This cohort study assesses whether identifiable support buffers the vulnerability of a health shock in older adults who live alone.
Key Points
Question
What is the association between social support and functional outcomes (ie, activities of daily living dependence, prolonged nursing home stay, death) in older adults who live alone?
Findings
In this cohort study of 4772 older adults who live alone, identifiable social support was associated with a lower likelihood of prolonged nursing home stays but only in the setting of a sudden change in health (eg, hospitalization, new cancer diagnosis). Support was not associated with new activities of daily living dependence or death.
Meaning
Identifiable support may buffer the risk of prolonged nursing home stays when facing a new health stressor while living alone.
Abstract
Importance
Older adults who live alone are at risk for poor health outcomes. Whether social support mitigates the risk of living alone, particularly when facing a sudden change in health, has not been adequately reported.
Objective
To assess if identifiable support buffers the vulnerability of a health shock while living alone.
Design, Setting, and Participants
In this longitudinal, prospective, nationally representative cohort study from the Health and Retirement Study (enrollment March 2006 to April 2015), 4772 community-dwelling older adults 65 years or older who lived alone in the community and could complete activities of daily living (ADLs) and instrumental ADLs independently were followed up biennially through April 2018. Statistical analysis was completed from May 2020 to March 2021.
Exposures
Identifiable support (ie, can the participant identify a relative/friend who could help with personal care if needed), health shock (ie, hospitalization, new diagnosis of cancer, stroke, heart attack), and interaction (multiplicative and additive) between the 2 exposures.
Main Outcomes and Measures
The primary outcomes were incident ADL dependency, prolonged nursing home stay (≥30 days), and death.
Results
Of 4772 older adults (median [IQR] age, 73 [68-81] years; 3398 [71%] women) who lived alone, at baseline, 1813 (38%) could not identify support, and 3013 (63%) experienced a health shock during the study. Support was associated with a lower risk of a prolonged nursing home stay at 2 years (predicted probability, 6.7% vs 5.2%; P = .002). Absent a health shock, support was not associated with a prolonged nursing home stay (predicted probability over 2 years, 1.9% vs 1.4%; P = .21). However, in the presence of a health shock, support was associated with a lower risk of a prolonged nursing home stay (predicted probability over 2 years, 14.2% vs 10.9%; P = .002). Support was not associated with incident ADL dependence or death.
Conclusions and Relevance
In this longitudinal cohort study among older adults who live alone, identifiable support was associated with a lower risk of a prolonged nursing home stay in the setting of a health shock.
Introduction
Older adults who live alone are an at-risk population. Studies have found that living alone is predictive of a wide range of poor health outcomes.1,2,3,4,5,6,7,8,9 From poor outcomes following cardiovascular events3,4,5 to increased depressive symptoms6,9 to higher rates of all-cause mortality,7,8 living alone is an indicator of risk.
However, older adults who live alone are not homogeneous. Some older adults who live alone are socially well supported and may not be as vulnerable as their living status suggests.9,10,11,12 Irrespective of whether they live alone, among all older adults, higher levels of quantifiable or perceived social support are associated with improved quality of life, better physical and mental health, and lower rates of frailty and mortality.13,14,15 Also, social support is beneficial when recovering from an acute illness (eg, acute myocardial infarction16), in preventing exacerbation of chronic illnesses (eg, heart failure exacerbation17), and in averting institutionalization in older adults.18
While living alone is a documented risk factor and social support is broadly beneficial to older adults, little is known about the circumstances in which social support may protect older adults who live alone.19 Social support may influence health in 2 circumstances: social support may be beneficial at all times, or the benefit may be principally realized when facing a stressor (eg, a hospitalization, a new cancer diagnosis).20 Characterizing the settings where social support is beneficial may advance our understanding of social support and identify opportunities to develop and test interventions to augment support.
To assess whether identifiable support is associated with mitigation of the vulnerability of living alone, we examined a cohort of nationally representative older adults who live alone. First, we determined whether identifiable support was associated with 3 health outcomes—incident activities of daily living (ADL) impairment, prolonged nursing home stay, or death. Then we determined whether identifiable support buffers the vulnerability of facing a new health stressor while living alone.
Methods
Design, Setting, and Participants
We performed a longitudinal analysis of older adults who live alone to (1) evaluate the association between identifiable support and functional outcomes (ie, ADL dependence, prolonged nursing home stay, death) and (2) assess if identifiable support acts as a buffer for functional outcomes during sudden changes in health status. Ethical approval for data analysis was obtained from the University of California, San Francisco, Committee on Human Research, which waived the patient consent requirement because the data are deidentified.
We used data from the Health and Retirement Study (HRS), a nationally representative, longitudinal study of older adults.21 Participants aged 50 years and older in the HRS are interviewed every 2 years to measure changes in disability, health, and wealth as they transition from work to retirement. The study interviews participants on 4 broad areas of aging: income and wealth; health, function, and cognition; work and retirement; and family connections. The HRS sampling methods were designed to create a cohort representative of community-dwelling older adults in the US.22 The HRS-based population estimates of mortality, education, and nursing home residence match estimates from the US Census.23
To create the study cohort, we included HRS participants 65 years or older who lived alone in the community and were functionally independent. We defined living alone irrespective of marital status as a household size of 1. We defined functional independence as not receiving help with any ADL or instrumental ADL; we did not exclude those who could complete activities with difficulty. We included participants who met inclusion criteria in any interview wave from March 2006 through April 2015, and we ascertained outcomes at the follow-up interviews, approximately every 2 years; follow-up interviews took place from February 2008 through April 2018. We excluded participants who were missing data on their living arrangements, outcomes, or comorbidities. We also excluded participants whose baseline interview was completed by proxy because depression, an important confounder, is not reliably measured by proxy interview. In each wave, less than 5% of participants were excluded for any reason (eTable 1 in the Supplement).
Measuring Identifiable Support
At the index interview, we determined if participants could identify support. Identifiable support was measured using the question, “Suppose in the future, you needed help with basic personal care activities like eating or dressing. Do you have relatives or friends besides your spouse/partner who would be willing and able to help you over a long period of time?” We categorized participants’ responses as the following: we took “yes” to mean “can identify support” and “no,” “refused,” or “do not know” to mean “cannot identify support.”
Health Shock
We identified participants who experienced a health shock, a sudden change in health status, during biennial follow-up interviews. Like prior investigators, we defined a health shock as a new-onset major illness or hospitalization.24 We defined major illness as a self-reported diagnosis of cancer (excluding skin cancer), heart disease, or stroke that was absent at an index interview and present on the following interview. Hospitalization was assessed through self-report; we included hospitalizations that lasted 2 or more days.
Outcomes
We examined 3 primary outcomes during biennial follow-up interviews: incident ADL dependency, prolonged nursing home stay, and death. The HRS obtains information from respondents about nursing home stays and moves to long-term care. To distinguish from short-term skilled nursing rehabilitation stays, we defined prolonged nursing home stays 30 days or more over 2 years consistent with definitions used by prior investigators.25 For participants who died, the HRS uses information from next-of-kin interviews and National Death Index data to determine the date of death. When participants died, we used data from a next-of-kin interview to determine if they also experienced a new disability or nursing home stay between the index wave and death. Participants could experience more than 1 outcome.
Confounder Measurement
We accounted for 3 domains for measurable confounders: sociodemographic, mental and physical health, and functional impairment. We used HRS interview data to characterize participants’ age, sex, self-reported race and ethnicity, education, and financial net worth. We characterized health status using the following measures: self-reported fair or poor health, significant pain (self-report that they often have trouble with pain and that the pain is moderate or severe),26 common medical comorbidities (self-report of physician’s diagnosis of hypertension, diabetes, cancer, lung disease, heart disease, stroke, or arthritis), depression (score ≥4 on the Center for Epidemiologic Studies Depression scale),27,28 and tobacco use. We characterized function using the following measures: difficulty with any of 6 ADLs (getting out of bed, bathing, dressing, eating, toileting, walking), cognitive status (using Langa-Weir score),29 visual impairment, and hearing impairment.
Statistical Analysis
To compare categorical baseline variables, we used the χ2 test. To compare continuous baseline variables, we used the Wilcoxon rank sum test for variables not normally distributed. We used generalized estimating equations with a logit link and accounted for time-varying confounders and repeated observations with an exchangeable correlation structure to measure the association between identifiable support and functional outcomes. All outcomes were assessed at the follow-up interview, on average 2 years after the index interview. We excluded participant observation with missing outcome data at the follow-up interview, on average 5% (eTable 1 in the Supplement). Outcomes other than death were treated as repeated outcomes (ie, participants could experience the same nondeath outcome more than once). We report the predicted population rates and the average marginal effect (AME) of having identifiable support from these results.30 The AME is the change in the predicted outcome (eg, death) associated with a change in exposure (eg, going from no support to having support) calculated for each observation and averaged over the study population.31 We identified confounders a priori; we did not use significance testing to determine which confounders to include in the regression models, consistent with epidemiologic best practices.32
To test if identifiable support buffers against poor outcomes (eFigure in the Supplement), we added an interaction term between health shock and identifiable support in the regression models described above. We tested for a multiplicative interaction (relative scale) and an additive interaction (absolute scale). We tested for a multiplicative interaction using the ratio of odds ratios (ORs), OR11 / (OR10 × OR01). We tested for an additive interaction by calculating the relative excess risk owing to interaction (RERI), where RERIOR = OR11 − OR10 − OR01 + 1.33,34,35 A RERIOR greater than 0 indicates an additive interaction; that is, the probability of the outcome with both exposures (ie, no support and health shock) is greater than would be expected if each exposure had an independent effect on the outcome. We report all results with 95% CIs. All P values are 2-sided, and the a priori significance threshold was P < .05. We performed analyses using SAS, version 9.4 (SAS Institute) and R, version 3.4.4 (R Foundation for Statistical Computing). We calculated predicted probabilities and average marginal effects using the SAS-supported Margins macro (version 1.8).36
Results
Patient Characteristics and Support
This study examined 4772 older adults (median [IQR] age, 73 [68-81] years; 3398 [71%] women) who lived alone and were functionally independent (Table 1). Participants were followed up for a mean of 4.9 years (23 187 person-years total). At baseline, 1813 (38%) could not identify support. Those who lived alone and could not identify support were older (median [IQR] age, 74 [68-82] years vs 73 [67-80] years), more likely to identify as White (1553 of 1813 [86%] vs 2350 of 2959 [79%]), and more likely to be high school graduates (1452 of 1813 [80%] vs 2227 of 2959 [75%]). In both groups, most were women, and a minority of patients were married. Participants who could not identify support had worse health and function across nearly all measures; they more often reported their health as fair or poor (609 of 1813 [34%] vs 726 of 2959 [25%]), were more often depressed (409 of 1813 [23%] vs 471 of 2959 [16%]), and had higher sensory impairment rates (visual impairment: 455 of 1813 [25%] vs 566 of 2959 [19%]; hearing impairment: 421 of 1813 [23%] vs 606 of 2959 [20%]). Cognitive impairment did not differ by support.
Table 1. Characteristics of Older Adults Who Live Alone in the Study Cohort.
| Characteristic | No. (%) | P value | |
|---|---|---|---|
| Cannot identify support (n = 1813) | Can identify support (n = 2959) | ||
| Sociodemographics | |||
| Age, median (IQR), y | 74 (68-82) | 73 (67-80) | <.001 |
| Sex | |||
| Female | 1257 (69) | 2141 (72) | .03 |
| Male | 556 (31) | 818 (28) | |
| Race | |||
| Black | 199 (11) | 528 (18) | <.001 |
| White | 1553 (86) | 2350 (79) | |
| Othera | 61 (3) | 81 (3) | |
| Hispanic ethnicity | 136 (8) | 191 (6) | .18 |
| Married | 84 (5) | 126 (4) | .59 |
| High school graduate | 1452 (80) | 2227 (75) | <.001 |
| Net worth less than cohort median | 911 (50) | 1383 (47) | .02 |
| Health status | |||
| Fair or poor self-reported health | 609 (34) | 726 (25) | <.001 |
| Significant painb | 138 (8) | 132 (4) | <.001 |
| Hypertension | 1110 (61) | 1873 (63) | .16 |
| Diabetes | 398 (22) | 604 (20) | .22 |
| Cancer (excluding minor skin cancer) | 361 (20) | 507 (17) | .02 |
| Lung disease | 191 (11) | 297 (10) | .62 |
| Heart disease | 530 (29) | 765 (26) | .01 |
| Stroke | 162 (9) | 251 (8) | .63 |
| Arthritis | 1190 (66) | 1952 (66) | .84 |
| Depressionc | 409 (23) | 471 (16) | <.001 |
| Ever tobacco use | 1051 (58) | 1653 (56) | .16 |
| Function | |||
| Difficulty getting out of bed | 78 (4) | 92 (3) | .04 |
| Difficulty bathing | 93 (5) | 62 (2) | <.001 |
| Difficulty dressing | 169 (9) | 159 (5) | <.001 |
| Difficulty eating | 31 (2) | 22 (1) | .003 |
| Difficulty toileting | 130 (7) | 110 (4) | <.001 |
| Difficulty walking | 98 (5) | 114 (4) | .01 |
| Cognitiond | |||
| Intact | 1350 (74) | 2213 (75) | .32 |
| Impairment, not dementia | 384 (21) | 593 (20) | |
| Dementia | 79 (4) | 153 (5) | |
| Visual impairmente | 455 (25) | 566 (19) | <.001 |
| Hearing impairmentf | 421 (23) | 606 (20) | .03 |
Other includes self-reported race as American Indian, Alaska Native, Asian, Native Hawaiian, Pacific Islander, or something else.
Participants were considered to have significant pain if they said they were troubled by pain and that most of the time, the pain is severe.
Depression = Center for Epidemiologic Studies Depression Scale score of 4 or higher.
Participants’ cognitive status was determined using the Langa-Weir score.
Visual impairment was defined as answering “fair” or “poor” to the question, “Is your eyesight excellent, very good, good, fair, or poor using glasses or corrective lenses as usual?”
Hearing impairment was defined as answering “fair” or “poor” to the question, “Is your hearing excellent, very good, good, fair, or poor (using a hearing aid as usual)?”
Functional Outcomes
In unadjusted analyses, after 2 years of follow-up, participants who lived alone and could not identify support were more likely to develop a new ADL dependency (predicted probability, 11.1% vs 9.4%; AME, 1.7%; 95% CI, 0.5%-2.9%), more likely to have a prolonged nursing home stay (predicted probability, 7.3% vs 4.9%; AME, 2.5%; 95% CI, 1.5%-3.4%), and more likely to die (predicted probability, 8.0% vs 6.5%; AME, 1.5%; 95% CI, 0.5%-2.5%) (Table 2). After accounting for confounders, those who lived alone and could not identify support were more likely to have a prolonged nursing home stay (predicted probability, 6.7% vs 5.2%; AME, 1.4%; 95% CI, 0.5%-2.4%) in the following 2 years but were no more likely to develop a new ADL dependency or to die (Table 3).
Table 2. Association of Identifiable Support and Functional Outcomes in Older Adults Who Live Alone, Unadjusteda.
| Outcome | % (95% CI) | ||
|---|---|---|---|
| Predicted probability | Average marginal effect | ||
| No support | Support | ||
| ADL dependency | 11.1 (10.1-12.0) | 9.4 (8.7-10.1) | 1.7 (0.5-2.9) |
| Nursing home stay | 7.3 (6.5-8.1) | 4.9 (4.3-5.4) | 2.5 (1.5-3.4) |
| Died | 8.0 (7.2-8.8) | 8.0 (7.2-8.8) | 1.5 (0.5-2.5) |
Abbreviation: ADL, activities of daily living.
Results are unadjusted and account for repeated observations. Support is defined as answering “yes” to the question, “Suppose in the future, you needed help with basic personal care activities like eating or dressing. Do you have relatives or friends besides your spouse/partner who would be willing and able to help you over a long period of time?”
Table 3. Association of Identifiable Support and Functional Outcomes in Older Adults Who Live Alone, Adjusteda.
| Outcome | % (95% CI) | ||
|---|---|---|---|
| Predicted probability | Average marginal effect | ||
| No support | Support | ||
| ADL dependency | 9.8 (9.0 to 10.6) | 10.3 (9.6 to 11.0) | −0.5 (−1.6 to 0.6) |
| Nursing home stay | 6.7 (5.9 to 7.4) | 5.2 (4.7 to 5.8) | 1.4 (0.5 to 2.4) |
| Died | 7.3 (6.6 to 8.0) | 7.1 (6.5 to 7.7) | 0.2 (−0.7 to 1.2) |
Abbreviation: ADL, activities of daily living.
Results adjusted for age, gender, race, education, income, pain, visual impairment, hearing impairment, hypertension, cancer diagnosis, lung disease, heart disease, stroke, arthritis, cognitive impairment, ADL difficulty, and account for repeated observations. Association with nursing home stay is statistically and clinically significant. Differences in death and functional impairment are neither statistically nor clinically significant. Support is defined as answering “yes” to the question, “Suppose in the future, you needed help with basic personal care activities like eating or dressing. Do you have relatives or friends besides your spouse/partner who would be willing and able to help you over a long period of time?” Full regression results can be found in eTables 3-5 of the Supplement.
Functional Outcomes and Health Shock
During the study, 3013 of 4772 (63%) older adults who lived alone experienced a health shock. Adjusting for confounders, identifiable support was not associated with a health shock in the following 2 years (predicted probability, 36.6% vs 36.4%; AME, −0.2%; 95% CI, −2.0% to 1.6%).
The interactive effects of support and health shocks on functional outcomes are shown in the Figure. There was no evidence of interaction on the additive or multiplicative scale when examining the interaction of support and health shock on incident ADL dependency or death.
Figure. Association of Identifiable Support and Functional Outcomes in Older Adults Who Live Alone Stratified by Health Shock, Adjusted.
When examining death and disability, there was no significant interaction between support and health shock. When examining nursing home stay, there was an additive interaction between support and health shock. Those without identifiable support had an excess risk of nursing home stay. Support was defined as answering “yes” to the question, “Suppose in the future, you needed help with basic personal care activities like eating or dressing. Do you have relatives or friends besides your spouse/partner who would be willing and able to help you over a long period of time?” Health shock was defined as hospital stay of 2 or more days, new cancer diagnosis, new stroke, or new heart attack. The predicted likelihood is a population estimate adjusted for age, sex, race and ethnicity, education, income, pain, visual impairment, hearing impairment, hypertension, cancer diagnosis, lung disease, heart disease, stroke, arthritis, cognitive impairment, and activities of daily living (ADL) difficulty. Full regression results can be found in eTables 6-8 of the Supplement.
There was an additive interaction (absolute scale) between support and health shock on the likelihood of a prolonged nursing home stay. The predicted probability of a prolonged nursing home stay was higher in the group that could not identify support and experienced a health shock than would be expected if each exposure were considered independently (RERIOR, 3.3; 95% CI, 0.8-5.8). Specifically, in the absence of a health shock, support was not associated with the risk of prolonged nursing home stay in the following 2 years (1.9% vs 1.4%; AME, 0.5%; 95% CI, −0.1% to 1.1%). However, in the presence of a health shock, lacking support was associated higher risk of a prolonged nursing home stay over 2 years (14.2% vs 10.9%; AME, 3.3%; 95% CI, 1.2%-5.4%). There was no evidence of a multiplicative interaction (relative scale). In a sensitivity analysis, we examined 3 added nursing home stay thresholds (1 day, 60 days, 100 days). Consistent with the primary results, we found an additive interaction between support and health shock on nursing home stays using a 1-day and 60-day threshold; we did not find an interaction using a 100-day threshold (eTable 2 in the Supplement).
Discussion
In a nationally representative cohort of older adults, we found that 2 of every 5 older adults who live alone could not identify support should they need help with personal care activities in the future. After accounting for observable differences, we found that identifiable support was not associated with incident ADL dependency or death. We found that identifiable support was associated with reduced need for prolonged nursing home stays, and the magnitude of the association was substantial in the setting of a health shock. When older adults experienced a sudden change in health, the risk of a prolonged nursing home stay was reduced from 14% in those without identifiable support to 11% in those with identifiable support. Further, sensitivity analyses indicated that the association may be more substantive for short- and medium-length nursing home stays than long-term nursing home stays where needs may be more profound. These findings suggest meaningful clinical value in learning if a patient who lives alone can identify a source of support.
These findings build on ongoing work to understand the role of social support in older adults who live alone to remain in the community. Studies have demonstrated that living alone, social support, and acute medical illnesses are important predictors of nursing home use and long-term institutionalization.18,37,38,39,40,41 We build on this literature in 3 ways. First, prior studies measured support with measures such as the presence of adult children, social network size, and participation in social activities to infer if an older adult may have the support needed to remain in the community. We used a more direct measure—a question asking those who live alone if they can identify someone to help with personal care needs if needed, a deficit that often results in nursing home use. Further, this measure can be more easily obtained in a clinical interaction. Second, while prior analyses7,42 identified living alone as a risk factor for nursing home care, in the present study, we found that among those who live alone, social support was associated with a meaningful decrease in the likelihood of prolonged nursing home stays. Third, we found that the benefit of social support in older adults who live alone was principally realized following a significant health event (eg, hospitalization, new cancer diagnosis).
While institutional settings may be appropriate for some older adults, many older adults prefer to age in place and recover from acute illness in their homes.43,44,45 Beyond the social and psychic costs, even short- and medium-term institutional care poses a tremendous financial cost to older adults and insurers. When older adults use their Medicare skilled nursing facility benefit, their average out-of-pocket expense exceeds $2000, and the typical cost to Medicare is greater than $10 000 per admission.46,47 Programs that support informal caregivers may benefit older adults and insurers and acknowledge the complex, unpaid work of families and friends who support older adults to remain in the community. For example, an analysis of California’s Paid Family Leave law showed that even a mere 6 weeks of mandated paid leave reduced nursing home stays in older adults by 11%.48 Caregivers also report that nonfinancial programs, like respite care supported by the National Family Caregiver Support Program, allow older adults to delay or avoid institutionalization.49 These results should also prompt clinicians, researchers, patient advocates, and policy makers to identify and test what type of support may be beneficial to the 40% of older adults who live alone and cannot identify support, particularly following a health shock.
Limitations
The study design and data have limitations that are important to consider when interpreting the results. First, while we infer that those who can identify support are more likely to obtain support when needed, we cannot confirm this. Second, this analysis did not account for the chronicity of living alone. Third, we could not account for changes to living alone or social support between the baseline interview and outcome. This exposure misclassification could bias estimates of support toward the null. Fourth, while we believe that we identified a comprehensive list of confounders of nursing home need, we cannot empirically verify that we have fully accounted for confounding. Finally, although the HRS is nationally representative, findings from this study may not generalize to populations typically underrepresented in survey studies, such as patients receiving hospice care or racial and ethnic minority groups, such as Asian American individuals.
Conclusions
Prior work demonstrates that older adults who live alone are vulnerable. In this cohort study, we found that identifiable support in older adults was associated with a lower risk of a prolonged nursing home stay, primarily in the setting of a health shock. Supporting informal caregivers may meet the desires of many older adults to remain in the community and reduce nursing home costs borne by patients and insurers. For the nearly 40% of older adults who live alone without identifiable support, we need to identify what type of support may be beneficial, especially following a health shock.
eTable 1. Cohort inclusion table by baseline interview wave
eTable 2. Nursing home cutoff sensitivity analysis
eTable 3. ADL dependence outcome corresponding to Table 3 in main manuscript
eTable 4. Nursing home outcome corresponding to Table 3 in main manuscript
eTable 5. Mortality outcome corresponding to Table 3 in main manuscript
eTable 6. ADL dependence outcome health shock interaction analysis corresponding to Figure 1 in main manuscript
eTable 7. Nursing home outcome health shock interaction analysis corresponding to Figure 1 in main manuscript
eTable 8. Mortality outcome health shock interaction analysis corresponding to Figure 1 in main manuscript
eFigure. Conceptual diagram demonstrating the interaction between social support and outcome with and without a stressor
References
- 1.Mahoney JE, Eisner J, Havighurst T, Gray S, Palta M. Problems of older adults living alone after hospitalization. J Gen Intern Med. 2000;15(9):611-619. doi: 10.1046/j.1525-1497.2000.06139.x [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Kharicha K, Iliffe S, Harari D, Swift C, Gillmann G, Stuck AE. Health risk appraisal in older people 1: are older people living alone an “at-risk” group? Br J Gen Pract. 2007;57(537):271-276. [PMC free article] [PubMed] [Google Scholar]
- 3.Case RB, Moss AJ, Case N, McDermott M, Eberly S. Living alone after myocardial infarction. impact on prognosis. JAMA. 1992;267(4):515-519. doi: 10.1001/jama.1992.03480040063031 [DOI] [PubMed] [Google Scholar]
- 4.Bucholz EM, Rathore SS, Gosch K, et al. Effect of living alone on patient outcomes after hospitalization for acute myocardial infarction. Am J Cardiol. 2011;108(7):943-948. doi: 10.1016/j.amjcard.2011.05.023 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Reeves MJ, Prager M, Fang J, Stamplecoski M, Kapral MK. Impact of living alone on the care and outcomes of patients with acute stroke. Stroke. 2014;45(10):3083-3085. doi: 10.1161/STROKEAHA.114.006520 [DOI] [PubMed] [Google Scholar]
- 6.Dean A, Kolody B, Wood P, Matt GE. The influence of living alone on depression in elderly persons. J Aging Health. 1992;4(1):3-18. doi: 10.1177/089826439200400101 [DOI] [Google Scholar]
- 7.Pimouguet C, Rizzuto D, Schön P, et al. Impact of living alone on institutionalization and mortality: a population-based longitudinal study. Eur J Public Health. 2016;26(1):182-187. doi: 10.1093/eurpub/ckv052 [DOI] [PubMed] [Google Scholar]
- 8.Gopinath B, Rochtchina E, Anstey KJ, Mitchell P. Living alone and risk of mortality in older, community-dwelling adults. JAMA Intern Med. 2013;173(4):320-321. doi: 10.1001/jamainternmed.2013.1597 [DOI] [PubMed] [Google Scholar]
- 9.Russell D, Taylor J. Living alone and depressive symptoms: the influence of gender, physical disability, and social support among Hispanic and non-Hispanic older adults. J Gerontol B Psychol Sci Soc Sci. 2009;64(1):95-104. doi: 10.1093/geronb/gbn002 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Perissinotto CM, Covinsky KE. Living alone, socially isolated or lonely—what are we measuring? J Gen Intern Med. 2014;29(11):1429-1431. doi: 10.1007/s11606-014-2977-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Thompson EE, Krause N. Living alone and neighborhood characteristics as predictors of social support in late life. J Gerontol B Psychol Sci Soc Sci. 1998;53(6):S354-S364. doi: 10.1093/geronb/53B.6.S354 [DOI] [PubMed] [Google Scholar]
- 12.Renwick KA, Sanmartin C, Dasgupta K, Berrang-Ford L, Ross N. The influence of low social support and living alone on premature mortality among aging Canadians. Can J Public Health. 2020;111(4):594-605. doi: 10.17269/s41997-020-00310-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Blazer DG. Social support and mortality in an elderly community population. Am J Epidemiol. 1982;115(5):684-694. doi: 10.1093/oxfordjournals.aje.a113351 [DOI] [PubMed] [Google Scholar]
- 14.Gobbens RJJ, van Assen MA. The prediction of quality of life by physical, psychological and social components of frailty in community-dwelling older people. Qual Life Res. 2014;23(8):2289-2300. doi: 10.1007/s11136-014-0672-1 [DOI] [PubMed] [Google Scholar]
- 15.Woo J, Goggins W, Sham A, Ho SC. Social determinants of frailty. Gerontology. 2005;51(6):402-408. doi: 10.1159/000088705 [DOI] [PubMed] [Google Scholar]
- 16.Reifman A. Social relationships, recovery from illness, and survival: a literature review. Ann Behav Med. 1995;17(2):124-131. doi: 10.1007/BF02895061 [DOI] [PubMed] [Google Scholar]
- 17.Krumholz HM, Butler J, Miller J, et al. Prognostic importance of emotional support for elderly patients hospitalized with heart failure. Circulation. 1998;97(10):958-964. doi: 10.1161/01.CIR.97.10.958 [DOI] [PubMed] [Google Scholar]
- 18.Oh A, Patel K, Boscardin WJ, et al. Social support and patterns of institutionalization among older adults: a longitudinal study. J Am Geriatr Soc. 2019;67(12):2622-2627. doi: 10.1111/jgs.16184 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Bucholz EM, Krumholz HM. Loneliness and living alone: what are we really measuring? Arch Intern Med. 2012;172(14):1084-1085. doi: 10.1001/archinternmed.2012.2649 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Cohen S, Wills TA. Stress, social support, and the buffering hypothesis. Psychol Bull. 1985;98(2):310-357. doi: 10.1037/0033-2909.98.2.310 [DOI] [PubMed] [Google Scholar]
- 21.Juster FT, Suzman R. An overview of the Health and Retirement Study. J Hum Resour. 1995;30:S7-S56. doi: 10.2307/146277 [DOI] [Google Scholar]
- 22.Sonnega A, Faul JD, Ofstedal MB, Langa KM, Phillips JW, Weir DR. Cohort profile: the Health and Retirement Study (HRS). Int J Epidemiol. 2014;43(2):576-585. doi: 10.1093/ije/dyu067 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Sonnega A. The Health and Retirement Study: an introduction, the view from 38000 feet. Presented at: The Health and Retirement Study Workshop; June 2, 2017; University of Michigan Institute for Social Research. [Google Scholar]
- 24.Prinz D, Chernew M, Cutler D, Frakt A. Health and Economic Activity Over the Lifecycle: Literature Review. National Bureau of Economic Research; 2018. doi: 10.3386/w24865 [DOI] [Google Scholar]
- 25.Kim DH, Glynn RJ, Avorn J, et al. Validation of a claims-based frailty index against physical performance and adverse health outcomes in the Health and Retirement Study. J Gerontol A Biol Sci Med Sci. 2019;74(8):1271-1276. doi: 10.1093/gerona/gly197 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Smith AK, Cenzer IS, Knight SJ, et al. The epidemiology of pain during the last 2 years of life. Ann Intern Med. 2010;153(9):563-569. doi: 10.7326/0003-4819-153-9-201011020-00005 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Steffick D. Documentation of Affective Functioning Measures in the Health and Retirement Study. Institute for Social Research, University of Michigan; 2000. Accessed August 13, 2019. https://hrs.isr.umich.edu/publications/biblio/5411
- 28.Zivin K, Llewellyn DJ, Lang IA, et al. Depression among older adults in the United States and England. Am J Geriatr Psychiatry. 2010;18(11):1036-1044. doi: 10.1097/JGP.0b013e3181dba6d2 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Crimmins EM, Kim JK, Langa KM, Weir DR. Assessment of cognition using surveys and neuropsychological assessment: the Health and Retirement Study and the Aging, Demographics, and Memory Study. J Gerontol B Psychol Sci Soc Sci. 2011;66(suppl 1):i162-i171. doi: 10.1093/geronb/gbr048 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Norton EC, Dowd BE, Maciejewski ML. Marginal effects—quantifying the effect of changes in risk factors in logistic regression models. JAMA. 2019;321(13):1304-1305. doi: 10.1001/jama.2019.1954 [DOI] [PubMed] [Google Scholar]
- 31.Norton EC, Dowd BE. Log odds and the interpretation of logit models. Health Serv Res. 2018;53(2):859-878. doi: 10.1111/1475-6773.12712 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Sun G-W, Shook TL, Kay GL. Inappropriate use of bivariable analysis to screen risk factors for use in multivariable analysis. J Clin Epidemiol. 1996;49(8):907-916. doi: 10.1016/0895-4356(96)00025-X [DOI] [PubMed] [Google Scholar]
- 33.VanderWeele TJ, Knol MJ. A tutorial on interaction. Epidemiol Methods. 2014;3(1):33-72. doi: 10.1515/em-2013-0005 [DOI] [Google Scholar]
- 34.Hosmer DW, Lemeshow S. Confidence interval estimation of interaction. Epidemiology. 1992;3(5):452-456. doi: 10.1097/00001648-199209000-00012 [DOI] [PubMed] [Google Scholar]
- 35.Correia K, Williams PL. Estimating the relative excess risk due to interaction in clustered-data settings. Am J Epidemiol. 2018;187(11):2470-2480. doi: 10.1093/aje/kwy154 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.SAS Institute Inc . Sample 63038: predictive margins and average marginal effects. Accessed August 6, 2021. http://support.sas.com/kb/63/038.html
- 37.Wolinsky FD, Callahan CM, Fitzgerald JF, Johnson RJ. The risk of nursing home placement and subsequent death among older adults. J Gerontol. 1992;47(4):S173-S182. doi: 10.1093/geronj/47.4.S173 [DOI] [PubMed] [Google Scholar]
- 38.Miller EA, Weissert WG. Predicting elderly people’s risk for nursing home placement, hospitalization, functional impairment, and mortality: a synthesis. Med Care Res Rev. 2000;57(3):259-297. doi: 10.1177/107755870005700301 [DOI] [PubMed] [Google Scholar]
- 39.Gaugler JE, Duval S, Anderson KA, Kane RL. Predicting nursing home admission in the US: a meta-analysis. BMC Geriatr. 2007;7:13. doi: 10.1186/1471-2318-7-13 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.Kasper JD, Pezzin LE, Rice JB. Stability and changes in living arrangements: relationship to nursing home admission and timing of placement. J Gerontol B Psychol Sci Soc Sci. 2010;65(6):783-791. doi: 10.1093/geronb/gbq023 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Charles KK, Sevak P. Can family caregiving substitute for nursing home care? J Health Econ. 2005;24(6):1174-1190. doi: 10.1016/j.jhealeco.2005.05.001 [DOI] [PubMed] [Google Scholar]
- 42.Rudberg MA, Sager MA, Zhang J. Risk factors for nursing home use after hospitalization for medical illness. J Gerontol A Biol Sci Med Sci. 1996;51(5):M189-M194. doi: 10.1093/gerona/51a.5.m189 [DOI] [PubMed] [Google Scholar]
- 43.Barrett L. Home and Community Preferences of the 45+ Population 2014. AARP Research; 2015. doi: 10.26419/res.00105.001 [DOI] [Google Scholar]
- 44.Sefcik JS, Ritter AZ, Flores EJ, et al. Why older adults may decline offers of post-acute care services: a qualitative descriptive study. Geriatr Nurs. 2017;38(3):238-243. doi: 10.1016/j.gerinurse.2016.11.003 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45.Burke RE, Jones J, Lawrence E, et al. Evaluating the quality of patient decision-making regarding post-acute care. J Gen Intern Med. 2018;33(5):678-684. doi: 10.1007/s11606-017-4298-1 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 46.Centers for Medicare & Medicaid Services . Medicare Skilled Nursing Facility (SNF) transparency data (CY2013). Published March 9, 2016. Accessed August 29, 2021. https://www.cms.gov/newsroom/fact-sheets/medicare-skilled-nursing-facility-snf-transparency-data-cy2013
- 47.Noel-Miller C. Medicare Beneficiaries’ Out-of-Pocket Spending for Health Care. AARP Public Policy Institute; 2020. doi: 10.26419/ppi.00105.001 [DOI] [Google Scholar]
- 48.Arora K, Wolf DA. Does paid family leave reduce nursing home use? the California experience. J Policy Anal Manage. 2018;37(1):38-62. doi: 10.1002/pam.22038 [DOI] [PubMed] [Google Scholar]
- 49.Avison C, Brock D, Campione J, et al. Outcome Evaluation of the National Family Caregiver Support Program. Westat; 2018:111. Accessed October 12, 2021. https://acl.gov/sites/default/files/programs/2018-12/Caregiver_Outcome_Evaluation_Final_Report.pdf [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
eTable 1. Cohort inclusion table by baseline interview wave
eTable 2. Nursing home cutoff sensitivity analysis
eTable 3. ADL dependence outcome corresponding to Table 3 in main manuscript
eTable 4. Nursing home outcome corresponding to Table 3 in main manuscript
eTable 5. Mortality outcome corresponding to Table 3 in main manuscript
eTable 6. ADL dependence outcome health shock interaction analysis corresponding to Figure 1 in main manuscript
eTable 7. Nursing home outcome health shock interaction analysis corresponding to Figure 1 in main manuscript
eTable 8. Mortality outcome health shock interaction analysis corresponding to Figure 1 in main manuscript
eFigure. Conceptual diagram demonstrating the interaction between social support and outcome with and without a stressor