher burden treatment, modeled with an interaction of declining physical functioning and worsening depression over a 6-year period, adjusting for age, sex, medical comorbidity, baseline physical functioning, baseline depression, and bodily pain. Data from the Johns Hopkins Precursors Study, 1992 to 1998.OR=odds ratio; CI=confidence interval.
None (reference category)
In this large, longitudinal study of older physicians, about one-third of physicians preferred to have high-intensity life-sustaining treatment, and these preferences were associated with having experienced a clinically significant decline in physical functioning. The association persisted after controlling for potential confounders. Additionally, it was found that physicians who experienced both physical functioning decline and worsening depression were much more likely to prefer high-burden treatment than the group without declining function or depression.
This study of the relationship between worsening health and life-sustaining treatment preferences addressed many limitations of previous investigations. Because of cross-sectional designs, many previous studies were unable to assess changes in functional status.23,24 Additionally, several prospective studies suffered from low response rates4 or had study samples limited to seriously ill patients5 or to young patients with acquired immunodeficiency syndrome.7,8 Finally, although previous prospective studies have investigated the associations between treatment preferences and changes in physical and mental health,5,6 no previous studies have reported on the interaction between worsening depression and declining physical functioning in association with life-sustaining treatment preferences.
Three studies that have examined the relationship between changes in physical health and depression and preferences for life-sustaining treatment produced conflicting results.5,6,8 A study of patients who were infected with the human immunodeficiency virus found that changing emotional status did not affect the stability of treatment preferences but that change in functional status was associated with preferences for more treatment at follow-up than at baseline.8 However, this finding did not differentiate persons with declining function from those with improving function.8 In the Study to Understand Prognoses and Preferences for Outcomes and Risks of Treatments (SUPPORT), investigators found no effect of changes in function on the stability of resuscitation preferences but found that improvement in depression was associated with increased likelihood of preferring resuscitation at follow-up.5 Findings from SUPPORT are inconsistent with the findings of the first study,8 but the samples in the two studies were different. In the first study, the patients were relatively young and not hospitalized.8 The patients in SUPPORT were seriously ill and hospitalized at the baseline interview and were reinterviewed just 2 months later. The third study, a large, 2-year prospective study of the stability of preferences for life-sustaining treatment in community-dwelling older patients, found that respondents who preferred more interventions at follow-up than at baseline were more likely to have experienced worsening physical mobility or worsening mental health.6 These three studies evaluated the effect of changing health on change in treatment preferences, whereas the current study examined the association between changes in health and the burden associated with preferred treatments assessed at one point in time. Nonetheless, the current study expands on this previous work and perhaps explains the contradictions of prior studies by demonstrating that the interaction between changes in physical and mental health has a greater association with treatment preferences than changes in function or depression alone.
Several studies have focused on the relationship between recovery from episodes of major depression and treatment preferences.25–28 Although one study found no influence of recovery from depression on treatment preferences,28 three studies found that participants were more likely to decline resuscitation when depressed than after recovery from depression.25–27 The differences in the levels of depression of the study samples chosen may explain the contradiction between studies finding preferences for more treatment after recovery from depression and the current investigation demonstrating the relationship between worsening depression and functioning and preferences for more treatment. The hopelessness that frequently afflicts persons with major depression may account for the wish to forgo cardiopulmonary resuscitation (CPR) associated with major depression and the change to acceptance of CPR when depression improves. However, for persons not suffering from major depression, worsening of depressive symptoms may reflect a sense of increasing vulnerability. Although patients respond to hopelessness by preferring less life-sustaining treatment, patients may respond to vulnerability by stating preferences for more life-sustaining treatment, as suggested by the third study.6 It can be hypothesized that the interaction between worsening physical function and depression found in the current study occurred because the physicians who experienced decline in both physical and mental health felt more vulnerable than others and responded to the vulnerability by choosing more intensive treatment.
The current study findings must be considered in the context of potential limitations. First, the study sample was a cohort of predominantly male older physicians who graduated from one medical school and have participated in a longitudinal study. The generalizability of the results may be limited to physicians or to men of high socioeconomic status. Nevertheless, because older physicians are highly likely to have had professional experience with the life-sustaining treatments and the complications of severe irreversible brain damage as presented in the questionnaire, participant misunderstanding of the conditions and interventions asked about in the questionnaire are unlikely to have influenced the results. Second, the measures of physical functioning and depression were based on subjective assessment by the respondents and not on objective evaluation. Yet, the participants were physicians who have been shown to provide accurate reports of their health.29 Third, it was decided to categorize the life-sustaining treatments according to the burden of the interventions. The potential for misclassification bias exists because of the categorization of life-sustaining interventions into three levels of treatment burden. Although there is no standard method for categorizing life-sustaining treatments, the methodology was consistent with that of other investigators.17,18 Moreover, by separating the interventions into categories according to the potential burden each intervention imposes, this categorization is more clinically relevant than methods that evaluated the difference in the actual number of interventions accepted6,28 or assessed only acceptance or rejection of CPR.5,25,30 Fourth, as addressed earlier, information about life-sustaining treatment preferences before 1998 is lacking, and it was not possible to track changes in preferences over time. Fifth, by eliciting treatment preferences using a hypothetical illness scenario, the preferences elicited might not accurately predict treatment decisions during actual illness. However, because hypothetical scenarios are frequently used in advance directive documents, this methodology reflects common medical practice regarding planning for end-of-life care.
Understanding the relationship between changes in health and older persons' treatment preferences has important clinical and policy implications. This study poses a challenge to the use of written documents for advance care planning, because many standard documents are unresponsive to changing health status. On the clinical side, these findings call attention to the need for continuous discussion of preferences for care, particularly in the context of dynamic changes in health status. With regard to policy implications, it is necessary to question the role of advance care planning that does not incorporate a means of reevaluating preferences for future medical care in light of changing health conditions. Further research should seek to clarify how the full spectrum of physical and mental health influences patient preferences for end-of-life medical care to better inform practitioners who must make clinical decisions in the face of uncertainty.