Older Adults without Desired Surrogates in a Nationally Representative Sample

Abstract

Background/Objectives:

Little is known about older adults who have intact capacity, but do not have a desired surrogate to make decisions if their capacity became impaired.

Design:

Cross-sectional study of a nationally representative sample.

Setting and Participants:

Community-dwelling older adults without known cognitive impairment, aged 57–85, interviewed as part of the National Social Life, Health, and Aging Project (NSHAP) from 2005–2006 (n=2,767).

Measurements:

We examined demographic, medical, and social connectedness characteristics associated with answering “no” to this question: “do you have someone who you would like to make medical decisions for you if you were unable, as for example if you were seriously injured or very sick?” Because many states permit nuclear family to make decisions for persons with no legally appointed health care agent, we used logistic regression to identify factors associated with individuals who were ill-suited to this paradigm in the sense that they had nuclear family, but did not have a desired surrogate.

Results:

Among NSHAP respondents, 7.5% (95% CI, 6.4–8.7) did not have a desired surrogate. Nearly 90% of respondents without desired surrogates had nuclear family. Compared to respondents with desired surrogates, those without desired surrogates had lower indicators of social connectedness. On average, however, they had four confidantes, approximately 70% socialized at least monthly, and more than 90% could discuss their health with a confidante. Among respondents who had nuclear family, few characteristics distinguished those with and without desired surrogates.

Conclusion:

Nearly 8% of older adults did not have a desired surrogate. Most had nuclear family and were not socially disconnected. Older adults should be asked explicitly about a desired surrogate, and strategies are needed to identify surrogates for those who do not have family or would not choose family to make decisions for them.

Introduction

The term “adult orphan” has been used to describe older persons who are able to make their own medical decisions but would not have a surrogate decision-maker if their capacity became impaired.^1–3 When they lose capacity, such individuals are said to be “unbefriended,” and require a stranger, such as a professional guardian, to make decisions on their behalf.^4–7

The words “adult orphan” and “unbefriended” give the impression that lacking a surrogate is an issue that affects people who do not have family or friends. These terms reinforce the common belief that, for an individual with family, a surrogate decision-maker will be available if one is needed. This is the implicit view of the default surrogate consent laws that are present in most states. These laws empower family members to make decisions for an incapacitated person who has not appointed a health care agent.⁸

Our previous research suggests that there are persons for whom the situation is more complex. In work on older adults who had become “unbefriended” and required a professional guardian to make decisions, we learned that two-thirds, in fact, had family, but their family members were unwilling, inappropriate, or unavailable to make decisions.⁹ Virtually nothing is known about older persons who still have capacity, but do not have a desired surrogate in the sense that there is no one whom they would choose to make decisions for them if their capacity were to become diminished. While the absence of a desired surrogate may be an issue faced primarily by the growing number of aging Americans without nuclear family,^{10, 11} clinical experience suggests that there are also individuals with family who would not want their family members to make decisions on their behalf, or whose family members do not want this role. Others have family members who, for a variety of reasons, would not be able to carry out the responsibility of making decisions that reflect their loved ones’ values and interests.

An improved understanding of this phenomenon is needed in the clinical setting, where surrogate decision-making is common. Nearly half of hospitalized older adults, for example, require assistance from another person to make health care decisions.¹² In the present study, we sought to characterize older adults with capacity, but without a desired surrogate, using nationally representative data from the National Social Life, Health, and Aging Project (NSHAP). We also sought to identify factors associated with persons who do not fit in the current decision-making paradigm in the sense that they reported that they did not have a desired surrogate but had at least one nuclear family member who would be legally recognized to make decisions in most states.

Methods

Data Sources and Study Population

We used data from Wave 1 of NSHAP, a study designed to explicate the role of social support and intimate relationships in aging. Details of the NSHAP study design have been described previously.¹³ Briefly, investigators drew upon the household screening carried out by the Health and Retirement Study in 2004 to identify a nationally representative multistage probability sample of community-dwelling persons age 57 to 85. African Americans, Hispanics, men, and the oldest old (75–84 years of age at time of screening) were oversampled. Respondents were excluded if they could not provide informed consent or if the interviewer felt the respondent had inadequate cognitive ability to complete the survey.¹⁴ Of 4,017 eligible persons, 3,005 were successfully interviewed, yielding a weighted response rate of 75.5%. In-home interviews were conducted in English and Spanish by professional interviewers between July 2005 and March 2006. To minimize the length of the in-person interview, additional questions were asked via a paper questionnaire that interviewers left behind for respondents to complete; the return rate for this questionnaire was 84%. Respondents were randomly assigned to six paths in which different sections of the survey were administered via the questionnaire, rather than in person.¹⁵

For the present study, our sample consisted of the 2,767 respondents (92.1%) with data about our outcome of interest. Because it involved secondary analysis of data that are publicly available and deidentified, the research plan was deemed exempt from review by the institutional review board at Yale University.

Outcome

The outcome of interest was a respondent’s answer to the following question: “do you have someone who you would like to make medical decisions for you if you were unable, as for example if you were seriously injured or very sick?”¹⁶ Those who answered “yes” were categorized as having a desired surrogate. Those who answered “no” were categorized as not having a desired surrogate.

Among respondents, 2,005 answered this question in person, while 762 completed the leave-behind questionnaire. Missing data for the outcome variable were more common among those who received the leave-behind questionnaire than those who answered the question in person (22% versus 1%), with respondents who were male, non-white, had lower education levels, and reported poorer self-rated physical and mental health more likely to be missing outcome data. We determined that these associations were consistent with the missing at random assumption and included these variables a priori in all analyses.

Social Connection: Relationship Structure, Function, and Quality

To assess the complex, multifactorial ways in which each respondent was connected to other people, we used the framework put forward by the National Academy of Medicine.¹⁷ This framework organizes an individual’s social connectedness in terms of structural, functional, and quality indicators.¹⁸ Structural indicators of social connectedness are quantitative and refer to the type, number, and strength of an individual’s relationships. Functional indicators measure the aid and other resources that relationships provide to an individual. Quality indicators capture the positive and negative ways that social relationships contribute to an individual’s life.

The presence or absence of nuclear family was a key structural indicator because of the important role of nuclear family members in state default consent statutes. To determine this, we utilized the social network rosters collected in NSHAP. Respondents were asked to list up to five people with whom they had discussed “important matters” within the past 12 months. Additional prompts were used to identify anyone else “who is very important to you” as well as any other household members. The nature of the respondent’s relationship to each person was recorded. We categorized a respondent as having nuclear family if a spouse, parent, sibling, or child was listed on the network roster or if a respondent had indicated elsewhere in the survey that he or she was married or had living children. We considered a respondent to have confidantes, but no nuclear family, if the network roster included only persons who were not nuclear family members and the respondent was unmarried and had no living children. We determined that a respondent had neither nuclear family nor confidantes if there was no one listed on the network roster and he or she was unmarried and had no living children.

We examined several additional structural indicators. We ascertained each respondent’s social network size, defined as the total number of individuals named in the network rosters. We also calculated network density, which is a measure of how connected each member of a person’s social network is with the other members of the social network. We used the standard method for undirected ties, 2T/N(N-1), where T is the number of ties between individuals in the network and N is the number of individuals in the network.¹⁹ Values are continuous between 0 and 1, where 0 indicates that no possible pairs of network members have any contact and 1 indicates that all possible pairs have maximum contact. We also measured community involvement using survey items about the frequency of attendance at religious services, attendance at community group meetings, and volunteer work in the past 12 months. These items were combined into a scale ranging from 0 (no participation) to 18 (frequent participation) in these activities. We assessed social activity using a survey item about how frequently respondents had socialized with friends or relatives in the past year, on a scale ranging from 0 (hardly ever) to 6 (several times a week). Finally, we identified residents who lived alone, versus having others in their household.

We identified three functional indicators of social connectedness in the NSHAP data. We ascertained a respondent’s social support using questions about whether he or she could rely on or open up to a spouse, family, or friends. Following previous work,²⁰ we created a composite scale ranging from 0 (no social support) to 18 (maximum support from all three sources). Because unmarried adults were asked these items for friends and family only, their scores were averaged and rescaled. Loneliness, a concept that is distinct from social isolation because it represents the perceived absence of support, was measured using a shortened version of the Revised UCLA Loneliness Scale in which respondents were asked to rate how often they felt a lack of companionship, left out, or isolated from others, with scores ranging from 3 (never or hardly ever) to 9 (very often).²¹ Finally, we determined whether respondents had any confidante with whom they could talk about their health issues, using an item asked separately from the question about whether they had a desired surrogate.

Finally, we examined several indicators of the quality of the respondent’s social connections. We measured social strain using questions about how often the respondent’s spouse, family and friends either (1) made too many demands, (2) were critical of them, or (3) got on their nerves. These items were combined into a composite scale ranging from 0 (no strain) to 27 (frequent social strain). Scores for unmarried adults were again averaged and rescaled. We also determined whether a respondent had suffered elder abuse, defined as either verbal, physical, or financial mistreatment in the past 12 months.²²

Chronic Medical Conditions

To measure a respondent’s chronic medical conditions, we calculated the modified Charlson Comorbidity Index (CCI), previously adapted for NSHAP.²³ Modified CCI scores ranged from 0–16.

Missing Data

Missing data for independent variables ranged from less than 1% to nearly 10% and were handled with multiple imputation using the fully conditional specification method (M=10) implemented by the multiple imputation procedure in SAS 9.4 (SAS Institute, Cary, NC), which was used to perform all analyses.²⁴ Rubin’s formulas were used to combine model estimates into a single set of results using the MIANALYZE procedure in SAS.²⁵

Statistical Analysis

We described the demographic characteristics, health, and social connectedness of respondents with and without desired surrogates. We summarized categorical variables with counts and percentages and continuous variables with means and standard deviations. We then examined bivariate associations between these characteristics and the presence of a desired surrogate, using chi-square tests for categorical variables and t-tests for continuous variables.

We conducted a logistic regression analysis among the subgroup of respondents who had at least one nuclear family member (n=2,694) to identify factors associated with those who did not have a desired surrogate. Such persons are ill-matched to the decision-making paradigm in most states, which empower nuclear family members to make decisions for individuals who have not appointed a health care agent. We selected explanatory variables a priori, using clinical judgment. Gender, education, race, and self-rated physical and mental health were included by default because of the associations with missing outcome data, as described above. We dichotomized responses to questions about self-rated physical and mental health (“poor/fair” versus “good/very good/excellent”) because small numbers of respondents viewed their health negatively. There was concern for potential multicollinearity between four structural indicators of respondents’ social connectedness - network size, network density, frequency of socializing, and community involvement - because these variables were moderately or strongly correlated (e.g., network size and density r=0.77, p<0.001; frequency of socializing and community involvement r=0.27, p<0.001). To address this issue, we used principal components analysis to create composite variables.²⁶ Two composites accounted for 75.1% of the total variance: (1) network structure (on which network size and density loadings were ≥ 0.60) and (2) social engagement (on which frequency of social activity and community involvement loadings were ≥ 0.60). Scores for each composite variable were calculated and used in the multivariable analysis. All analyses incorporated the survey design variables and analytic survey weights to account for differential nonresponse and the complex sampling design.

Results

Among NSHAP respondents, 227 of 2,767 reported that they did not have a desired surrogate. This represents an overall rate of 7.5% (95% CI, 6.4–8.7) after accounting for the survey design. Table 1 compares their demographic and medical characteristics to those of respondents with desired surrogates. Non-white respondents were more likely than white respondents to report that they did not have a desired surrogate, but there were no other significant differences between the two groups, including no difference in age or health status.

Table 1.

Demographic and health characteristics of NSHAP respondents

Characteristic	Without desired surrogate (N=227)	With desired surrogate (N=2540)	p-valuea
Age, % (95% CI)b			0.28
57–64	46.5 (38.1–54.9)	40.7 (38.2–43.2)
65–74	32.4 (25.2–39.6)	35.3 (33.2–37.5)
75–85	21.1 (15.3–26.9)	24.0 (22.0–26.0)
Male, % (95% CI)	55.6 (46.9–64.2)	47.1 (44.8–49.4)	0.05
Race, % (95% CI)			0.01
White	71.8 (62.8–80.8)	81.8 (78.5–85.1)
Black	14.1 (6.7–21.5)	9.2 (7.0–11.3)
Hispanic, non-Black	10.8 (4.7–16.9)	6.3 (3.4–9.1)
Other	2.2 (0.1–4.2)	2.5 (1.5–3.5)
Education, % (95% CI)			0.45
No degree	22.7 (16.8–28.7)	20.0 (16.9–23.0)
High school degree	33.1 (24.6–41.5)	38.3 (35.1–41.5)
College degree	44.2 (33.3–55.1)	41.7 (38.1–45.4)
Household income, % (95% CI)			0.49
< $25,000	33.6 (24.9–42.4)	25.0 (21.5–28.4)
$25–49,999	25.4 (16.5–34.3)	27.9 (25.2–30.6)
$50–$99,999	21.3 (9.6–32.9)	23.4 (21.1–25.7)
≥ $100,000	10.8 (1.6–19.9)	13.0 (10.4–15.5)
Missing	8.9 (4.7–13.1)	10.7 (8.7–12.7)
Self-rated physical health, % (95% CI)			0.34
Poor or Fair	28.5 (17.4–39.5)	23.4 (20.9–25.9)
Good/Very Good/ Excellent	71.5 (60.5–82.6)	76.3 (73.7–78.8)
Self-rated mental health, % (95% CI)			0.12
Poor or Fair	12.9 (7.2–18.6)	9.3 (7.7–10.9)
Good/Very Good/ Excellent	86.7 (80.9–92.5)	90.5 (88.9–92.1)
Modified Charlson Index, mean (95% CI)c	1.6 (1.3–1.9)	1.7 (1.6–1.8)	0.41

^aChi-square test for categorical variables and student’s t-test for continuous variables.

^bPercentages are weighted and were derived with the use of sampling weights from the NSHAP survey. Totals may not sum to 100% because of rounding.

^cModified Charlson Index scores range from 0 (no chronic conditions) to 16 (all measured conditions).

As shown in Figure 1, an overwhelming majority of respondents without a desired surrogate −202 of 227 (88.9%; 95% CI, 84.5–93.2) - had nuclear family. Just 6 of 227 (3.0%; 95% CI, 0.0–6.2) had neither nuclear family nor confidantes.

Figure 1.

NSHAP respondents without desired surrogates. Respondent numbers are unweighted. Because respondents could have multiple relationships, the numbers of respondents with each individual relationship do not sum to the number of respondents in each category.

Compared to respondents with desired surrogates, there were significant differences in the presence of nuclear family and in all other structural indicators of social connectedness (Table 2). Respondents without desired surrogates were less likely to have others living in their household (55.8% [95% CI, 46.1–65.6] versus 78.1% [95% CI, 75.9–80.2]; p<0.001), had smaller network sizes (3.7 members [95% CI, 3.3–4.0] versus 4.3 members [95% CI, 4.2–4.4]; p<0.001), and had lower levels of network density and community engagement. A higher percentage of respondents without desired surrogates (24.9% [95% CI, 14.3–35.5]) than those with desired surrogates (14.7% [95% CI, 13.1–16.3]) socialized with friends and family more than once a week. Approximately 70% of respondents in both groups did so at least once a month.

Table 2.

Social connectedness of NSHAP respondents with and without desired surrogates

Characteristic	Without desired surrogate (N=227)	With desired surrogate (N=2540)	p-valuea
Structural indicators
Had nuclear family, % (95% CI)b	88.9 (84.5–93.2)	98.4 (97.8–99.0)	<0.001
No family or confidantes, % (95% CI)	3.0 (0.0–6.2)	0.0 (0.0–0.1)	<0.001
Others living in household, % (95% CI)	55.8 (46.1–65.6)	78.1 (75.9–80.2)	<0.001
Network size, mean (95% CI)	3.7 (3.3–4.0)	4.3 (4.2–4.4)	<0.001
Network density, mean (95% CI)c	0.1 (0.1–0.2)	0.2 (0.2–0.2)	<0.001
Community involvement, mean (95% CI)d	7.3 (6.4–8.2)	8.2 (8.0v8.4)	0.04
Frequency of socializing, % (95% CI)			0.02
Never	2.2 (0.0–5.6)	1.3 (0.8v1.7)
Less than once a year	1.7 (0.0–3.6)	0.5 (0.2–0.8)
Once or twice a year	2.9 (0.7–5.2)	3.4 (2.6–4.2)
Several times a year	12.8 (9.2–16.5)	14.9 (13.1–16.7)
Once a month	20.0 (14.2–25.8)	20.1 (18.1–22.1)
Every week	25.4 (17.7–33.0)	34.3 (31.7–37.0)
Several times a week	24.9 (14.3–35.5)	14.7 (13.1–16.3)
Missing	10.2 (5.2–15.2)	10.7 (9.4–12.1)
Functional indicators
Social support, mean (95% CI)e	13.5 (13.1–13.9)	14.7 (14.6–14.8)	<0.001
UCLA Loneliness Scale, mean (95% CI)f	4.3 (4.0–4.6)	4.0 (3.9–4.0)	0.03
Could discuss health with ≥1 confidante, % (95% CI)	91.3 (85.6–97.0)	98.9 (98.4–99.4)	<0.001
Quality indicators
Social strain, mean (95% CI)g	8.1 (7.7–8.4)	7.8 (7.7–7.9)	0.13
History of elder abuse, % (95% CI)	20.7 (9.6–31.8)	13.4 (11.6–15.1)	0.15

^aChi-square test for categorical variables and student’s t-test for continuous variables.

^bPercentages are weighted and were derived with the use of sampling weights from the NSHAP survey. Totals may not sum to 100% because of rounding.

^cNetwork density values are continuous between 0 and 1, where 0 indicates that no possible pairs of network members have any contact and 1 indicates that all possible pairs have maximum contact.

^dCommunity involvement scores range from 0 (no participation in religious services, community meetings, or volunteer work in the past 12 months) to 18 (frequent participation in all of these).

^eSocial support was measured on a composite scale from 0 (no support from spouse, family, or friends) to 18 (maximum support for all three sources). Because unmarried adults were asked about family and friends only, their scores were averaged and rescaled.

^fLoneliness scores range from 3 (never or hardly ever felt a lack of companionship, left out, or isolated from others) to 9 (very often felt a lack of companionship, left out, and isolated from others).

^gSocial strain scores range from 0 (spouse, family and friends never made too many demands, were critical, or got on nerves) to 27 (spouse, family, and friends all frequently made too many demands, were critical, and got on nerves). Because unmarried adults were asked about family and friends only, their scores were averaged and rescaled.

There were significant differences in all functional indicators of social connectedness. Respondents who did not have a desired surrogate reported lower levels of social support and higher levels of loneliness. While they were also less likely to have a confidante with whom they could discuss a health issue, more than 90% of respondents in both groups were able to name such a confidante (91.3% [95% CI, 85.6–97.0] of respondents without desired surrogates and 98.9% [95% CI, 98.3–99.4] of respondents with desired surrogates). Elder abuse was more common among persons without desired surrogates, but the difference was not statistically significant.

Figure 2 shows the results of a logistic regression analysis limited to the respondents in both groups who reported that they had nuclear family. Four factors were associated with respondents who had nuclear family but did not have a desired surrogate. Such respondents had lower odds of social support (aOR 0.92; 95% CI, 0.86–0.99). They had higher odds of living alone (aOR, 2.39; 95% CI, 1.44–4.00) and of being male (aOR, 1.53; 95% CI, 1.10–2.12) and Hispanic (aOR 2.12; 95% CI, 1.30–3.47). In multivariable analysis, there was no association with other factors, including having a confidante to discuss health issues, network structure, or social engagement. Full results appear in Supplementary Table S1.

Figure 2.

Associations between selected factors and the absence of a desired surrogate among respondents with nuclear family. Statistically significant associations are marked with **. Because several measures of social connectedness were moderately or strongly correlated, principal components analysis was used to create two composite variables. Network structure was created from network size and density measures, and social engagement was created from community involvement and socializing measures. Values range from 0 to 1, with greater values indicating stronger network structure and greater social engagement.

Discussion

In this study of a nationally representative sample of older adults without known cognitive impairment, we found that nearly 8% did not have a desired surrogate to make medical decisions for them if they lost capacity. Almost all had at least one nuclear family member who would be legally recognized to make decisions in most states. Compared to older adults with desired surrogates, those without desired surrogates were less socially connected across multiple domains. Most, however, were not profoundly isolated. On average, such persons had four confidantes in their social networks, approximately 70% reported socializing with family and friends at least once a month, and more than 90% had at least one confidante with whom they were able to discuss their health issues.

Health care providers traditionally have expected family members to act as surrogate decision-makers for patients who develop impaired capacity, and default surrogate consent laws explicitly empower family members to take on this role. In considering the limitations of this paradigm, previous work has examined the negative relationship between social isolation and participation in advance care planning,²⁷ and there is a growing body of literature dedicated to improving decision-making for individuals who have no family or friends.^{3, 28} Our findings complicate this view of surrogate decision-making by showing that patients who are truly at risk of becoming “unbefriended,” in the literal sense that they have no intimate relationships, represent a small fraction of a far larger group of older persons who have no one whom they would choose as a surrogate. Only 3% of NSHAP respondents without a desired surrogate reported that they had no family and no confidantes. Furthermore, our results suggest that data obtained by health care providers in the social history, such as information about family or household structure, do not reliably identify patients who do not have a desired surrogate to represent them. Among respondents with nuclear family, living alone was the strongest predictor of not having a desired surrogate, but more than half of respondents without desired surrogates had others in their household. These findings suggest that health care providers should not make assumptions about the presence of a desired surrogate based on what they know about their patients’ families and social environment. Particular attention to this issue may be important among patients from racial and ethnic minority backgrounds, since, in our study, non-white respondents were more likely not to have a desired surrogate, and Hispanic ethnicity was a predictor of not having a desired surrogate among respondents with nuclear family. These findings may help to explain why non-white patients are more likely to be represented by professional guardians.⁹ Overall, however, the best way for health care providers to identify patients who do not have a desired surrogate is to ask all older persons about this issue directly.

No clear guidance exists to help providers know what to do, and what sort of advance care planning to recommend, when patients report that they have no one whom they would want to make decisions on their behalf. Further work is needed to develop an effective strategy to address this problem. Such a strategy is likely to have at least three components. First, providers will need to encourage such individuals to think broadly and creatively about non-traditional surrogates - persons who may not be an intimate part of their social context, but who would nevertheless be well positioned to make decisions for them. Patients with deeply held religious beliefs, for example, may have clergy who are qualified to take on this role. Second, it will be particularly important to ensure that patients without desired surrogates participate in robust efforts to document their values, preferences, and wishes. Work is ongoing to determine the best form of such documentation, with efforts focusing on values assessment tools,²⁹ video diaries,³⁰ and improved advance directives.^{31, 32} Finally, sound options need to be developed to identify surrogates for people who do not have family or do not want their family to make decisions for them. One proposal has been to encourage the licensing of health care fiduciaries — professionals with training in surrogate decision-making whom an older adult could hire to participate in discussions about his or her treatment preferences and then, if necessary, to serve as a surrogate decision-maker and assist the medical team in making the best possible decisions.³³ This could be an extension of the role already served by Board-Certified Patient Advocates.³⁴ A second option would be to expand the voluntary guardianship programs that already exist in several parts of the country, in which community members make decisions for incapacitated persons with oversight from the court.³⁵ Such programs could invite patients with capacity but without a desired surrogate to develop a relationship with a volunteer guardian, so that, if necessary, the guardian would be well equipped to speak on their behalf.

Our study has several limitations. First, although we were able to describe respondents without desired surrogates, information was not collected in the NSHAP survey about why respondents did not have desired surrogates or about whether they had advance directives or had participated in other forms of advance care planning. In addition, individuals with known cognitive impairment were excluded, even though many such persons still possess the capacity to identify a desired surrogate. Future work should address these knowledge gaps. Finally, the NSHAP Wave 1 data were collected more than a decade ago. It is possible that the prevalence and characteristics of older adults without desired surrogates have changed as the baby boom generation has entered older age. To our knowledge, however, more recent data are not available to study this important issue.

In summary, in a large, nationally representative sample, nearly 8% of older adults did not have a desired surrogate. Most had nuclear family and were not socially disconnected. These findings suggest that health care providers should ask all older adults whether they have a desired surrogate, and that work is needed to develop strategies to identify surrogates for persons who do not have family or would not choose family to make decisions for them.