Spouse and Child Availability for Newly Disabled Older Adults: Socioeconomic Differences and Potential Role of Residential Proximity

Abstract

Objectives

To examine the potential role of child and spousal availability in facilitating community-based care for disabled older adults.

Method

We used the Health and Retirement Study, a nationally representative longitudinal study of older adults. The analysis sample included older adults who were nondisabled at baseline, but who then developed at least one activity of daily living (ADL) limitation over the subsequent 2-year period (N = 2,094). Using multivariate, multinomial logistic regression, we examined the association of child and spouse availability prior to disablement of the older adults with ADL care receipt status after the onset of disablement, after adjusting for other sociodemographic and clinical characteristics.

Results

Lower socioeconomic status (SES) was associated with less availability of a spouse but greater availability of children at baseline. Compared with older adults who had no children nearby (i.e., all children lived further than 30 miles), older adults who had at least one child living with or near them prior to the onset of the ADL limitation were less likely to go to a nursing home (adjusted odds ratio [AOR]: 0.26 for coresident; AOR: 0.44 for 1– 30 miles) and less likely to depend on formal care (AOR: 0.39 for coresident; AOR: 0.51 for 1–30 miles) after the onset of new ADL limitation/s.

Discussion

Understanding SES variations in the informal care resources, and potential role of child geographic availability, may inform the development of cost-effective community-based care programs and policies.

Informal care from family and friends reduces the use of paid home health care and delays nursing home entry (Van Houtven & Norton, 2004; Sasso & Johnson, 2002). Many studies have examined the prevalence and profiles of informal care providers (Bianchi, Hotz, McGarry,& Seltzer, 2008; Hayman et al., 2001; Hickenbottom et al., 2002; Langa et al., 2002; Piette, Rosland, Silveira, Kabeto, & Langa, 2010; Spillman & Pezzin, 2000; Stone, 2000; Wolff & Kasper, 2006). The relative prevalence of care received from spouses and children varies by individual and family factors. Wives are the most common primary caregivers for disabled older men, but children are the most common primary caregivers for disabled older women (Bianchi et al., 2008; Katz, Kabeto, & Langa, 2000; Wolff & Kasper, 2006). Older white adults are more likely to receive help from a spouse than are older black adults, and black adults are more likely to rely on people outside the family for informal care than are whites (Burton et al., 1995). Children caring for same-gender parents are less likely to transfer out of the primary caregiver role than are children caring for a parent of the opposite gender (Allen, Lima, Goldscheider, & Roy, 2012).

There are differences in provision of informal care by socioeconomic status (SES). Using a British retirement survey of adults aged 55–69, Glaser and Grundy (2002) examined social class differences in care provision for respondents’ spouse and their own parents. Broese van Groenou and coworkers (2006) showed SES differences in older adults’ informal (defined as help from family members outside the household, neighbors, and friends) and formal help in four European countries. However, these studies did not differentiate between informal care from a spouse and informal care from a child. Because spouse and adult children are in different positions in their life courses, the opportunity cost of caring for older adults is likely to differ between spouse and adult children. Adult children of older adults are mostly young adults or middle-aged adults, whereas spouses of older adults are mostly near or at retirement age. To understand burdens of families and individuals caring for disabled older adults, it is important to disentangle SES differences in care sources between spouse and children.

Another important knowledge gap in the literature on informal care resources for older adults is geographic availability of children. As spouses mostly live together, the geographic availability of a spouse to a disabled older person is not usually an issue if a spouse is present. Some studies discussed the importance of close residential proximity of primary caregivers for functionally dependent older adults requiring specific care on a daily basis (Lee, Dwyer, & Coward, 1990; Litwak & Kulis, 1987). When older parents’ physical health declines, parents and children tend to move closer to each other (Silverstein, 1995), and older parents tend to select a target child with greater potential to provide support (Silverstein & Angelelli, 1998). Changing residences responding to older adults’ health declines may be a burden to the moving families and individuals, financially and nonfinancially, especially if health declines occur dramatically and unexpectedly. Close proximity to children prior to care need may play a significant role in promoting community-based care for older adults by lowering the risk of older adults moving to a nursing home during and after the onset of the disablement.

Accordingly, the aim of this study is to fill the knowledge gap in the literature on informal care resources, in particular, SES differences in child and spouse availability, and the role of residential proximity to children, by asking two questions: (a) Is an older adult’s SES associated with availability of a spouse and children for older adults, even after adjusting for health conditions and other characteristics? and (b) Do child and spousal availability facilitate activity of daily living (ADL)-specific care for older adults in the community after the onset of disability and lower the risk of disabled older adults’ transitioning to a nursing facility?

We hypothesize that (a) lower SES older adults have less availability of spouse but greater residential proximity (close distance) to children; (b) close proximity to a child reduces the risk of absence of ADL care receipt and risk of transition to a nursing home. Our hypotheses are based on prior research on SES variations in fertility, marriage, and health (Dye, 2010; Hyattsville, 2012; Kreider & Ellis, 2011), which are likely to shape informal care availability in older ages. Higher mortality and poor health in lower SES families will lead to less availability of spousal care resources in older ages because of absence of, or poor health of, spouse. Further, having a greater number of children and having children with greater restriction of geographic mobility among lower income families would imply a greater probability of lower SES older adults having a child in close proximity.

By disentangling informal care between intragenerational (spousal) support and intergenerational (child) support, results from the study will contribute to the informal care literature and will inform policy discussion on how best to address the burden of informal care for disabled older adults. By examining the role of child and spouse availability for care provision in the community, study findings also can inform the development of cost-effective, community-based programs.

Method

Data and Analysis Sample

We used the Health and Retirement Study (HRS), a biennial longitudinal survey of a nationally representative cohort of U.S. adults aged 51 and older (http://hrsonline.isr.umich.edu). Telephone or in-person interviews with HRS participants are conducted every 2 years. HRS collects extensive information about sociodemographic characteristics, health status, and health care. In recent survey years, HRS also has collected geographic information about residence of respondents’ children. Residential location of children is a key variable of interest for this article and was only available for the 2004–2008 surveys while we were preparing for this manuscript. Therefore, we used three waves of the HRS (2004, 2006, and 2008) for analyses. Our initial sample criteria include (a) aged 55 and older in 2004; (b) had at least one living child; (c) newly disabled in 2006 or 2008 (i.e., no ADL limitation in the last interview but at least one ADL limitation in the current interview); and (d) community dwelling during the last interview (i.e., 2 years prior to disablement). We have 2,094 respondents who meet the initial sample criteria and are representative of 7,717,730 older adults in the United States.

Analyses

For the descriptive table, we provided the weighted distributions of availability of child and spouse stratified by older adults’ health, demographic characteristics, and SES. The chi-square test was used to assess overall variations of child and spouse availability by each demographic and SES group. To determine SES differences in child and spouse availability prior to older adults’ disablement after adjusting for demographic characteristics and ADL severity, multinomial logistic regressions were used. Sample weights from 2004 survey were applied for all summary statistics and analysis estimates.

To assess the potential role of close residential proximity to a child for receipt of community-based care, we utilized the longitudinal structure of the HRS. To reduce potential endogeneity in measuring the role of prior residential proximity to a child after the onset of the ADL limitation, we first restricted the analysis sample to older adults who reported no ADL limitation during the prior interview (i.e., 2 years earlier) but reported at least one ADL limitation during the current interview. More specifically, to observe 2006 newly disabled older adults, we excluded those with an ADL limitation in 2004. To observe 2008 newly disabled older adults, we excluded those with an ADL limitation in 2006. Therefore, one respondent can only appear once in the analysis sample to meet this criterion. Our main predictors are measured prior to disablement of older adults. For example, geographic availability of children was measured when older adults did not have any ADL limitation 2 years prior to development of ADL limitation/s. In addition, a comorbidity index during the last interview (2 years prior) also was added to adjust for further potential health differences in the analysis sample even after the sample criterion of no ADL. A comorbidity index was preferred to other health measures such as self-rated health status because it is a relatively more objective measure. A sensitivity analysis by adding depression measure as well as self-rated health status measure did not change the effect size in the geographic proximity (results are not shown). We also controlled for demographic characteristics, SES, number of children, and working status of children.

The unit of observation for this analysis is respondents, and the sample weight from 2004 survey was applied for summary statistics and analysis estimates. As a sensitivity check concerning missing responses in covariates, we also repeated analyses by employing multiple imputation methods based on iterative switching regression techniques (Royston, 2005a,b). Because results remained consistent between nonimputed and imputed data set, we reported results from the nonimputed data. All analyses were conducted using STATA 12.

Measures

ADLs Limitations

Limitations in ADLs were identified with the following question: “Because of a health or memory problem, do you have difficulty with [bathing/ eating/dressing/walking across a room/getting in or out of bed/using the toilet]?” Each item variable is a dichotomous indicator of having difficulty. The ADL index was obtained by summing those six items, ranging from 0 to 6 limitations.

Availability of Spouse

We assessed the availability of a spouse based on (a) the presence of a spouse and (b) the spouse’s ADL limitation status. The presence of a spouse is defined based on whether the respondent had a married spouse or a partner during the interview. The ADL status of the spouse is defined using the ADL index, as described earlier. In sum, values of spousal availability include the following: 1 = Spouse with 0 ADL limitation; 2 = Spouse with at least one ADL limitation; 3 = No spouse present.

Availability of Children

Children in this article included biological and adopted children of respondents. The main purpose of excluding stepchildren in our definition is to reduce potential measurement error in child availability and reduce direct correlation with spousal availability. Stepchildren can only be defined by respondents’ spousal status, whereas biological/adopted children are defined independently from spousal status although is often related with it. Because we include both spouse and child availability in each estimation equation, it is rational to exclude stepchildren to reduce potential collinearity between child availability and spousal availability.

This article utilizes newly available location information of each child of respondents, which is a valuable complement to the existing proximity information. Without this additional information, the measure of distance to each child was available only at a 10-mile level. Although the 10-mile distinction has been used to measure close proximity in previous studies, the meaningful cut point might differ by specific outcome of interest. In addition, relying only on the proximity response by respondent is more likely subject to the endogeneity issue than is the location response. Respondents may be more likely to report a child living less than 10 miles distant if the child is in frequent contact, but less likely if the child is not. By using location information besides proximity information, this endogeneity issue may be minimized. In sum, we constructed a measure of residential proximity of children based on three questions in HRS. Using the household roster, we first identified coresident children of respondents. To identify distance between respondents and their non-coresident children, we utilized two sets of information. First, HRS asked family respondents whether each of their children lived within 10 miles or not. Since the 2004 survey, HRS also asked family respondents what those children’s zip codes were if they reported their children living further than 10 miles from the family respondents. Based on these responses, we obtained residential proximity between respondents and their children at levels—coresident, 10 miles, and distance between centroids of zip codes. Because our unit of analysis is respondents rather than children, we needed to summarize children’s geographic distance at the respondent level. More specifically, we used the residential distance to the closest child. The overall pattern of proximity between older adults and their adult children is consistent with descriptions from other nationally representative data (Lin & Rogerson, 1995).

After testing different cut points for close proximity (e.g., at 10 miles vs at 30 miles) and examining the sample size in each category, three exclusive proximity distinctions—coresident, less than or equal to 30 miles, and greater than 30 miles—were used in the analyses. As a result, the final measure of availability of children at a respondent’s level is (a) at least one child coresident; (b) at least one child within 30 miles; and (c) all children > 30 miles.

ADL Care Receipt, Nursing Home, and Primary ADL Care Provider

To measure respondents who received any help for ADLs, we use the following question: Does anyone ever help you [get across a room, dress, bath, eat, and get in and out of bed, or use the toilet]? Based on this question, we created an indicator of care receipt for completing at least one ADL. Nursing home residence was determined if the respondent is living in a nursing home or other health care facility at the time of interview. Using these, we created a categorical variable of ADL care receipt for an analysis: 1 = living in a community and received ADL care; 2 = living in a community and did not receive ADL care; 3 = moved to a nursing home.

To identify respondents’ primary helpers in completing ADL/s, we used responses to the following questions: Who most often helps with getting across a room, dressing, bathing, eating, and getting in and out of bed, or using the toilet? What is that person’s relationship to you? In the analysis, we focused on primary ADL care receipt from child or not (1 if they reported at least one child as a primary ADL care provider; 0 otherwise). We also examined an outcome of reporting a formal care as a primary ADL care provider (1 if they reported at least one formal care as a primary ADL care provider; 0 otherwise). Formal care includes care from an organization, employee of institution, paid helper, and professionals.

Socioeconomic Status

Education, occupation, income, and wealth are standard measures of SES. Because the majority of our sample were retired, occupation and labor income are not appropriate to reflect financial resources in our study sample. And, wealth measure includes assets from both respondent and spouse, which is likely to be correlated with spousal availability outcome, in particular, the presence of spouse. Therefore, we used levels of education of respondents as SES measure of older adults. The education levels include less than high school, GED/high school graduate, and at least some college education.

Covariates: Comorbidity, Demographics, Number of Children, and Working Status of Children

The comorbidity score (0–8) is the sum of eight indicators of whether a physician has ever told the respondent that he or she has ever had a particular disease. The eight diseases are high blood pressure, diabetes, cancer, lung disease, heart disease, stroke, psychiatric problems, and arthritis. We included respondents’ age (55–64, 65–74, 75–84, and 85 or older), gender, and race (white, black, and other) as demographic variables in all multivariate analyses. Number of children is a categorical variable of each respondent having one child, two children, and three or more children. Working status of children is an indicator of respondent having at least one child not working in the labor market.

Results

Variations in Spouse and Child Availability for Older Adults Who Currently Are Not disabled But About to Disable in 2 Years

Our analysis sample includes older adults who do not have any ADL limitation during interview in year T (=2004 and/or 2006), but who will have at least one ADL limitation in the subsequent 2 years, T + 2 (in 2006 or 2008). Using this sample, Table 1 presents the weighted distribution of child and spouse availability stratified by older adults’ characteristics in T (i.e., 2 years prior to disablement). Older adults who had a greater number of comorbidities are more likely to have children living in close proximity: 89% of nondisabled older adults who had four or more diseases had at least one child coresident or living nearby, whereas 78% of those who had none or one disease did have. In contrast, older adults who had a higher number of comorbidities tend to have less availability of spouse: only 42% of those with 4–6 diseases had nondisabled spouse, whereas 56% of older adults with 0–1 disease did have. Although the association between age and child availability is not clear, older age is significantly associated with lower availability of spouse: Over 50% of the adults aged 74 or younger had a nondisabled spouse, but only 19% of the adults aged 85 or older had a nondisabled spouse. Relative availability of child and spouse significantly differ by gender and race. Older women are more likely to have at least one child living with/in close proximity (85%) compared with older men (77%). However, older men are more likely to have a nondisabled spouse (65%) compared with older women (35%). Compared with whites, blacks are more likely to have at least one proximate child (90% vs 80%), but less likely to have a nondisabled spouse (30% vs 48%).

Table 1.

Weighted Percentages of Child and Spouse Availability Prior to Disablement by Comorbidity, Demographic Characteristics, and Socioeconomic Status

	Availability of children, T				Availability of spouse, T
	%	Have at least one child		All children > 30 miles	%	Spouse present and spouse’s ADL status		No spouse present
		coresident	living 1–30 miles			0 limitation	1–6 limitations
		%	%	%		%	%	%
Total	100	24.4	57.6	17.9	100	45.9	9.9	44.2
Comorbidity index, T
0–1	100	22.8	55.2	21.9	100	56.2	11.4	32.4
2	100	22.2	58.1	19.7	100	41.9	10.6	47.5
3	100	26.8	55.5	17.8	100	44.8	10.4	44.8
4–6	100	26.6	62.1	11.3	100	42.4	7.4	50.2
Age, T0
55–64	100	33.6	51.0	15.4	100	57.1	11.5	31.4
65–74	100	19.9	60.8	19.2	100	53.4	8.9	37.7
75–84	100	21.4	59.7	18.9	100	36.9	9.8	53.3
≥85	100	17.2	63.5	19.3	100	18.5	7.8	73.7
Gender
Male	100	21.4	55.9	22.8	100	64.5	14.6	20.9
Female	100	26.2	58.6	15.2	100	35.4	7.2	57.4
Race
White	100	21.7	58.9	19.4	100	48.0	10.3	41.8
Black	100	40.9	49.3	9.8	100	30.1	8.2	61.7
Other	100	39.7	52.4	7.9	100	41.1	5.9	53.0
Education
Less than high school	100	31.9	58.9	9.3	100	37.6	10.4	52.1
High school/GED certificate	100	21.9	59.9	18.1	100	45.9	8.9	45.2
At least some college	100	20.8	53.9	25.4	100	53.3	10.5	36.2

Notes. ADL = activity of daily living; GED = General Educational Development. Analysis sample criteria are (a) interviewed during 2004 and at least one more follow-up interview (in 2006 or 2008); (b) aged 55 and older in 2004; (c) had at least one living child; (d) reported at least one ADL limitation in a given year (2006, 2008); (e) reported no ADL limitation in the previous survey wave; (f) community dwelling in the previous survey wave (in 2004 or 2006). N = 2,094.

Does SES Predict Variations in Child and Spouse Availability for Older Adults?

There is a significant difference in relative availability of child and spouse by older adult’s education level, as shown in Table 1. Compared with those with at least some college education, older adults with less than high school education are more likely to have at least one proximate child (91% vs 75%) but are less likely to have a nondisabled spouse (38% vs 53%). To assess whether SES disparity in relative informal resources is mainly explained by health and demographic characteristics, we examined results from multivariate, multinomial regressions. Table 2 presents adjusted relative risk ratio (RRR) and adjusted predictions of child and spouse availability in association with respondents’ education level. Using the status of having all children further than 30 miles as the base outcome, lower education is significantly associated with a greater probability of having a coresident child (RRR = 3.66 for less than high school; RRR = 1.40 for high school/GED) and associated with greater probability of having a child in close proximity (RRR = 2.75 for less than high school; RRR = 1.48 for high school/GED). However, lower education is significantly associated with lower probability of having a nondisabled spouse (RRR = 0.58 for less than high school; RRR = 0.76 for high school/GED, base outcome = spouseless) compared with a college education.

Table 2.

Relative Risk Ratios and Predicted Probabilities of Child and Spouse Availability (Multinomial Logistic Regressions)

	Availability of child, T (multinomial logistic regression)			Availability of spouse, T (multinomial logistic regression)
	Coresident child	At least one child ≤ 30 miles	All children > 30 miles (base outcome)	Spouse with 0 ADL limitation	Spouse with at least one ADL limitation	No spouse present (base outcome)
Adjusted odds ratio (95% CI)
Less than high school	3.66*** (2.41, 5.56)	2.75*** (1.92, 3.94)	—	0.58*** (0.43, 0.78)	0.84 (0.55, 1.30)	—
High school/GED	1.40† (0.95, 2.07)	1.48* (1.09, 2.00)	—	0.76† (0.57, 1.01)	0.77 (0.49, 1.20)	—
At least some college	(reference)	(reference)	—	(reference)	(reference)	—
Adjusted prediction
Less than high School	32%	59%	9%	38%	10%	52%
High school/GED	22%	60%	18%	46%	9%	45%
At least some college	21%	54%	25%	53%	11%	36%
Model Wald χ2		147.5***			293.94***
N	2,078			2,063

Note. ADL = activity of daily living; CI = confidence interval. Analysis sample criteria are same as in Table 1. All covariates are listed in Table 1.

*p < .05. ***p < .001. ^†p < .10.

Does Prior Availability of Child and Spouse Predict a Community-based ADL Care Receipt and Risk of Transition to a Nursing Home?

Table 3 presents results of assessing the role of prior availability of child and spouse for older adults’ ADL care receipt after the onset of disablement. We controlled for number of children and prior working status of children as well as all covariates listed in Table 1. Prior availability of child is a significant predictor of community-based care receipt and transitioning to a nursing home. The adjusted prediction of remaining in a community and receiving ADL care is 32% if a newly disabled older adult had a coresident child, 26% if he/she had a proximate child, and 21% if he/she had no proximate child. The adjusted risk of moving to a nursing home is 4% if a newly disabled older adult had a coresident child, but 9% if he/she had no children nearby. A greater availability of spouse is significantly associated with lowering the risk of no ADL care receipt in a community setting and lowering the risk of transitioning to a nursing home. The adjusted prediction of remaining in a community and receiving ADL care is 31%–35% if he/she has a spouse but 20% if he/she does not. And the adjusted prediction of transitioning to a nursing home after the onset of ADL limitation/s is 3% if the older adult has a healthy spouse, but 8% if he/she does not have a spouse.

Table 3.

Relative Risk Ratios, Odds Ratios, and Predicted Probabilities of ADL Care Receipt Status, Transition to Nursing Home, and Primary ADL Helper

	Sample I: Previously community-dwelling, newly disabled older adults (multinomial regression)			Sample II: Additional criterion—having at least one ADL helper (logistic regression)
	Remaining in a community (between interviews T and T + 2)		Outcome 3: Transition to a nursing home between interviews T and T + 2	Receive primary ADL help from
	Outcome 1: At least one ADL helper, T + 2 (base outcome)	Outcome 2: No ADL helper, T + 2		Child, T + 2	Formal care, T + 2
RRR/OR (95% CI)
Child availability, T
Coresident	—	0.59 (0.40, 0.88)**	0.26 (0.13, 0.54)***	13.7 (5.83, 32.33)***	0.39 (0.20, 0.78)**
1–30 miles	—	0.80 (0.57, 1.12)	0.44 (0.26, 0.76)**	3.91 (1.78, 8.58)***	0.51 (0.30, 0.87)*
>30 miles	—	(reference)	(reference)	(reference)	(reference)
Spousal availability, T
Spouse with 0 ADL	—	0.46 (0.35, 0.61)***	0.24 (0.14, 0.42)***	0.24 (0.14, 0.41)***	0.21 (0.12, 0.37)***
Spouse with 1–6 ADLs	—	0.52 (0.35, 0.79)***	0.42 (0.19, 0.90)*	0.53 (0.27, 1.06)†	0.53 (0.27, 1.02)†
No spouse	—	(reference)	(reference)	(reference)	(reference)
Adjusted prediction
Child availability, T
Coresident	32%	64%	4%	52%	19%
1–30 miles	26%	69%	5%	28%	23%
>30 miles	21%	69%	9%	11%	34%
Spousal availability, T
Spouse with 0 ADL	35%	62%	3%	20%	12%
Spouse with 1–6 ADLs	31%	64%	5%	32%	24%
No spouse	20%	73%	8%	44%	36%
Model Wald χ2	262.7***			146.0***	70.24***
N	2,063			681	681

Note. ADL = activity of daily living; CI = confidence interval; OR = odds ratio; RRR = relative risk ratio. Covariates include number of children and employment status of children as well as all variables listed in Table 1.

*p < .05. **p < .01. ***p < .001. ^†p < .10.

Among those who reported at least one ADL helper, the probability of ADL care receipt from a child is significantly greater if the older adult had a child coresident (adjusted odds ratio [AOR] = 13.7) or living in close proximity (AOR = 3.9) prior to ADL limitation/s. If the older adult had a healthy spouse, the odds of primary care receipt from a child significantly decreases (AOR = 0.24). Accordingly, the probability of using a formal ADL care is significantly lower if an older adult had a healthy spouse or a proximate child. The adjusted prediction of formal care utilization is 19% if the older adult had a coresident child, 23% if he/she had a proximate child, and 34% if he/she had no proximate child prior to ADL limitation/s.

Discussion

Although children’s care role for disabled older adults becomes even more important in a rapidly aging society, the role of prior geographic availability of children has been underexplored. Using newly disabled older adults, findings in this article suggest that child availability prior to an older adult’s disablement may prevent the older adult from lacking assistance for daily activity and from transitioning to a nursing home after the onset of disablement. Prior geographic availability of a child also reduced dependence on formal care after the onset of disablement. These are important findings that inform the development of community-based care systems for disabled older adults. Residential changes during the onset of an older parent’s health decline might be costly and thus prevent families from providing care for daily activities. Family care plans and residential adjustment prior to an older parent’s disablement may have significant effects on moderating family burdens, hence promoting community-based care provision for older adults. This also has important policy implications. Public policies might contribute to lowering health care costs for a rapidly growing older population by encouraging family involvement in discussions about older adults’ health conditions and by supporting residential adjustments prior to an older adult’s disablement.

Lower SES older adults are more likely to have a child available at close proximity prior to disablement, even after controlling for comorbidity and demographic characteristics. Because low-income families are more likely to rely on Medicaid assistance to cover residence in nursing facilities, efficiently leveraging care resources of adult children is likely to be cost-effective. However, children of lower SES older adults are more likely to carry burdens of caring for disabled parents than are children of higher SES older adults; this may be attributable to the fact that lower SES older adults tend to lack spousal care resources. If lower SES adult children, who were not working in the labor market and who lived with/close to a parent, tend to become primary ADL caregivers once the parent becomes disabled, their future job attachments would be undermined. This would further increase the risk of intergenerational transmission of economic disadvantage. By focusing on older adults’ outcomes in this article, we did not assess burdens of the care-providing children. However, it is important that future studies assess children’s outcomes and SES disparity in the association with residential proximity to the older adult by using children as the unit of observation.

Our analyses have several potential limitations. The proximity measure was constructed based on older adults’ responses at different geographic proximity units—coresident, 10 miles, and zip code. To some extent, intergenerational, residential distance in this article is prone to measurement errors due to older adults’ perceptions and due to limited knowledge about the exact residential location of children. However, the overall pattern of proximity between older adults and their adult children is consistent with descriptions from other nationally representative data. We used distance in miles as the base of proximity measurement in order to make interpretation easier. Using jurisdictional boundaries—such as county—might be more informative for the discussion on public policy implementation.

In assessing the potential role of prior proximity to a child, there might be unobserved omitted variables that lead to endogeneity even after controlling for an extensive set of covariates and restricting the sample to newly disabled older adults. Moreover, children might have moved closer to parents in anticipation of imminent parental disablement. Although such endogeneity lessens after controlling for comorbidity and other covariates, we are not directly interpreting the estimates as causal relationships, but rather as a potential role of prior geographic availability of children. More rigorous methodological approaches are needed to identify causal relationships.

There might be significant variations by other individual and family characteristics in the potential role of prior spatial availability of children. Although these are important, we did not have a large enough sample to conduct stratified analyses or to test interaction terms. For example, although stratified analyses by gender provided similar patterns in terms of the potential role of child geographic proximity (results not shown), we did not have enough sample size to detect statistical significance in some estimates, especially among men.

Among older adults who reported at least one ADL limitation, more than half reported they did not receive any help. This might be attributable to the ADL limitation being relatively minor, but also might be attributable to the lack of availability of care resources. Future studies should further investigate the nature of care absence among disabled older adults. Although we focused on care provided by spouses and children, about one third of primary care providers were other individuals. Future studies should further assess the potentially significant roles of other informal care resources, especially for disabled older adults who have neither spouse nor children. And, we did not examine instrumental activities of daily living (IADL) such as using telephone, taking medication, handling money, shopping, and preparing meals. Close residential proximity of a child is likely to play a significant role in facilitating care for IADL as well. However, the role of a child’s geographic availability may vary widely by specific tasks of IADL (shopping vs preparing meals) compared with those of ADL. Thus, in this article, we chose to focus on ADLs.

Despite these limitations, this article provides new insights into informal care among newly disabled older adults. Our findings clarify variations in the availability of informal caregivers by SES and provide evidence on the significant contribution of prior residential proximity of a child to reducing formal care dependence after the onset of disablement. Study findings also can contribute to developing a cost-effective community-based health care system for the growing number of disabled older adults.