Measuring Spatial Availability of Children for Older Adults with Disability

Abstract

Background:

Adult children are often crucial care providers for older adults in need of help with daily activities. Previous research showed that having an adult child living with or nearby may help prevent parents with disability from moving to nursing homes and reduce using paid care services. Despite the profound implications for care resources and utilization among older adults, there is no measure summarizing the spatial availability of all adult children for an older adult, which accounts for potential care associated with spatial proximity to each of the adult children. Our study aims to develop a holistic measure of children’s spatial availability to assess potential family care resources for older adults with disability.

Methods:

Data were obtained from the population-based, longitudinal cohort study in the U.S., the Health and Retirement Study (HRS). We selected a nationally representative sample of older adults 55 years and older who has a limitation with any daily activities. Using multivariable two-part regression models, we developed a care-weighted child spatial availability for caregiving (CSAC) index, which summarizes the spatial dispersion of all adult children accounting for the level of caregiving associated with distance to each adult child. We also constructed a secondary or reference index of child spatial proximity (CSP) focusing only on spatial distance (i.e., without incorporating the care aspect associated with distance) by employing Gaussian kernel weighting functions on parent-adult child distances.

Results:

CSAC index (primary index) highlights the great importance of having an adult child in the same household (vs. nearby or far) for receiving care among older adults with disability, compared to the CSP index (secondary index), showing relatively a gradual decline over the spatial distance. Both CSAC and CSP indices vary substantially by older adults’ sociodemographic attributes.

Conclusions:

The holistic indices of child spatial availability will contribute to assessing care resources for older adults, albeit future development is needed to extend the utility of the indices tailored for specific care needs.

INTRODUCTION

Family members play a vital role in providing care to older adults with disability (Bianchi et al., 2010; Lin & Wolf, 2020; Schulz & Eden, 2016), which has profound implications for public healthcare as the U.S. older population is growing and the demand for long-term care is increasing substantially. Evidence suggests physical and cognitive health of older adults have not improved in recent decades and some health and function even worsened in recent years (Chen & Sloan, 2015; Choi et al., 2016, 2018; Freedman et al., 2013; Martin et al., 2010; Tipirneni et al., 2020) with poorer condition among younger birth cohorts of middle and older adults than older cohorts (Choi & Schoeni, 2017). A higher proportion of older adults is likely to require long-term care services associated with chronic disability as they age, and thus, out-of-pocket long-term care costs may result in a substantial financial burden on families (Favreault & Dey, 2016).

Many older adults with limitations in physical and cognitive function do not have a spouse or have a spouse who also has difficulty with daily activities due to poor health, but most of them have at least one adult child (Choi et al., 2015b; Choi, Heisler, et al., 2021). Having an adult child in the same household or nearby has been identified as a significant factor in lowering the likelihood of an older adult with a disability moving to a residential care facility and reducing other paid care services used by older adults (Artamonova et al., 2023; Charles & Sevak, 2005; Choi et al., 2015a; Choi, Heisler, et al., 2021; Van Der Pers et al., 2015). Even if living far away, adult children provide instrumental support such as managing bills and grocery shopping via online services (Albers et al., 2022; Bei et al., 2020; Cagle & Munn, 2012). However, close spatial proximity (living in the same house or neighboring) may be a critical facilitator to assist older adults with daily activities such as walking, bathing, and using the toilet.

While literature showed the evident role of spatial proximity of adult children in caregiving, there were limitations in measuring the spatial availability of parents and adult children. For example, to summarize the spatial availability of adult children, previous studies focused on the spatial proximity of a parent to the nearest adult child or median distance (Artamonova et al., 2023; Choi et al., 2015b; Compton & Pollak, 2015; Van Der Pers et al., 2015) instead of accounting for the distances to all adult children. Choi et al. (2020) adopted a relatively more holistic approach in measuring the spatial dispersion of adult children by estimating the share of parents who had all adult children at a given distance. However, to our best knowledge, no prior study developed a measure of child spatial availability that directly incorporates geographic proximity to all their adult children by accounting for the care amounts that an older adult received from each of all adult children associated with the proximity (i.e., child spatial proximity in the caregiving context).

This study aims to develop an adult-child spatial availability index, which is a weighted summary scale for each older adult with a disability, accounting for potential care availability associated with spatial proximity to all adult children. We used population-based cohort study that collects a nationally representative sample of older adults in the U.S. We included a broad relationship type of children (i.e., biological, adopted, stepchildren, and in-laws) to address an increasingly diverse and complicated family structure in the U.S. in recent years (Wiemers et al., 2019; Pearce et al., 2018). The measure of a care-weighted index of adult-child spatial availability will advance the literature by improving the understanding of family care resources accessible for older adults with disability – from all adult children rather than focusing on one, such as the closest adult child. Furthermore, the index will contribute to identifying older adults at risk of fewer family care resources which may be associated with their demographic and socioeconomic status.

METHOD

Data and Sample

The Health and Retirement Study (HRS), approved by the University of Michigan Institutional Review Board, is a national, longitudinal study that biennially surveys adults aged 51 and older in the U.S. since 1992. We used the biennial data of children from 2010 to 2018 for the following two considerations. First, the sequence of questions about children’s geographic information is consistent from 2010 onward, though the locations of all children have been collected since 2004 (Child ZIP Codes 2004–2020, 2023). Second, children’s sociodemographic characteristics harmonized across survey waves are not available beyond the survey year of 2018 in the RAND HRS Family Data (Bugliari et al., 2023), which is one of the primary data sources. Therefore, we used the children’s sociodemographic and geographic information from the survey years of 2010 – 2018, to ensure measurement consistency over time.

Our baseline sample of older adults includes sample persons 55 and older with disability. Disability was defined as the status of having some difficulties in at least one of the 11 activities: six ADLs (dressing, eating, bathing, walking across the room, getting in/out of bed, using the toilet) or the five instrumental ADLs (IADLs; using the telephone, taking medication, shopping, preparing meals, managing money). A small fraction of children’s records (roughly 2.3%) were excluded from the analysis sample because the children were not uniquely identified across survey waves, e.g., duplicates of child identifiers (Bugliari et al., 2023). There are 9,029 older adults and 27,201 dyads of parents and adult children in the study sample.

Measures

We constructed a semi-continuous variable of the residential distances of parent-adult child dyads according to a three-fold process. First, using household roster and parent self-reports of proximity to each child, we identified the dyads that lived in the same house or within 10 miles from one another. Then, we indicated the dyads living in the same ZIP Code area, which is the most granular information on geographic areas available for adult children. Further, we computed dyad residential great-circle distances between the centroids of the ZIP Code areas where a parent and each adult child resided if the dyad was in different ZIP Code areas or more than 10 miles apart. We reclassified the semi-continuous variable into eight groups: 1) coresident, 2) within 10 miles or in the same ZIP Code area, 3) within 30 miles, 4) 30 (exclusive) to 100 miles, 5) 100–200 miles, 6) 200–500 miles, 7) 500–1,000 miles, and 8) beyond 1,000 miles. The distances are top-coded to 2,150 miles (99 percentile value) to address the influence of extreme values.

The amount of care that a parent received from each child was measured as monthly hours of ADL or IADL help after incorporating multiple variables – hours of care per day and associated information on reference period (i.e., daily or not; the number of days per month or per week). In the original questionnaire, each parent was asked about one of the three questions related to the number of days of help per helper (e.g., child): a) “How many days of help did the helper help during the last month?”; b) “How many days per week did the helper help during the last month?”; or c) “Did the helper help every day during the last month?” Then, parents reported the number of hours on those days that a helper had helped in the last month. Assuming 8 hours of sleep, the maximum daily hours of help per helper was set to 16 hours. The final amount of care from each child is the number of days of help in the last month multiplied by the corresponding number of hours per day. Zero hours of care represent either no help from the child at all or no help in the previous month of the survey. For our analysis, we used the data of care amount that were surveyed two years after the measure of dyad proximity to reduce the endogeneity and reverse relationship; for example, parents and adult children may have moved closer to each other to facilitate the care. In the original questionnaires, care amount was measured during the preceding month of the interview (e.g., “During the last month, how many days did the helper help you?”), and spatial proximity was measured at the time of the interview (e.g., “Do any of your children who do not live with you live within 10 miles of you?”). We thus used the lagged proximity as the predictor to ensure it was measured prior to the care outcome in the estimation model.

Other covariates, measured at the same time-point as the measure of dyad residential distances (biennially during 2010 and 2018), are survey year, parents’ sociodemographic characteristics and health status, as well as adult children’s sociodemographic characteristics as follows: parents’ age, gender, race and ethnicity (non-Hispanic White, non-Hispanic Black, non-Hispanic Others, Hispanic), foreign-born status, categorical years of schooling (<12, 12, 13–15, 16+ years), marital status (married/partnered, separated/divorced/never married, widowed), and the number of limitations in ADLs and IADLs; adult children’s age (linear and quadratic terms), gender, child type (biological or adopted children, step-children, and children-in-law or others), categorical years of schooling (<12, 12, 13–15, 16+ years), marital status (married/partnered vs. not), and employment status (full-time employment vs. not). Other types of children may include former step-children, former children-in-law, and any type of children other than those specified in the original survey questionnaire (e.g., foster children).

Approach

Our primary aim is to develop an index of the adult child spatial availability for care (CSAC) to measure the care-weighted spatial availability of all adult children for parents with a disability. As a comparison measure, we also created a secondary index of adult child spatial proximity (CSP) based only on dyad residential distances (i.e., without incorporating the caregiving amount associated with the distance). The constructions of CSAC and CSP indices consist of two overall steps: 1) computing spatial availability scores at the dyad level (dyad of parents and their adult children); and 2) aggregating the scores over dyads and assigning zero values to those who do not have any adult children to create a summary scale at the older adult level. In the first step of estimating dyad-level scores, we employed regression models (for CSAC) or formulas (for CSP) on residential proximity between parents and each of their adult children. This means that dyads of a parent and minor child (under the age of 18) and those older adults who did not report any children alive at the survey time were excluded (3,953 dyad-year observations). In the second step summarizing spatial availability of adult children, we assigned zero index values to those older adults with no adult child (209 older adults), reflecting the lowest adult child spatial availability.

Construction of the Child Spatial Availability for Care (CSAC) Index (primary index)

The dyad-level CSAC scores are the estimated amount of ADL/IADL care received by older adults from each adult child corresponding to the dyad residential proximity category. We adopted two analysis approaches to measuring dyad-level CSAC scores - the base and interaction models. For the base model, we performed a multivariable two-part model (Belotti et al., 2015), where the first part is a logit and the second part is a generalized linear model (GLM) with a log link, given the massive amount of zero hours in the ADL/IADL care. The model is adjusted for sampling weights, household clusters, and survey sampling stratum to account for nonresponse rates and dependency between dyads per household, as well as project the estimates into the U.S. older population in each survey wave. The specification of the model is in formula (1).

$$g(y_{ij}|{Proximity}_{ij},X_{ij})=\left\{\begin{array}{cc}1-P(y_{ij}>0|{{Proximity}_{ij},X}_{ij})\hfill & \hfill if{y}_{ij}=0\\ P(y_{ij}>0|{{Proximity}_{ij},X}_{ij})\times \mathrm{f}({\beta }_0+{\beta }_1{Proximity}_{ij}+{\beta }_2{X}_{ij}+{ϵ}_{ij})\hfill & \hfill if{y}_{ij}>0\end{array}\right.$$ (1)

The outcome ($y_{ij}$) is the monthly hours of ADL/IADL help received by parent $i$ from adult child $j$, measured two years succeeding the primary predictor of categorical dyad residential proximity (${Proximity}_{ij}$). Other model covariates ($X_{ij}$) are survey years and sociodemographic characteristics of parents and adult children listed above. The first part of the logit model determines the likelihood of a parent receiving any care (positive vs. zero care hours) from each adult child. The second part of the GLM with a log link further fitted the number of care hours among parents who received help. The coefficients of the spatial proximity category estimating monthly care hours in both parts of the base model are provided in Supplemental Table 1.

The predicted values based on the estimated effects of dyad spatial proximity from the two-part model were calculated, holding the other covariates at their mean values of the sample (formula (2)).

$$S_{ij}=\left\{\begin{array}{cc}1-{\left[1+\exp \left({-{\beta }_0^{Logit}-\beta }_1^{Logit}{Proximity}_{ij}-{\beta }_2^{Logit}\stackrel{-}{X}\right)\right]}^{-1}\hfill & \hfill if{y}_{ij}=0\\ {\left[1+\exp \left(-{\beta }_0^{Logit}-{\beta }_k^{Logit}{Proximity}_{ij}-{\beta }_2^{Logit}\stackrel{-}{X}\right)\right]}^{-1}\times \exp \left({\beta }_0^{GLM}+{\beta }_k^{GLM}{Proximity}_{ij}+{\beta }_2^{GLM}\stackrel{-}{X}\right)& \hfill if{y}_{ij}>0\end{array}\right.$$ (2)

The predicted values serve as the dyad-level care-weighted CSAC scores. A larger score reflects greater spatial availability of an adult child to care for the parent. More details on the dyad-level CSAC scores by spatial proximity are mentioned in the results section.

The care involvement of an adult child may be affected by the pool of caregivers in the family. For example, the care amount received by an older adult from each adult child may be less if the older adult has multiple adult children although the total care amount increases with the number of adult children. This may reflect diminishing marginal productivity or care allocation among adult children. To account for the potentially differing effect of proximity by the number of adult children, we extended the base model in estimating the CSAC dyad-level scores by including interaction terms between the proximity category and the number of children (1, 2, 3+); hereafter, the interaction model. Due to the small sample size among the far distance categories (especially for the second part in the two-part model focusing on those with positive care hours), we collapsed the proximity categories beyond 100 miles.

The older adult-level CSAC index is the sum of the dyad-level scores for each older adult with a disability. We assigned zero values to the CSAC index for those older adults without any adult child (i.e., having no child or having minor children only) at the time of the interview. The final CSAC index was rescaled to a range of 0 to 100; the fraction of the original CSAC index over the difference in the highest and lowest index value, which was multiplied by 100.

Construction of the Child Spatial Proximity (CSP) Index (secondary index)

The CSP index is a function of residential distances between parents and adult children and the number of adult children; however, it does not account for the potential level of caregiving associated with distance. We first calculated a dyad-level score using a spatial decaying model, Gaussian kernel weighting function (Brunsdon et al., 1998) as specified below.

$$\mathrm{e}\mathrm{x}\mathrm{p}\left\{-\frac{{d_{ij}}^2}{2b^2}\right\}$$ (3)

$d_{ij}$ is the square-root transformed dyad residential distances between parent $i$ and adult child $j$; and $b$ is the bandwidth of the sampling-weighted interquartile range of $d_{ij}$. A larger score reflects greater spatial availability to an adult child for the parent.

Next, the older adult-level CSP index is the aggregate of CSP dyad-level scores of all adult children for each older adult, ranging from 0 to 100. Zero values were assigned to the CSP index for those who did not have any adult children at the time of the interview. The CSP is possibly sensitive to the selection of bandwidth in the Gaussian kernel weighting function. To investigate this, we conducted a sensitivity analysis to re-construct the CSP index by using different bandwidths; one, two, and three weighted standard deviations (S.D.s) of the square-root transformed dyad residential distances, separately. We found that the bandwidths barely changed the CSP scores where relevant (Supplemental Figure 1).

Sensitivity and Auxiliary Analyses

We employed several auxiliary and sensitivity analyses to provide more nuanced evidence and assess variations in the estimates of CSAC. First, the significance of an older adult having a coresident adult child in regard to care receipt may be lessened if the older adult has a spouse without disability, who is likely to be the primary caregiver. W thus reran the base model stratified by parents’ spousal disability status (no spouse or partner, a spouse/partner with disability, and a spouse/partner without disability) to assess the extent to which the potential effect of spatial proximity differ by the older adult’s spousal availability. Second, the spatial proximity of parents and adult children is likely affected by their life stage. We thus reran the base model stratified by parents’ age groups (<65, 65–74, 75–84, and 85+ years of age) to provide insights into the potential implications of proximity for care across age groups. Third, the role of close spatial distance is likely to differ by specific care activities associated with care needs. For example, assistance with ADLs (e.g., bathing and walking across rooms) may be more contingent on close distance (including coresidence) compared to assistance with IADLs (e.g., managing bills and grocery shopping). We reran separate base model specific to ADLs and IADLs to examine the extent to which the dyad-level CSAC score differs by the type of care. Fourth, we reran the base model with a contemporary design meaning the estimated model for dyad-level CSAC included the main predictor (spatial proximity category) and the outcome (care hours) surveyed in the same year. The comparison of the contemporary design and the lagged design (base model) would help examine the sensitivity to the potential endogeneity or simultaneity issues (e.g., the convergence of parents and their adult children due to the increasing care needs of the parents).

RESULTS

Summary statistics of spatial proximity to children, by parents’ and children’s characteristics

We summarized the spatial proximity by the sociodemographic characteristics of older adults and adult children in our study sample in Tables 1 and 2, respectively. About seventy percent of older adults with disability have at least one adult child either coresiding (25.9%, Table 1) or living within 10 miles/the same ZIP Code area (43.1%). Almost thirty percent of parents have all (i.e., farthest) adult children either coresiding (6.6%) or nearby (within 10 miles/the same ZIP Code area; 22.1%). The greater share of females (vs. males) have at least one coresident adult child (28.7% vs. 21.9%), as well as, have all adult children coresident (7.4% vs. 5.5%) or within 10 miles/same ZIP Code (23.9% vs. 19.5%). Compared to other racial and ethnic groups, non-Hispanic Whites are least likely to have any coresident adult child (20.7% vs. 32.8–41.8%); and are more likely to have the farthest child living 100 miles away (44.1% vs. 33.5–39.0%, Table 1). The P-values are all from Chi-square tests and F tests, if specified, and are statistically significant.

Table 1.

Older adults’ spatial proximity to the closest and farthest adult child, by older adults’ characteristics (Sample N: 18,429 older adult-year observations)

	N	Older adults without any adult childa	Closest Adult Child Spatial Proximity					P b	Farthest Adult Child Spatial Proximity					P b
			Coresidence	Within 10 miles or same ZIP Code	Residential distances (miles)				Coresidence	Within 10 miles or same ZIP Code	Residential distances (miles)
					<=30	30–100	>100				<=30	30–100	>100
Overall	18,429	3.3	25.9	43.1	12.0	5.4	10.3	-	6.6	22.1	15.5	11.0	41.5	-
Age group, %								<0.001						<0.001
55–64 years	5,355	3.2	32.5	37.1	11.5	5.6	10.1		9.6	24.9	17.0	10.8	34.5
65–74 years	4,216	2.1	22.9	44.9	13.2	5.7	11.3		5.2	23.2	16.0	11.8	41.7
75–84 years	5,398	2.8	21.8	48.0	11.8	5.4	10.2		4.6	18.9	15.0	11.1	47.5
85+ years	3,460	6.4	22.8	46.0	11.1	4.7	9.1		5.7	18.8	12.2	9.9	47.1
Gender, %								<0.001						<0.001
Male	6,949	3.3	21.9	43.1	13.3	6.9	11.6		5.5	19.5	15.0	12.2	44.5
Female	11,480	3.3	28.7	43.2	11.0	4.4	9.4		7.4	23.9	15.9	10.2	39.4
Race/Ethnicity, %								<0.001						<0.001
Non-Hispanic White	10,621	3.0	20.7	45.5	12.7	6.5	11.6		5.2	20.4	14.5	12.7	44.1
Non-Hispanic Black	4,126	4.5	32.8	40.5	10.5	3.8	7.9		7.7	25.1	17.1	8.7	36.8
Non-Hispanic Others	554	1.1	37.9	36.1	8.5	4.1	12.3		15.2	20.7	15.4	8.6	39.0
Hispanic	3,101	4.0	41.8	35.6	10.7	2.2	5.7		10.2	27.9	19.2	5.3	33.5
Foreign born status, %								<0.001						<0.001
No	15,559	3.1	23.6	44.7	11.9	5.9	10.7		5.9	21.4	15.0	11.6	43.0
Yes	2,859	4.4	41.3	32.3	12.2	2.4	7.5		11.6	26.6	19.3	7.2	31.0
Marital status, %								<0.001						<0.001
Married/Partnered	9,360	2.2	22.8	44.0	13.7	6.6	10.7		5.1	19.0	16.2	12.2	45.4
Separated/Divorced/Never Married	3,438	4.2	28.5	41.4	10.9	4.3	10.8		10.7	31.0	15.9	9.0	29.3
Widowed	5,617	5.0	30.6	42.6	8.9	3.7	9.1		6.9	22.0	13.6	10.1	42.3
Education years, %								<0.001						<0.001
<12 years	6,170	4.5	33.6	43.0	10.6	2.9	5.3		7.4	25.6	18.2	10.1	34.2
12 years	5,866	2.3	23.3	48.1	11.5	6.0	8.7		6.1	23.9	16.0	12.0	39.7
13–15 years	3,831	2.8	23.9	41.1	13.2	6.3	12.8		6.5	20.3	13.6	11.2	45.6
16+ years	2,508	3.9	20.2	36.4	13.5	7.5	18.5		6.5	14.5	12.6	10.6	51.9
Number of ADLs/IADLs
Mean (SD)	18,427	3.65 (2.86)	3.06 (2.72)	2.75 (2.28)	2.62 (2.20)	2.40 (1.95)	2.61 (2.17)	<0.001	3.25 (2.75)	3.04 (2.46)	2.80 (2.44)	2.57 (2.08)	2.62 (2.27)	<0.001
Median (IQR)	18,427	3 (4)	2 (3)	2 (3)	2 (2)	1 (2)	2 (2)	<0.001	2 (4)	2 (3)	2 (3)	2 (2)	2 (2)	<0.001

Note. All older adults who were 55 years or older in 2010, 2012, 2014, 2016, and 2018 and had disability are included, regardless of having any adult child. For older adults with only one adult child, their proximity to the child is presented in both sections of the table about the closest and farthest adult child. The sampling weights and the complex survey design were incorporated to adjust for standard errors.

^aOlder adults had minor children (i.e., child<18 years of age) only or no living child.

^bF test P-values for mean values and Chi-square test P-values for all the others

Table 2.

Spatial proximity of parent- adult child dyads by adult children’s characteristics (Sample N: 51,582 dyad-year observations)

	N	Parent-adult child Dyad Spatial Proximity					P a
		Coresidence	Within 10 miles or same ZIP Code	Residential distances (miles)
				<=30	30–100	>100
Overall	51,582	11.8	34.1	18.1	10.0	26.0	<0.001
Adult child’s age, mean (SD)	51,582	39.0 (14.2)	45.4 (11.2)	44.9 (11.1)	46.0 (10.9)	46.8 (11.3)	<0.001
Adult child’s gender, %							<0.001
Male	25,006	12.6	31.9	17.8	10.4	27.2
Female	26,576	11.0	36.2	18.4	9.7	24.8
Adult child’s relationship to parent, %							<0.001
Biological/adopted children	43,146	13.2	35.5	17.3	9.4	24.5
Stepchildren	8,141	4.1	27.2	22.1	12.9	33.8
Children-in-law or others	281	10.1	21.4	26.0	16.6	25.8
Adult child’s married/partnered, %							<0.001
No	18,736	25.1	29.1	15.4	8.7	21.8
Yes	32,832	4.7	36.8	19.6	10.7	28.2
Adult child’s education years, %							<0.001
<12 years	6,365	16.0	37.6	19.7	9.8	16.9
12 years	19,168	11.4	37.6	19.1	9.7	22.2
13–15 years	11,940	13.5	35.4	17.7	9.2	24.2
16+ years	12,101	7.7	26.3	16.0	11.8	38.2
Adult child’s full-time employed, %							<0.001
No	18,113	19.5	33.0	17.0	9.3	21.3
Yes	30,571	7.9	35.4	18.3	10.3	28.1

Note. The sampling weights and the complex survey design were incorporated to adjust for standard errors.

^aF test P-values for mean values and Chi-square test P-values for all the others

Adult children living in the same house as their parents are relatively younger than those not coresiding (mean age: 39.0 vs. 44.9–46.8 years, Table 2). Compared to step-children, the greater share of biological/adopted children live with the older adult (13.2% vs. 4.1%) or nearby (35.5% vs. 27.2% for ‘within 10 miles/same ZIP Code’). Adult children are less likely to coreside with parents if married/partnered vs. otherwise (4.7% vs. 25.1%) and full-time employed vs. not (7.9% vs. 19.5%). However, more than one-third of adult children who are married (36.8%) and employed full-time (35.4%) live within 10 miles/same ZIP Code area. Also, adult children with fewer years of schooling tend to live close to their parents; more than half of adult children with education less than 12 years (53.6%) live either in the same house or nearby (within 10 miles/same ZIP Code), while about 34% of those with 16 years of education or more do (Table 2). The P-values are from Chi-square and F tests and are statistically significant.

Child Spatial Availability for Care (CSAC) and Child Spatial Proximity (CSP), by Spatial Proximity Category

Table 3 summarizes the estimated dyad-level CSAC scores (from the base and interaction models) and dyad-level CSP scores, corresponding to each spatial proximity category. The dyad-level CSAC score primarily highlights the importance of the proximity of living in the same residence, which is the main contributing factor for greater child spatial availability for caregiving. There is a drastic difference in the dyad-level CSAC score between coresidence and all other proximity categories. However, there is only a small or no variation in the dyad-level CSAC score over the proximity categories beyond coresidence: dyad-level CSAC score=16.5 for a coresident adult child vs. 1.5–3.7 for all the others (Base model in Table 3). Results from the interaction model suggest the dyad-level CSAC score decreases as the number of children increases for an adult child coresident or living nearby: e.g., dyad-level CSAC score of coresident child=31.3 for parents with one adult child, 21.7 for parents with two adult children, and 16.3 for parents with three or more adult children (Interaction Model in Table 3).

Table 3.

Summary of parent-adult child dyad-level CSAC and CSP scores by spatial proximity (Sample N: 51,582 dyad-year observations)

Spatial Proximity	CSAC				CSP
	base model	Interaction Model: a base model with dyad spatial proximity interacted with parents’ number of children
		One child	Two children	Three or more children
Coresidence	16.5	31.3	21.7	16.3	6.3
Within 10 miles/ same ZIP Code	3.7	7.9	4.4	3.9	6.1
<=30	2.6	3.6	3.1	2.9	6.0
>30 & <=100	2.5	1.9	1.5	3.0	5.2
>100 & <=200	1.6	4.4	2.1	1.8	3.9
>200 & <=500	1.8	-	-	-	2.1
>500 & <=1,000	1.5				0.53
>1,000	1.8				0.03

Note. The dyad-level CSP scores represent the scores at the distance of 0 miles for coresidence; 10 miles for the category of within 10 miles/same ZIP Code, and the median distance within each of the other categories.

^aThe dyad spatial proximity categories beyond 100 miles are combined to have enough sample size.

In contrast to the pattern of the dyad-level CSAC, the dyad-level CSP score decreases gradually associated with farther distances (or less proximate categories): CSP score=6.3 for coresidence; 6.1 for ‘within 10 miles/same ZIP Code’; 6.0 for ‘<=30 miles’; and 3.9 or below for any dyad residential distances more than 100 miles (CSP in Table 3). Given that the CSP score was built on continuous distances, the values of CSP presented in Table 3 are the dyad-level scores at the distance of 0 miles for coresidence, 10 miles for ‘within 10 miles/same ZIP Code’, and 1000 miles for ‘>1,000 miles’, as well as, at the median distance within each of the other spatial proximity categories.

Supplemental Figure 2 shows the distributions of the final CSAC (base model) and CSP indices which are summary indices for each older adult with disability (as opposed to parent-child dyad-level scores). Both indices have right-skewed distributions with median values of 7 in CSAC (base model) and 12 in CSP. To better interpret the indices, we provided the index value corresponding to each hypothesized scenario in Figure 1. It demonstrates the values of CSAC (from the base model and the interaction model) and CSP indices related to the number of adult children and proximity composition for hypothetical older adults. Overall, the CSAC index is largely driven by the presence of coresident adult children. For example, older adult A with one coresident adult child may have a CSAC index (16.5, Figure 1) similar to older adult J with five adult children living within 10 miles or in the same ZIP Code area from the base model (18.5, Figure 1). The CSAC index is comparable between the models (base and interaction) for older adults with three or more children or far distances (greater than 10 miles and outside the ZIP Code area); however, for older adults with one or two children who coresident or living nearby, CSAC index using the interaction model is greater than CSAC index using the base model (older adults A, B, F, G in Figure 1). Overall, CSP index is relatively lower for older adults with coresident adult children (older adults A, B, C, D, E in Figure 1) but higher for older adults with adult children non-coresident and living far (e.g., older adults G, H, I, J, M in Figure 1), compared to CSAC indices.

Figure 1. CSAC and CSP indices by hypothesized scenarios (Sample N: 18,429 older adult-year observations).

Note. The CSAC and CSP indices are the sum of child-level CSAC and CSP scores respectively. Both CSAC and CSP indices range from 0 to 100, with a higher value indicating greater spatial availability.

Variation in Child Spatial Availability for Care (CSAC) Score and Index by Older Adults’ Characteristics

Figure 2 presents the median values of the older adult-level CSAC index (base model) by older adults’ characteristics. The levels of CSAC index differ greatly by older adults’ race and ethnicity, foreign-born status, and years of schooling. Non-Hispanic Whites (7.3, 95%CI= [7.2, 7.5]) have less child spatial availability than non-Hispanic Blacks (11.0, 95%CI= [10.3, 11.8]) and Hispanics (13.4, 95%CI= [11.6, 15.1]). Older adults born in the U.S. (7.3, 95%CI= [7.2, 7.5]) have less child spatial availability than their peers born outside the U.S. (11.1, 95%CI= [9.5, 12.7]). Fewer than 12 years of schooling are associated with more child spatial availability than 16 years or more schooling (11.0, 95%CI= [10.4, 11.6] vs. 6.2, 95%CI= [5.8, 6.5]).

Figure 2. Median values of CSAC index (base model) by older adults’ characteristics (Sample N: 18,429 older adult-year observations).

Note. The CSAC index is the sum of dyad-level CSAC scores. Zero index values were assigned to the parents with only minor children or no child alive. The data labels represent the estimated median values of CSAC and the 95% confidence intervals around them (in parentheses). All the estimates were sampling-weighted, and the standard errors of the confidence intervals accounted for the complex survey design.

The sociodemographic patterns of child spatial availability based on CSP were mostly consistent, but the CSP index shows more drastic differences than CSAC for some subgroups: between ages 85+ vs. others (9.8 vs. 11.7–12.1); either separated, divorced, or never married vs. married or partnered (8.2 vs. 12.1); and 13–15 years of schooling vs. 12 years (10.0 vs. 11.9, Supplementary Figure 3).

Results of CSAC (base model) from Sensitivity Analyses

All the results of the CSAC dyad-level scores from the sensitivity analysis models (Supplementary Tables 2–6) are similar to those from the base model with some additional findings. In the stratified analysis by parents’ spousal disability status, coresident adult children may have a smaller contribution to child spatial availability for parents having a spouse without disability than for those having no spouse and having a spouse with disability (CSAC score=14.9 vs. 16.6–16.7, Supplemental Table 2), but the non-coresident adult children living nearby or less than 30 miles may have a higher contribution (CSAC score for ‘within 10 miles/same ZIP Code’=5.1 vs. 3.4–4.0; CSAC score for ‘<=30’=5.3 vs. 1.2–2.9).

In the stratified analysis by parents’ age, we observed that the CSAC score seems lower for parents >=85 years than those younger for a coresident dyad (15.3 vs. 16.7–17.5, Supplementary Table 3). In the contemporary model (vs. the lagged design base model), the decline in the dyad-level CSAC score between coresidence and nearby proximity is more drastic (18.4 for coresident dyad vs. 0.05–1.8 for non-coresident dyad, Supplementary Table 5). The differences in the results from the contemporary and the lagged models may be attributable to the endogenity or reverse relationship in the estimation model which is likely to be more pronounced with the contemporary design (e.g., geographic convergence between parents and their adult children due to increasing parents’ needs of care).

DISCUSSION

We developed holistic indices – CSAC and CSP – to quantify the spatial availability of all adult children of the U.S. adults aged 55 and older living with disability. The two indices have distinct focuses on the spatial availability of adult children. The CSAC index, our primary measure incorporating the level of caregiving, highlights the importance of having at least one adult child living in the same house, which contributes to substantially higher overall child spatial availability to care for older adults with disability. In other words, the number of adult children and spatial proximity beyond coresidence have relatively low or little contributions to the overall adult child spatial availability to provide ADL/IADL care for older adults with disability. For example, the CSAC score is 16.5 for coresident, which decreased to 3.7 for the nearby proximity (within 10 miles/the same ZIP Code) and to 2.5 for 30–100 miles. There is no decreasing trend over the spatial distance beyond 100 miles (CSAC score range of 1.5–1.8). The finding is consistent with the prior evidence suggesting that ADL/IADL care support is strongly contingent on very close proximity, such as coresidence or less than 5 miles (Schoeni et al., 2022). Furthermore, we observed a diminishing marginal utility or productivity of adult children in the CSAC score for a given proximity level, which was mainly pronounced for a coresident dyad of parent and adult child; the value of the CSAC score for a coresident dyad was 31 if the older adult had one adult child but 16 if the older adult had three or more adult children. There may be one adult child playing a primary role in providing assistance with daily activities (ADLs and IADLs), and an additional adult child may be involved relatively less in the care activities for an older adult with disability. Or multiple adult children may share the care amount to meet the care needs of the older adult with disability, resulting in smaller care amount provided by each adult child if the older adult has a greater number of adult children, given the total amount of care needs.

Compared to the CSAC, the CSP index (without incorporating potential care received from adult children associated with distance) suggests that both the spatial proximity composition of adult children and the number of adult children are important to the overall child spatial availability for older adults with disability. While CSP index is valuable in summarizing spatial dispersion of adult children, it may not depict well the child spatial availability to care for people with disability as CSP tends to underestimate the contribution of coresident adult children and overestimate the contribution of non-coresident nearby or moderate distance (e.g., <200 miles).

Both the CSAC and CSP indices vary substantially depending on older adults’ sociodemographic status. Our findings on substantially greater spatial availability among racial/ethnic minority populations and those with low education are consistent with the prior evidence on population differences in intergenerational spatial proximity (Choi et al., 2020; Reyes et al., 2020; Choi, Schoeni, et al., 2021). Racial/ethnic minorities and those with lower education have relatively limited financial resources, which might increase the incentive for informal care support. We found that child spatial proximity (CSP index) for older adults aged 85+ is lower compared to other age groups (55–84); this may be partly attributable to a relatively higher rate of older adults 85+ living in an assisted living or nursing facility (in our analysis sample, 4.7% for older adults aged 85+ vs. 0.4% for older adults aged 55–84), likely to be far away from adult children’s residence (e.g., greater than 10 miles). We also found that a higher percentage of older adults aged 85+, compared to the other age groups, have no adult child (6.4% vs. 2.1%-3.2%, Table 1).

There are some limitations in this study. First, the measure of dyad spatial proximity lacks granularity; hence we could not discern adult children living within very close spatial proximity (e.g., on the same block, less than 1 mile). Prior evidence suggests the level of ADL/IADL care received by a parent varied primarily over the dyad residential proximity level within a very close distance level such as 0–5 miles (Schoeni et al., 2022). Thus, the spatial proximity category of within 10 miles or in the same ZIP Code area may be too coarse to distinguish the important variation of child spatial availability at a very close distance. Also, the measure of dyad residential distances may be subject to reporting errors as older adults self-reported the residential locations of all their children (in the form of ZIP Code, city, town). Second, the high level of reported ADL/IADL care received by older adults from the coresident adult children may not necessarily reflect the care hours for the activities responsive to older adults’ needs only; it may also include the time spent on activities that need to be done routinely as they live in the same house (such as grocery shopping) but may have been viewed as part of ADL/IADL care by the sample person (i.e., older adults). The potential measurement error related to this is likely less with ADL care (e.g., walking across rooms, bathing) but more with IADL care (e.g., grocery shopping, managing bills). However, the results from our sensitivity analysis suggests that the levels and proximity patterns of CSAC scores using IADL care hours are similar to ADL care hours. Third, although a lagged analysis approach is adopted for constructing CSAC to reduce reverse causation of dyad spatial proximity and parents’ care needs, it may not fully address the endogeneity associated with geographic convergence prior to the study periods, considering the lack of detailed history of coresidence for each dyad of parent and adult child. Fourth, older adults with minor children (under 18 years old) only and people without any children are equally regarded as having zero child spatial availability, which may be a limitation as some minor children – especially just under 18 - may contribute to caring for parents with a disability. However, due to the small sample size (e.g., only 0.96% children aged 15–17), we are unable to test the extent to which they provide ADL/IADL care associated with their spatial distances.

The study may contribute to the literature about intergenerational caregiving for mid- to late-life older adults with ADL/IADL disability. It advances the understanding of family care resources for the aging population by developing holistic indices of child spatial proximity and availability. The CSAC and CSP indices help summarize the potential care availability that an older adult with disability has by accounting for all adult children living at different distances regardless of the type of relationship to the older adult. Moreover, using a nationally representative sample of older adults, the CSAC index based on sample-weighted model coefficients may be generalizable to the U.S. population aged 55 and older with at least one ADL or IADL limitation. While the CSP index summarizes the spatial availability of children in a more general perspective associated with spatial distance, the CSAC index gives important insight into the potential family care resources available to older adults with any disability, particularly focusing on ADL or IADL care availability. The additional care-weighted feature of the CSAC index may inform health policy which assesses care resources to meet specific care demands for older adults with a disability. This understanding of family care resources (such as child availability) for older adults is critically important in developing effective and sustainable long-term care programs because it has profound implications for health care use (e.g., nursing facility) and associated health care costs (Artamonova et al., 2023; Charles & Sevak, 2005; Choi et al., 2015a; Choi, Heisler, et al., 2021; Friedman et al., 2019; Van Der Pers et al., 2015). However, future investigation is warranted to enhance the utility of indices by extending those with different types of support (e.g., emotional and financial) or subpopulations (e.g., people living with dementia). It is also important to assess the extent to which the index of spatial availability of children predicts the paid and unpaid health care used by older adults and associated care costs.