Physicians’ Views on the Importance of Patient Preferences in Surrogate Decision Making

Abstract

Objective

To explore the degree to which physicians report reliance on patient preferences when making medical decisions for hospitalized patients lacking decisional capacity

Design

Cross-sectional survey

Setting

One academic and two community hospitals in a single metropolitan area.

Participants

281 physicians who recently cared for hospitalized adults

Measurements

A self-administered survey addressing physicians’ beliefs about ethical principles guiding surrogate decision making and physicians’ recent decision making experiences.

Results

Overall, 73% of physicians identified a standard related to patient preferences as the most important ethical standard for surrogate decision making (61% identified advanced directives and 12% substituted judgment). Among the 73% of physicians who reported recently making a surrogate decision, 82% reported that patient preferences were highly important in decision making. However, only 29% reported that patient preference was the most important factor in the decision. Physicians were significantly more likely to base decisions on patient preferences when the patient was in the intensive care unit (odds ratio, 2.92; 95% confidence interval (CI) 1.15-7.45) and less likely when the patient was older (odds ratio, 0.76 for each decade of age; 95% CI 0.58-0.99). The presence of a living will, prior discussions with the patient, and the physicians’ beliefs about ethical guidelines did not significantly predict the physicians’ reliance on patient preferences.

Conclusion

Although a majority of physicians identified patient preferences as the most important general ethical guideline for surrogate decision making, physicians relied on a variety of factors when making treatment decisions for a recent patient lacking decisional capacity.

INTRODUCTION

Hospitalized adults often require major medical decisions at a time when they are unable to make them. Dementia, trauma, sedation or the effects of acute illness render as many as two in five hospitalized adults unable to make their own decisions.¹ In such cases, health care providers must work with close family members or others to make surrogate decisions for patients.

There are widely accepted ethical standards for making medical decisions for an adult who lacks decisional capacity.^{2, 3} They emphasize following the patient’s previously stated preferences through advance directives if they are available, or through a substituted judgment, in which others attempt to make the decision that the patient would have made. When patient wishes are unknown, decisions should be based on what is best for the patient. This approach has been codified into many state statutes and affirmed through several major court cases. However, little is known about how often physicians apply these ethical guidelines in their clinical practice. Prior research has either been qualitative in nature⁴ or has relied on hypothetical scenarios rather than physicians’ clinical experience.⁵

To examine physicians’ approaches to surrogate decision making in the hospital setting, we administered a survey to inpatient physicians at 3 diverse hospitals in a large Midwest metropolitan area. We asked each physician to describe the most recent patient for whom they had made a surrogate decision. Because ethical models for surrogate decision making specify that patient preferences should guide decision making whenever possible, a primary goal of the study was to determine the extent to which physicians actually rely on patient preferences when making decisions in the clinical setting. We hypothesized that physicians would be more likely to rely on patient preferences when the patient had a living will, when the patient had previously expressed his or her preferences, when these preferences were judged to be helpful in the decision at hand, or when the physician endorsed an ethical standard based on patient preferences (advance directives or substituted judgment) as the best guide for surrogate decision making.

METHODS

Survey design

We designed a quantitative, written survey (see Appendix) of physicians’ experiences and beliefs about surrogate decision making, based on a review of the literature^5-9 and the results of previous semi-structured interviews with inpatient physicians.⁴ Methods are described in detail elsewhere.¹⁰

On the survey, physicians were asked to indicate whether they had made any major medical decisions for any patients within the past month who “could not participate in the decision making process. This could be for any reason, for example, due to delirium, sedation, dementia or psychosis.” We then asked physicians to recall the most such recent patient they cared for. The physician provided basic demographic information about this patient and information about the nature of the most recent decision. Physicians rated the importance of a series of potential factors that influenced their decision making and selected the most important factor in decision making for that patient (Table 3).

Table 3.

Reliance on Patient Preferences in a Recent Clinical Decision According to Patient and Physician Characteristics (N=206).^*

	Total**	Number (%) who relied on patient wishes	Bivariate p-value	Logistic regression
Physician endorsed patient preferences as most important ethical standard
Yes	130	43 (33)	0.09	1.31 (0.50-3.44)
No	54	11 (20)		Referent
Patient had a living will
Yes	19	8 (42)	0.19	2.35 (0.65-8.51)
No	163	45 (28)		Referent
Patient had a prior discussion about his or her wishes
Yes	81	26 (32)	0.40	1.10 (0.18-6.55)
No	102	27 (26)		Referent
Either prior discussion or living will judged to be helpful
Yes	73	23 (32)	0.59	0.99 (0.17-5.82)
No	108	30 (28)
DEMOGRAPHIC VARIABLES
ICU
Yes	58	25 (43)	0.005	2.92 (1.15-7.45)***
No	126	29 (23)		Referent
Age
18-30	2	1 (50)
31-40	9	3 (33)
41-50	11	6 (55)	0.04	0.76 (0.58-0.99) for each decade***
51-60	27	8 (30)
61-70	38	12 (32)
71-80	41	14 (34)
80+	55	10 (18)

^*Analysis adjusted for additional patient (sex, race, hospital) and physician (sex, race, religion) characteristics

^**Columns may not sum to 206 due to missing data

^***P<0.05

We attempted to maximize the content validity of the decision making factors by basing the factors on well-established ethical guidelines for surrogate decision making^{2, 3} by incorporating the results of previous semi-structured interviews in surrogate decision making that were conducted with inpatient physicians similar to those in the current survey⁴ by pilot testing the initial survey with 4 national experts in end-of-life care and with 7 physicians who cared for adult inpatients, and by iteratively revising the survey based on this feedback.

We included a factor related to the patient’s own preferences for care, “what the patient would have wanted you to do.” This was based on the ethical principle of respect for autonomy and reflected language used by physicians when describing recent cases.⁴ To assess the extent to which physicians relied on patient preferences, we identified the proportion of physicians who selected this as the most important factor in their recent decision. We also included a factor, “what was best for the patient overall,” which was the language most commonly used by physicians when describing decision making based on best interests.⁴ Participants were also asked a series of questions about the general principles that ought to guide surrogate decision making for hospitalized adults.

Subjects and survey administration

We administered the survey between August 2006 and April 2007 to physicians from three diverse hospitals, including an academic medical center with an internal medicine residency, a Catholic community hospital that had residency training programs in internal medicine and family medicine, and a community hospital with an internal medicine residency.

The subject pool consisted of all attending and resident physicians who worked on the internal medicine, family medicine or intensive care unit services over a consecutive 4 month period at each hospital and all private practice physicians on staff at each hospital who cared for adult inpatients during the same four month period. We identified residents and attending physicians who rotated on service using monthly inpatient schedules for the internal medicine, family medicine and intensive care units at each hospital. We contacted every physician who had worked on any of the target services for at least one week. These physicians were surveyed no more than two weeks after their service had ended. To recruit private practice physicians, we obtained the complete medical staff lists of internal medicine, family medicine, and intensive care unit physicians from the community hospitals. Private practice physicians were eligible if they had treated one or more adult patients in the inpatient setting within the last month. A member of the research team telephoned each physician at least twice to determine eligibility and to request participation in the study. A member of the research staff brought a pencil and paper survey to each physicians’ office or hospital. Surveys were collected immediately upon completion. We obtained Institutional Review Board approval at each hospital.

Data Analysis

Our primary outcome was a dichotomous variable indicating whether the physician rated patient preferences as the most important factor guiding their most recent surrogate decision. First, we examined the bivariate relationships between each physician, patient, and decision characteristic and our outcome variable. Next, we performed multiple logistic regression to examine the independent association between predictor variables and our outcome of interest. In our final models we included variables approaching significance on bivariate analysis (p<.20), key physician and patient demographic characteristics, and other variables in which we had a theoretical interest. Because each of the potential decision making factors was designed to address a unique and separate element of decision making, we examined discriminant validity¹¹ by constructing a matrix of the Spearman correlation coefficients for responses to these survey items. We expected these correlations to be low to moderate. Data was entered into a Microsoft Access database and analyses were performed using STATA 10.0 (Stata Corp., College Station, TX).

RESULTS

Response rate and physician characteristics

Five hundred and thirty potentially eligible physicians were identified during the study. Of these, 74 subjects could not be contacted and 110 were ineligible because either they had not cared for inpatients in the recent past or had left the participating hospital. Of the 346 remaining, 281 completed the survey. The minimum response based on American Association for Public Opinion Research definitions was 281/420 or 67%. (This conservative estimate assumes all 74 unreachable subjects were eligible.) Compared with non-respondents, respondents were more likely to be house staff (73% v. 61%, p=.007) and internists (78% v. 61%, p<.001) and less likely to be in private practice (76% v. 52%, p<.001). The physician sample was 46% female. Forty-seven percent were attending physicians, and of these 57% were in private practice (Table 1).

Table 1.

Physician Characteristics and Beliefs about Surrogate Decision Making Guidelines (N=281)

CHARACTERISTICS	Percent
Female	46
Race
White	56
Asian	27
African American	7
Hispanic/Latino	3
Other	7
Training level
Intern	29
Resident	22
Fellow	2
Attending	47
University or residency faculty	43
Private practice	57
DECISION MAKING GUIDELINE RATED AS MOST IMPORTANT
Patient Preferences
Advance directives, if the patient has completed them in the past	61
Substituted judgment, or what the patient would have wanted	12
The best interests of the patient	25

Ethical Guidelines for Surrogate decision making

All surveyed physicians were asked to rank the importance of several ethical standards for surrogate decision making (Table 1). We found that 73% endorsed a standard that reflected patient preferences: 61% endorsed “advance directives” as the most important standard, and 12% endorsed “substituted judgment.” An additional 25% endorsed “the best interests of the patient,”

Decision making for the most recent patient

Of all physicians surveyed, 206 of 281 physicians (73%) had made a major medical decision in the past month for a patient who lacked decision-making capacity. Patients ranged in age from 18 to over 80 years old and 60% were female. The sample was 65% African American, 30% white, 3% Hispanic/Latino and 2% Asian. One in ten of these patients had a living will and 31% of patients had named a durable power of attorney for health care. Physicians reported that 44% of patients had previously discussed their wishes with anyone prior to losing decision making capacity. When the patient had a living will or had had a prior discussion of wishes, this information was judged to be helpful by 70% and 87% of physicians, respectively. Of all patients, 80/206 (39%) had at least one source of prior information judged to be helpful

Physicians were asked to rate the importance of a series of decision making factors in the decisions they had made for their most recent patient (Table 2). Ninety-nine percent of physicians rated “the patient’s prognosis” as extremely or very important followed by “what was best for the patient overall” (98%), “respecting the patient as a person” (97%) and “the patient’s pain and suffering” (95%). Eighty-two percent of physicians rated “what the patient would have wanted you to do” as extremely or very important. Inter-item correlations between the decision making factors ranged from -0.15 to 0.53, suggesting low to moderate relationships between the decision making factors. The inter-item correlation between two items reflecting key ethical concepts, “what was best for the patient overall” and “what the patient would have wanted you to do” was 0.22, suggesting that these variables reflect distinct concepts.

Table 2.

Physician evaluations of factors related to process of decision-making for a specific hospitalized adult lacking decisional capacity (N=206 physicians).

Factor	“Extremely Important” or “Very Important”	“Most Important” of all factors listed
Patient’s prognosis	99	12
What was best for the patient overall	98	33
Respecting the patient as a person	97	5
Patient’s pain and suffering	95	13
What the patient would have wanted you to do	82	29
Providing the standard of care	81	2
Respecting the wishes of the family or surrogate(s)	81	3
Following the law	69	1
The burden on the family	45	0
Religious beliefs of the patient	35	0
Religious beliefs of the family or surrogate(s)	29	0
The cost to society of caring for the patient	14	0
Physician’s own religious beliefs	11	0
Concerns about paying for medical care	9	0
Concern that the surrogate(s) might sue	8	1

When asked to identify the single most important factor in making decisions for their patient, physicians most commonly reported “what was best for the patient overall” (33%), “what the patient would have wanted you to do” (29%), “the patient’s pain and suffering” (13%), and “the patient’s prognosis” (12%). Of the 166 subjects who rated “what the patient would have wanted you to do” as extremely or very important, 46 (28%) selected “what was best for the patient overall” as the most important factor.

Bivariate and multivariate correlates of reliance on patient preferences

In bivariate analysis, physicians were more likely to select “what the patient would have wanted you to do” as the most important factor when the patient was in the ICU (X²= 7.73 p=0.005) or was younger (X² for trend=4.01, p=0.04 for each decade, Table 3). Neither the presence of a living will, prior conversations with the patient about care, the physician’s judgment that either the living will or prior conversations were helpful, or the physicians’ beliefs about the correct ethical guidelines for surrogate decision making significantly predicted reliance on patient preferences. A similar pattern of results was found in logistic regression controlling for patient and physician characteristics.

DISCUSSION

In this survey of inpatient resident and attending physicians at three urban hospitals, physicians relied upon a variety of ethical considerations when making surrogate decisions for hospitalized adults. Consistent with standard ethical models of decision making, a vast majority of physicians rated factors reflecting patient preferences and best interest as very or extremely important. Nevertheless, when asked to identify the single most important factor, many fewer (one-third) reported they relied primarily on patient preferences to guide decision making. There was no association between reliance on patient preferences and physicians’ beliefs about the ethical principles that should guide surrogate decision making. More surprisingly, there was no association between physicians’ reliance on patient preferences and whether the patient had expressed such preferences through a living will or prior discussion regarding their care, or when the physician had judged this information to be helpful in decision making. When these documents were present and judged to be helpful, physicians still selected patient preferences as most important less than half of the time.

These data do not mean that physicians are unaware of standard ethical guidelines for surrogate decision making. Three in four respondents identified either advance directives or substituted judgment as the standard that physicians should use. The data also do not mean physicians’ routinely disregard patient preferences, nor that in reporting an alternative guiding standard, they don’t take preferences into account. However, they do raise the question of the degree to which there is concordance between ethical standards, on the one hand, and clinical practice, on the other.

There are several possible explanations for any such lack of concordance. First, in many cases information about patient’s preferences was not available due to lack of a living will or prior conversations. Second, available living wills or prior conversations about care may not apply to the clinical situation at hand.^{12, 13} They may not address the specific clinical decision that must be made or they may have been written prior to major changes in the patient’s health status that could affect decision making. This is supported by our finding that even when living wills or prior conversations had occurred, physicians relied on patient preferences less than half the time.

Third, although physicians are taught that autonomy should be given priority in both patient decision making and surrogate decision making, physicians may perceive that acting in the patient’s best interests is at least equally important. Our prior qualitative work has identified that physicians feel a sense of duty to determine the patient’s best interests and to promote them in clinical care.⁴ At the bedside, this duty may be seen as equally important as ethical teachings about autonomy. As evidence for this, even when patients had living wills and evidence of prior conversations that were judged to be helpful, physicians selected patient preferences as the most important factor less than half of the time.

Fourth, patients who select “what was best for the patient overall” may have been making a more global assessment that included both best interests and patient preferences. We found that close to a third of those who rated “what was best for the patient overall” also rated patient preferences as highly important. It is possible these physicians regarded the former item as a broader concept that included patient preferences. A few small qualitative studies have found that surrogates most often rely on either patient best interests or on a combination of best interests and patient wishes, but do not clearly distinguish between these two ethical concepts.^{14, 15} Our prior qualitative work⁴ suggests that physicians sometimes take a similar clinical approach.

We did find evidence that patient preferences were more likely to be ranked as most important in the ICU setting. In the ICU, decisions with immediate life and death consequences arise on a regular basis and patients are often unable to make decisions. It is possible that when the benefits of aggressive life-sustaining interventions are uncertain, patient preferences are given increased importance.

We also found a decreased reliance on patient preferences with increasing age. This may reflect ageist assumptions that the preferences and values of older adults are less important than those of younger adults. Another possible explanation is that physicians may ascribe to a “natural life span” perspective, which includes the belief that death is an unavoidable event and that health care in advanced age should not focus primarily on extending life.¹⁶ Some physicians may hold this view, and would place less weight on the preferences of older adults when considering aggressive medical interventions than they would for a younger person.

In recent years, the reliance on advance directives and substituted judgment has been criticized based on substantial empirical evidence.¹⁷ Studies have found that patients want physicians and family members to have some leeway in decision making,^{18, 19} that patients change their own minds about treatment over time,^20-22 and that surrogates²³ and physicians^{24, 25} are poor predictors of patient wishes. These studies were all conducted with patients who were considering a future time when they would be unable to make decisions. By contrast, our study explored the factors that practicing physicians actually used to reach surrogate decisions. We found that physicians rely on a multitude of decision making factors, and that the factor selected as most important may vary from case to case.

This study has several limitations. Understanding the basis for any behavior is a complex task, and our work examining physicians’ self-reported motivations provides an incomplete picture of clinical decision making. Other factors, such as the surrogate’s reasoning, may have been quite different from the physician’s and also may have impacted the final decision. Second, we explored the beliefs and experiences of a limited number of physicians from a single metropolitan area. Third, although we attempted to limit the potential selection of particularly memorable or outlying cases by surveying physicians close to the time of surrogate decision making, physicians may have reported their most memorable or most distressing experience rather than the most recent one. This could have introduced bias. For example, cases may be more distressing if the patient’s own wishes are unknown or are not effectively incorporated into the decision making process. Finally, because our outcome variable, reliance on patient preferences, was common (occurring 29% of the time) the odds ratios in our analysis may be more extreme than the true risk ratios.

In conclusion, although widely accepted models of surrogate decision making rank patient preferences as the most important ethical guideline for surrogate decision making^{2, 3} and a majority of house staff and attending physicians identify this as the most important guideline, physicians rely on a variety of factors when making decisions in the hospital setting. Even when helpful information about the patient’s preferences is available, physicians appear to incorporate other decision making factors that may be at least equally important. This difference between ethical theory and physician practice could encourage us either to compel physicians to weigh patient preferences more heavily in surrogate decision making or to consider whether the ethical framework for surrogate decisions should be modified to allow for balancing of multiple decision making factors.