Living arrangements and disability-free life expectancy in the United States

Chi-Tsun Chiu

doi:10.1371/journal.pone.0211894

. 2019 Feb 8;14(2):e0211894. doi: 10.1371/journal.pone.0211894

Living arrangements and disability-free life expectancy in the United States

Chi-Tsun Chiu ^1,^*

Editor: Stefano Federici²

PMCID: PMC6368297 PMID: 30735515

Abstract

No studies have investigated the association between living arrangements and disability-free life expectancy in the United States, nor worldwide. This study aims to examine the differences in total and disability-free life expectancy among older Americans according to living arrangements. Data from the Health and Retirement Study (1998 to 2014) for non-Hispanic whites aged 50 and over (N = 21,612). Disability-free life expectancy by gender, living arrangement, and education are obtained from incidence-based multistate life tables. Overall, those who live only with their spouses/partners live 1–19 years longer with 3–25 more years without disability and 1–7 fewer years with disability than do those with other living arrangements. Among those with the same living arrangement, the higher educated live up to 6 years longer with up to 8 more years in a disability-free state and up to 2 fewer years in a disabled state. The study shows strong association between living arrangement and disability-free life expectancy by gender and education. Long-term care policy should take into account the length of life with/without disability by living arrangements and socioeconomic status and make use of the potential family resources.

Introduction

Rapid increases in the proportion of the population that is elderly is a global phenomenon. In the United States, the number and proportion of the elderly population, aged 65 and above, have increased over time and are projected to grow in the near future [1]. Meanwhile, family structures and household composition have gradually evolved over the past few decades, and the living arrangements of older Americans have changed as well. Household size decrease, and the proportion of one-person households has increased during the past decades [2,3]. In 2003, over 30% (10.5 million) of Americans aged 65 and above were living alone; among those who live alone, only one-quarter were men [4]. This is explained, mainly, by the fact that women have longer life expectancies than men. Older women are more likely than their male counterparts to be widowed, and older men are more likely to live with their spouse compared to older women [3,4].

Previous studies have provided possible explanations for why there is a significant association between living arrangement and health/mortality, including availability of resources (including social support and financial resources), degree of difficulty of living situations, regulation of health behavior, and demands on individual roles [5–9]. Most studies show that those who live alone are at greater risk of cognitive decline, poor physical/mental health, mortality incidence, and nursing home admissions compared to those not doing so [7,9–12], because living alone, compared to other co-residence living arrangements, is less favorable to social/financial support and adopting a healthy lifestyle [5,6,8,9]. There may also be psychological consequences to living alone, such as negative perceptions of health, depression, and loneliness [13,14]. However, living alone does not always mean being lonely or feeling lonely [15]. A previous study using data from China shows that older women living alone have fewer health problems compared to older women living with children and others, but the same does not apply to men; a possible explanation for this is the traditional caregiving role of women [16]. Since living arrangements are associated with health and mortality, it is crucial to know whether older Americans are living more healthily as well as longer in these new living arrangements.

This study aims to examine these matters by investigating the association between living arrangements and disability-free life expectancy. Despite extended life expectancy, older persons may suffer from disability and illness during their later years [17,18]. Living longer does not necessarily signify living healthier. Therefore, it is important to consider not only length of life, but also the quality of life (e.g., disability-free life expectancy) of older Americans in terms of population health and policy implications for public health. Disability-free expectancy–the average number of years that a person can expect to live without disability–combines information about both mortality and disability, and indicates a potential future need for health care [17,19,20]. Living arrangements represent a person’s immediate social environment and often imply how and where older people will receive formal and informal care [21,22]. It is worth examining how different types of living arrangements are associated with health, mortality and disability-free life expectancy.

Methods

Data

This study employs the Health and Retirement Study (HRS), a biennial survey spanning the years 1992 to 2014, primarily sponsored by the National Institute of Aging (NIA) and conducted by the University of Michigan. The HRS is a nationally representative longitudinal survey of Americans aged 51 years and older, and their spouses, in the United States. The institutionalized population was not initially included in the HRS sample in the baseline interview, but respondents were followed into institutions, so institutional respondents are also included in the current data. Since the HRS is a longitudinal survey, survival status is available from both National Death Index and follow-up interviews. In order to calculate life expectancy and disability-free life expectancy according to different living arrangements (including living in nursing homes), this study includes an institutional sample and uses the weight that is the combined individual-level weight and the nursing home weight. This study makes use of the HRS Rand file (version P) and Rand Core fat files (1998–2014), and the analytical sample comprises individuals aged 50 and over. The study only includes HRS data from 1998–2014 because of the consistency of disability measurement [23,24].

Disability measures the gap between personal capability and environmental demands, and can be mitigated on either side, by increasing capability or reducing demand [25]. In the study, disability is measured by difficulties encountered in undertaking the activities of daily living (ADL) that are necessary for survival [25], and with a composite summary measure for ADL, (RwADLA, wherein "w" refers to the wave number) provided by Rand which is used here [23]. The ADL measure includes items such as dressing, bathing, eating, walking, and getting in and out of bed. Respondents with any ADL disability will be coded as disabled.

Other socio-demographic measures include age, gender, and education. Age is recorded at the time of interview for those who are 50 and over. Gender is an important variable for this study because living arrangements have a stronger impact for men than women in terms of health/mortality [5,16]. All analyses are stratified by gender.

Education is often viewed as a fundamental cause [26,27] of health/mortality disparities and is highly correlated with living arrangements. It may be that income, wealth, and/or education are the reasons for differences in life expectancy and disability-free life expectancy between cohabiting and single older adults rather than living arrangements per se. Therefore, education, as the key socioeconomic status, is considered in the model. Educational attainment at the time of the interview was measured according to each respondent’s level of education in each survey. This study uses a binary version of the education variable based on previous research showing that high school (HS) graduates and those with a higher degree have significantly better health/mortality profile compared to those with less than a high school degree [28,29]. Due to small cell sizes, this study is not able to use a refined education variable.

All analyses in the study are restricted to non-Hispanic white respondents aged 50 and older with complete information on age, gender, education, living arrangement, and valid sample weights in 1998–2014. Those with information missing regarding these variables are excluded. Focusing on non-Hispanic whites is not only to avoid small cell sizes of other races and Hispanic groups, but also to improve data quality and to reduce population heterogeneity.

Differing living arrangements imply different demands and resources facing older Americans [7,30]. This is so for three reasons:

spouses/partners are oftentimes the immediate informal primary caregiver in the household [16,31];
living alone is strongly associated with health/mortality [5–11,13,14,16] and
nursing home admissions are associated with severe health problems [12].

The study distinguishes between five types of living arrangements and accounts for the changes in living arrangements during the survey period. Therefore, living arrangements at the beginning of the interval are not necessarily the same as those at the end of the interval.

Couple only: married/cohabiters living with their spouses/partners only.
Couple with children/others: married/cohabiters living with their spouses/partners, and children and/or others in the same household.
Single only: people (who are separated, widowed, divorced, never married, or not cohabiting) living alone.
Single with children/others: people (who are separated, widowed, divorced, never married, or not cohabiting) living with children/others in the same household.
Nursing home: people living in nursing homes.

The married and cohabiters are viewed as couples in the study because previous work has found no significant mortality differences between married and cohabitating couples [32]. Couples or singles who are living with their children may also be living with others. Previous studies have shown the significance of children’s socioeconomic status on parents’ health and mortality [33–36] and family members’ caregiving roles, especially spouses/partners not present in the same household [37]. Older adults who are single and live with children and/or others may in poor health [16]. Single adults living with others include a substantial fraction of older Americans living in another’s household [30]. Living arrangements could change across waves. The living arrangements framework in this study is inspired by the designs of previous studies [7,30]. In their studies using the first one or two waves of the HRS there were almost no cases of individuals living in nursing homes. However, as the HRS follows respondents into institutions, and this study makes use of HRS waves four through twelve, respondents living in nursing homes are included from the analysis.

Analytic approach

Total life expectancy (TLE), disability-free life expectancy (DFLE) and disabled life expectancy (DLE) are obtained from incidence-based multistate life tables (MSLTs) that are estimated using a relatively recent algorithm, Stochastic Population Analysis for Complex Events (SPACE) [38]. The SPACE is composed from SAS macros with SAS IML and CONNECT components. The SPACE approach offers some advantages over traditional ways of estimating MSLTs, namely, the use of microsimulation and the bootstrap method to estimate the sampling variability of MSLT functions. Here, a simulation cohort size of 100,000 is used along with 300 bootstrap samples. Via the bootstrap technique, statistically testing whether differences between groups are statistically significant becomes possible. Confidence intervals of 95% are constructed for the distribution of the between-group difference in estimates, which is from the 2.5th to the 97.5th percentile. The analyses are stratified by gender, with living arrangements and education serving as covariates in the model. Below are models to estimate age-specific state-dependent transition probabilities using multinomial logistic regression:

Model 1: age + living arrangement
Model 2: Model 1 + education + interaction of living arrangement and education

The possible six transitions are listed as follow:

Disability-free to disability-free
Disability-free to disabled
Disability-free to dead
Disabled to disabled
Disabled to disability-free
Disabled to dead

Living arrangement is treated as a time-varying covariate in the model. Within each respondent in the long-form analysis data file input to the SPACE program, each respondent is allowed to have different living arrangement states across different rows since living arrangements of respondents could change across waves. Therefore, the analysis is able to capture the changing nature of living arrangements and to reflect the complicated association between living arrangements and health expectancy.

SPACE calculates disability-free life expectancy with the input data and settings. SPACE provides both population-based and status-based MSLTs. For more detailed technical information about the SPACE program, please refer to Cai and colleagues’ work [38].

Results

There are 21,612 non-Hispanic whites included in the analysis sample during the last observations. Among these, 5,941 men and 7,798 women are alive; 3,771 men and 4,102 women are dead. Among 9,712 men, 1,708 have less than a high school degree, while 8,004 have a high school degree or above; among 11,900 women, 2,167 have less than a high school degree while 9,733 have a high school degree or above. There are 10,120 non-Hispanic whites (4,127 men and 5,993 women) who change their living arrangements during the analysis period.

Table 1 displays the numbers of interview records of time-varying living arrangements by gender for non-Hispanic whites. Among interview records of non-Hispanic white men, the most common living arrangement is married/cohabiters living only with their spouses/partners (30,191, 51.8%); followed by married/cohabiters living with their spouses/partners and children and/or others (9,566, 22.7%) and single men living alone (7,578, 16.6%); the least common is living in a nursing home (715, 1.0%). For non-Hispanic white women, the most common living arrangement is also married/cohabiters living only with their spouses/partners (29,146, 41.1%); followed by single women living alone (19,420, 28.9%) and married/cohabiters living with their spouses/partners and children and/or others (7,752, 14.2%); the least common is living in a nursing home (1,936, 2.2%). Table 2 shows the numbers of interview records by gender for non-Hispanic whites across waves. Among the total 119,105 interview records, about 9% - 13% of interview records appear in each wave for men, women, and both men and women combined.

Table 1. The numbers of interview records of living arrangement by gender for non-Hispanic whites.

Living Arrangements	Men		Women		Total
Living Arrangements	#	%	#	%	#	%
Couple only	30,191	51.8	29,146	41.1	59,337	46.0
Couple with children/others	9,566	22.7	7,752	14.2	17,318	18.1
Single alone	7,578	16.6	19,420	28.9	26,998	23.3
Single with children/others	3,703	7.9	9,098	13.6	12,801	11.0
Nursing home	715	1.0	1,936	2.2	2,651	1.7
Total	51,753	100.0	67,352	100.0	119,105	100.0

Open in a new tab

Note. Numbers of interview records (#) are unweighted, but proportions (%) are weighted. Data from Health and Retirement Study.

Table 2. The numbers of interview records by gender for non-Hispanic whites across waves.

Interview Waves	Men		Women		Total
Interview Waves	#	%	#	%	#	%
1998	6,779	9.9	8,529	10.0	15,308	10.0
2000	6,209	9.4	8,038	9.7	14,247	9.6
2002	5,672	9.0	7,461	9.5	13,133	9.3
2004	6,159	11.7	8,062	11.7	14,221	11.7
2006	5,664	11.2	7,471	11.2	13,135	11.2
2008	5,204	10.5	6,916	10.6	12,120	10.6
2010	5,906	13.4	7,618	13.1	13,524	13.2
2012	5,446	12.8	7,081	12.6	12,527	12.7
2014	4,714	12.1	6,176	11.7	10,890	11.9
Total	51,753	100.0	67,352	100.0	119,105	100.0

Open in a new tab

Note. Numbers of interview records (#) are unweighted, but proportions (%) are weighted. Data from Health and Retirement Study.

Table 3 compares vital statistics and estimated results by gender for Americans at age 50. The estimated total life expectancies are 29.02, and 32.86 years for non-Hispanic white men and women respectively, while life expectancies from vital statistics at age 50 for non-Hispanic white men and women are 28.96, and 32.63 years respectively. The differences between vital statistics and estimated total life expectancies are all within a half year, and vital statistics fall within 95% conference intervals.

Table 3. Vital statistics of life expectancy and estimated total, disability-free, and disabled life expectancy by gender for Americans at age 50.

	Vital Statistics	Estimated
Gender	TLE	TLE	95% CI
Men	28.96	29.02	(28.6–29.4)
Women	32.63	32.86	(32.5–33.3)

Open in a new tab

Note. The vital statistics are from the life table in 2006 from Tables 2 and 3 of Arias (2010) [51] with tables updated using revised intercensal populations. TLE = total life expectancy. 95% CI = 95% confidence interval. Data from Health and Retirement Study.

Tables 4–6 presents total, disability-free, and disabled life expectancy by gender and living arrangement for non-Hispanic whites at age 50, and different patterns can be seen according to gender and initial health state. In Table 4 for population-based estimates, among men, those living in a couple only survive the longest (TLE = 30.8), followed by those living in a couple with children/others (TLE = 29.8), single only (TLE = 28.6), single with children/others (TLE = 26.8), and nursing home (TLE = 15.2). Compared to couple only, living arrangements of singles and nursing home have shorter TLE, shorter DFLE, and longer DLE (p<0.05). Singles living with children or others also have shorter TLE, shorter DFLE, and longer DLE (p<0.05) than couples living with children or others. Men living with spouses/partners enjoys the longest years of total life and disability-free life (DFLE = 27.3) and the shortest disabled life (DLE = 3.4), where men in nursing homes are the opposite. Among women, those living in a couple only arrangement live the longest (TLE = 35.9), followed by single only (TLE = 34.8), couple with children/others (TLE = 32.7), single with children/others (TLE = 30.0), and nursing home (TLE = 16.6). Slightly different patterns are observed compared to men. Compared to couple only, all other living arrangements have shorter TLE, shorter DFLE, and longer DLE (p<0.05). Women living with spouses/partners enjoy the longest years of total life and disability-free life (DFLE = 31.5) and the shortest disabled life (DLE = 4.5), where women in nursing home are the opposite. Very similar patterns by gender can be observed in status-based estimates, whether initial health states are disability-free or disabled (Tables 5 and 6).

Table 4. Total, disability-free, and disabled life expectancy by gender and living arrangement for non-Hispanic whites at age 50: Population-based estimates.

Gender	Living Arrangement	TLE	SE		DFLE	SE		DLE	SE		DFLE(%)
Men	Couple only	30.8	(0.27)		27.4	(0.31)		3.4	(0.12)		89.0
	Couple with children/others	29.8	(0.64)		26.0	(0.60)		3.9	(0.32)		87.2
	Single only	28.6	(0.57)	^a ^b	24.3	(0.61)	^a ^b	4.3	(0.26)	^a ^b	85.0
	Single with children/others	26.8	(0.62)	^a ^b	22.0	(0.58)	^a ^b	4.8	(0.39)	^a	82.1
	Nursing home	15.2	(0.88)	^a	4.6	(0.72)	^a	10.6	(0.68)	^a	30.3
Women	Couple only	35.9	(0.34)		31.5	(0.34)		4.5	(0.17)		87.7
	Couple with children/others	32.7	(0.87)	^a	28.1	(0.70)	^a	4.6	(0.40)		85.9
	Single only	34.8	(0.34)	^a ^b	29.0	(0.33)	^a ^b	5.8	(0.19)	^a ^b	83.3
	Single with children/others	30.0	(0.41)	^a ^b	23.3	(0.51)	^a ^b	6.7	(0.37)	^a ^b	77.7
	Nursing home	16.6	(0.81)	^a	3.9	(0.57)	^a	12.8	(0.70)	^a	23.5

Open in a new tab

Note. TLE = total life expectancy. DFLE = disability-free life expectancy. DLE = disabled life expectancy. %DFLE = the proportion of disability-free life expectancy over total life expectancy. 95% CI = 95% confidence interval. Data from Health and Retirement Study.