Competing Risk Analysis of Time to Communal Residence for Elder Orphans

Abstract

Elder Orphans, socially/physically isolated older adults without caregiving support, are of interest in an aging population. Lack of caregivers for Elder Orphans may influence relocation to residential care facilities, including skilled nursing or assisted living facilities, compared to ageing in place. Using the National Health and Aging Trends Study (NHATS), Competing Risk Survival Analyses were performed to determine if Elder Orphans or those At Risk for becoming elder orphans had increased risk for residential care relocation over nine NHATS waves (2011–2019). Elder Orphans had significantly higher risk for moving to residential care facilities in unadjusted (SHR=1.780; p=0.001) and adjusted (SHR=1.571; p=0.043) models. Those At Risk for becoming an elder orphan had significantly decreased risk for residential care residence in unadjusted (SHR=0.517; p<0.001) and adjusted (SHR=0.726; p=0.037) models. As aging in place becomes prioritized in the US healthcare system, understanding caregiving needs of older adults is vital to their health outcomes.

Introduction

As the population of older adults increases in the United States (US) and worldwide, the concept of an elder orphan has been used to describe “aged, community-dwelling individuals who are socially and/or physically isolated, without an available known family member or designated surrogate or caregiver” (Carney et al., 2016). Although older adults may experience some of the factors for qualification as an elder orphan, including solo community residence, isolation, or lack of caregivers, the combination of these elements is an important distinction for status as an elder orphan because these individual factors may compound to cause increased risk of poor health outcomes or mortality for elder orphans. Previous studies have estimated the number of elder orphans living in the community in the US to be 2.62% of a nationally representative sample of older adults with Medicare, with an additional 21.29% of older adults determined to be At Risk for becoming elder orphans (i.e. fitting all the criteria for elder orphan, but still living with a spouse/partner) (Roofeh et al., 2020). Although some individuals who live in residential care facilities, including skilled nursing or assisted living facilities, may otherwise qualify as elder orphans, community residence is an important aspect in elder orphan categorization, as the services provided at residential care facilities may confound some of the main factors of elder orphanhood.

While dedicated study of health and social outcomes for elder orphans remains limited, examination of outcomes related to the contributing factors may be beneficial to assess outcomes for elder orphans. There is concern that older adults who experience social or physical isolation may face increased risk of mortality and poor health outcomes, including increased risk of skilled nursing facility residence (Chan et al., 2020; Djundeva et al., 2019; Luppa et al., 2010). Physical isolation in the form of homebound status has been associated with increased risk of mortality, even when factors like functional status and comorbidities are controlled for (Soones et al., 2017). There is also some evidence that social isolation and homebound status may combine to increase risk of overall mortality (Sakurai et al., 2019). Separate studies determined that restricted life spaces and reduction of life space mobility can increase risk of skilled nursing facility placement, even while controlling for the traditional factors associated with skilled nursing facility admission (Sheppard et al., 2013). In addition to physical isolation, previous literature has determined that increased emotional loneliness may decrease global cognitive function and predict all-cause mortality, particularly for older men (Boss et al., 2015; Holwerda et al., 2012; OʼSúilleabháin et al., 2019). A study of adults in Canada determined that older adults who felt low affection, emotional support, positive social interactions, or a weak sense of belonging had an increased mortality risk over a ten year period (Renwick et al., 2020). This increase in morbidity and mortality due to social isolation has been shown to increase Medicare spending by objectively isolated older adults through increases in hospital and skilled nursing facility costs (Shaw et al., 2017). Isolated older adults who also do not have close family to act as caregivers, often referred to as kinless older adults, have increased mortality hazard compared to older adults with spouses and/or children (Margolis & Verdery, 2017; Patterson et al., 2020). This population is at increased risk for dying in a skilled nursing facility, even though they have similar end of life symptom burden and treatment intensity (Plick et al., 2021).

While the literature characterizing elder orphan status continues to grow, as yet little work has examined health outcomes that prevent aging in place and necessitates admittance to residence in a residential care facility. Using the National Health and Aging Trends Study (NHATS) and previously categorized elder orphan status groups, this study aims to determine if elder orphan status is a risk factor for increased mortality or residential care facility residence.

Methods

Participants

The first nine Waves of the National Health and Aging Trends Study (NHATS), covering participant interviews from 2011–2019, were used for this analysis. The NHATS is a nationally representative sample of older adult Medicare recipients in the United States, with oversampling of Black older adults and the oldest old (age 85+ years) (Kasper & Freedman, 2020). Elder Orphan, At Risk for becoming an elder orphan, and Not an Elder Orphan categorization was determined based on NHATS participant status during Wave One.

Elder Orphan categorization was based on participant household composition, level of social and physical isolation, independent activity of daily living (IADL) and basic activity of daily living (ADL) assistance needs, and unpaid caregiver availability. Detailed methods for categorization of Elder Orphan, At Risk, and Not an Elder Orphan are described elsewhere (Roofeh et al., 2020). Of the original 8,245 individuals interviewed in the first Wave of the NHATS, 6,680 met criteria for inclusion for the elder orphan prevalence assessment. Previous analysis determined a 2.65% prevalence of Elder Orphans and a 21.29% prevalence of those At Risk for becoming elder orphans. The same 6,680 individuals were included for this analysis.

All participants included for categorization lived in the community and could participate in the NHATS interview independently, excluding those already living in skilled nursing or assisted living facilities and those requiring proxies to participate. Older adults living alone were assessed for Elder Orphan status, older adults living with only their spouse/partner were assessed for At Risk for becoming elder orphans, and thzose living with two or more household members were considered Not Elder Orphans (Table 1). Social isolation was measured by the participant’s reported number of social contacts, with 2 or fewer contacts (of up to 5 contacts) used as an indication of social isolation. Physical isolation was measured by the reported number of times per week the participant left their home for any reason, with 4 or fewer days per week used as an indication of physical isolation. Caregiving need was established if the participant required assistance with a minimum number of IADLs and ADLs for self-care and household function, including cooking, grocery shopping, laundry, banking, dressing, and personal hygiene, including washing and toileting. Caregiver availability was determined through the number of unpaid caregivers recorded, regardless of their relationship with the participant. Older adults without ADL assistance requirements were listed as not having caregiving needs, regardless of caregiver availability.

Table 1.

Definition of Elder Orphan status categories

	Elder Orphan Status
Category	Not an Elder Orphan	At Risk	Elder Orphan
Living Arrangement	• Spouse and 1+ others • 1+ others, non-spouse	Spouse only	Lives alone
Social Isolation
Social Network Size		≤ 2 individuals	≤ 2 individuals
Physical Isolation
Days Left House		≤ 4 days per week	≤ 4 days per week
Caregiving
IADL or ADL Assistance Needed	No	Yes	Yes
Caregiving Provided by	--	• Spouse • Paid Caregiver • Caregiving Not Received	• Paid Caregiver • Caregiving Not Received

IADL= Independent Activities of Daily Living; ADL= Basic Activities of Daily Living

IADL and ADL tasks include cooking, grocery shopping, laundry, banking, dressing, washing and toileting

Spouse is inclusive of spouse and unmarried partner

Elder Orphans lived alone in the community, had few social contacts, left home infrequently, and had caregiving needs but no unpaid caregiving support. Those At Risk for becoming elder orphans met all these criteria, with the exception that they continue to live with, and potentially receive care from, their spouse/partner.

Although the NHATS indicates residence in facilities where nursing care is available, residence in facilities with or without nursing care was analyzed together to account for the spectrum of services available in residential care facilities and the potential that available nursing services were not used by residents. Additionally, residence in residential care facilities with and without nursing services was combined because both types of facilities provide socialization and assistance with caregiving tasks that are included in the calculation to determine if an older adult has caregiving needs and/or receives caregiving. It was determined that once an older adult entered Residential Care, the receipt of these caregiving tasks and regular check-ins by staff would cause the older adult to fall out of the population being considered for elder orphan categorization.

Competing Risk Models

Data were weighted using Wave One sample weights to account for nonresponse bias (Montaquila et al., 2012). Demographic characteristics, including mean age, gender, race/ethnicity, number of comorbidities, and highest level of education were included. Number of common comorbidities was calculated for each participant using self-reported medical history. Mean years from Wave One interview to mortality and Wave One interview to moving to a residential care facility was calculated and separated by elder orphan category. Individual hazard ratios were calculated for potential confounding variables to determine which variables to include in the main survival models.

NHATS Wave One month and year interview date was used as the start date for both time to mortality and time to residential care facility move. Participants without a Wave One interview date were assigned June 2011, representing the midpoint of the interview season. Months from first interview to month and year of death was used for survival time to mortality. For participants with missing month of death but with year of death listed, a month of death was assigned based on year of death and previous interview dates. For participants with both month and year of death missing, a death month of January and associated year of NHATS Wave was assigned.

To determine time to residential care facility residence, months from Wave One interview to the first interview date where residential care facility residence was reported was used. Those who died in the community were censored at date of last interview with confirmed community residence (for example, a participant who had died by Wave 9 was censored on Wave 8 interview date).

As the factors that can lead to participant mortality or residential care facility residence are related, Competing Risk models were utilized for this analysis to account for any participant mortality prior to a move into a residential care facility residence. Because residential care facility residence must occur prior to mortality, the three Competing Risk Survival Analysis models were performed for time to residential care facility residence analysis, with mortality included as a competing risk. Elder Orphan and At Risk status were separated to determine risk for these groups individually. Model 1 included just Elder Orphan or At Risk status. Model 2 added participant age, gender, and race during Wave One and Model 3 added participant highest level of education.

Statistical analyses were conducted using STATA/ IC 15 software (StataCorp, College Station, Texas).

Results

Sociodemographic Characteristics

Table 2 shows the weighted demographic characteristics of the 6,680 NHATS participants included. For the overall sample, mean age was 75.13 years, 55.8% of participants were female, and 81.11% of participants were White, non-Hispanic race/ethnicity, and 78.6% of participants had at least a high school level education. The average number of comorbidities for the overall sample was 2.43 out of a possible 10. Previous analysis has determined that female gender, higher age, Black non-Hispanic and Hispanic race/ethnicity, worse health, higher education, and higher income are significantly associated with At Risk categorization and higher income is significantly associated with elder orphan categorization (Roofeh et al., 2020). For the overall sample, average time from Wave One interview to participant mortality was 4.48 years and average time from Wave One interview to residential care facility residence was 4.23 years.

Table 2.

Demographic characteristics of sample at Wave 1, categorized by elder orphan status

Weighted %	Total Population	Not Elder Orphans	At Risk	Elder Orphans
Age, mean years	75.1	75.6	73.7	77.2
Mean comorbidities	2.4	2.5	2.2	2.7
Gender, Female	55.8	64.2	31.3	55.6
Race / Ethnicity
White, non-Hispanic	81.1	79.5	86.6	73.0
Black, non-Hispanic	8.0	9.2	4.3	10.5
Hispanic	6.7	7.1	5.3	11.4
Other (Am. Indian, Asian, Hawaiian)	4.1	4.3	3.8	5.1
Highest Education Level
Less than High School	21.4	22.8	17.0	26.3
High School Graduate	27.1	27.3	25.9	33.2
Associate’s or Some College	26.6	26.5	27.2	22.8
Bachelor’s or Higher	24.9	23.4	30.0	17.7
Survival Time*, mean years	6.6	6.5	6.9	5.9
Community Residence Time†, mean years	4.3	4.3	4.0	4.2

Note: Values and percentages are weighted. Percentages sum to 100% with rounding error.

^*Of participants who did not move to a residential care facility over the study period

^†Of participants who moved to a residential care facility over the study period

Covariates

Individual hazard ratios were calculated for gender, age, race/ethnicity, and highest level of education to determine the variables to be included in the competing risk models (Table 3). Black non-Hispanic and Hispanic race/ethnicity decreased risk of residential care facility residence over the study period. Female gender and increasing age both increased risk of residential care facility residence.

Table 3.

Subhazard Ratios of individual variables

	SHR	95% CI	P Value
Age	1.008	1.007 – 1.009	<0.001
Gender (female)	1.688	1.383 – 2.060	<0.001
Race / Ethnicity
White, non-Hispanic		Reference
Black, non-Hispanic	0.640	0.491 – 0.834	0.001
Hispanic	0.543	0.309 – 0.953	0.033
Other (Am. Indian, Asian, Hawaiian)	0.781	0.383 – 1.590	0.496
Highest Education Level
Less than High School		Reference
High School Graduate	0.952	0.727 – 1.246	0.719
Associate’s or Some College	1.022	0.780 – 1.340	0.873
Bachelor’s or Higher	0.798	0.602 – 1.058	0.117

Bold values significant at the p=0.05 level

Competing Risk Models

Three Competing Risk Survival Analysis models were performed for time to residential care facility residence, with mortality as the competing risk for both elder orphans and those at risk for becoming elder orphans (Figure 1). Table 4 shows the three Competing Risk Survival Analysis models for the association between Elder Orphan or At Risk status and time to residential care facility residence. The unadjusted model (Model 1) shows a significant increased risk for Elder Orphan residential care facility residence (SHR=1.780; p=0.001), which remains significant (SHR=1.571; p=0.046) in Model 2 when age, gender, and race/ethnicity are added as covariates, and in Model 3 (SHR=1.587, p=0.043) when highest education is added. Table 4 also shows a decreased risk of residential care facility residence for those At Risk for becoming elder orphans. In the unadjusted model (Model 1) those At Risk for becoming elder orphans are at decreased risk for residential care facility residence (SHR=0.517, p<0.001). This decreased risk remains significant (SHR=0.730, p=0.039) in Model 2 with the addition of age, gender, and race/ethnicity as covariates. In Model 3, At Risk status remains significant (SHR=0.726, p=0.037) when highest level of education is included in the analysis.

Figure 1.

Cumulative incidence of time to residential care facility residence for Elder Orphans and those At Risk for becoming elder orphans, as compared to those who are not elder orphans. Participant mortality prior to residential care facility residence is the competing risk.

Table 4.

Competing Risk Survival Analysis models for time to Residential Care Facility residence. Participant mortality prior to Residential Care Facility residence is competing risk

	Sub-Hazard Ratio (95% Confidence Interval)
Variable	Model 1	Model 2	Model 3
Elder Orphan	1.780 (1.146 – 2.765)	1.571 (1.008 – 2.450)	1.587 (1.016 – 2.479)
At Risk	0.517 (0.390 – 0.684)	0.729 (0.540 – 0.985)	0.726 (0.538 – 0.981)
Age		1.007 (1.006 – 1.008)	1.007 (1.006 – 1.008)
Gender (female)		1.420 (1.154 – 1.748)	1.436 (1.161 – 1.777)
Race / Ethnicity
White, non-Hispanic		Reference	Reference
Black, non-Hispanic		0.620 (0.475 – 0.809)	0.672 (0.510 – 0.886)
Hispanic		0.527 (0.301 – 0.923)	0.598 (0.342 – 1.048)
Other (Am. Indian, Asian, Hawaiian)		0.901 (0.444 – 1.831)	0.966 (0.476 – 1.963)
Highest Education
Less than High School			Reference
High School Graduate			1.162 (0.883 – 1.529)
Associate’s or Some College			1.436 (1.091 – 1.890)
Bachelor’s or Higher			1.364 (1.018 – 1.827)

Bold values significant at the p=0.05 level

Discussion

Summary of Main Findings

The purpose of this study was to determine if Elder Orphans and those At Risk for becoming an elder orphan were at increased risk of residential care facility residence, as compared to those who are not elder orphans. Using the NHATS and previously categorized elder orphan groupings, we determined that Elder Orphans were at increased risk for residential care facility residence in unadjusted models, and this remained significant in the full analysis when additional covariates were introduced into the model. Analyses also determined that those At Risk for becoming an elder orphan had decreased risk of residential care facility residence in unadjusted models.

The results of this study are in line with current research surrounding those who are aging alone and experience isolation. Previous studies have noted that older adults who have continuously lived alone are at higher risk for moving to a nursing home, while those living with a spouse are at lower risk in unadjusted models (Kasper et al., 2010). Interestingly, these studies also note that those at lowest risk for nursing home residence are those who consistently live with an adult child, or an older adult who would not classify as an elder orphan. There is also indication that living alone is associated with higher risk of unplanned hospitalization (Pimouguet et al., 2017). This may serve as an entry point for residential care facility admission as the hospitals organize a safe discharge plan. Presumably these admissions are related to health conditions that cannot be managed by the older adult alone and the receipt of appropriate care in the facility is a measure to improve or extend the older adult’s life.

Older adults who are At Risk for elder orphanhood are at reduced risk for residential care facility admission as compared to those who are not elder orphans and potentially have multiple friends or family members to act as caregivers. Table 2 indicates that among the three categories, those At Risk for becoming elder orphans have the shortest community residence time prior to residential care facility admission and the longest survival time in the community if they are not admitted to a skilled nursing or assisted living facility. This suggests that if those At Risk for becoming elder orphans required residential care facility residence, it was identified and occurred earlier in the study period and those who had spouse/partners who could provide caregiving were able to remain safely in the community. Because it is known that those At Risk for becoming elder orphans have caregiving needs that are fulfilled by their spouse/partner, it can be presumed that the spousal caregiving provided was an important factor for the participant’s ability to remain in the community. Previous studies have shown that more than half of spousal caregivers are providing this care alone, without the assistance of other friends or family (Ornstein et al., 2019). Particularly for those At Risk of becoming elder orphans, these spouse pairs may assume increasing caregiving responsibilities or underreport need for assistance to avoid residential care facility admission and the emotional impact of involuntary separation (Glasier & Arbeau, 2019).

Although aging in place is often the preferred setting for older adults, this study indicates that those who do not have caregivers available in the community are more likely to be unable to age in place. A previous longitudinal study found similar results using the Melbourne Longitudinal Studies on Healthy Ageing. In this study, most older adults interviewed intended to age in place, but over the course of the 16-year follow up, those who were no longer living with a spouse or those who were never married were significantly less likely to remain in their home (Kendig et al., 2017). An analysis using the Health and Retirement Study brings an additional layer of nuance an older adult’s intention to move to a skilled nursing facility and the caregivers available to them in the community. This study determined that older married men had less intention of eventually moving to a skilled nursing facility, likely due to their ability to rely on spousal caregiving, as compared to men with other marital statuses who had higher intention of moving to a skilled nursing facility, likely related to limitations in available caregivers (Lu et al., 2020). The same did not hold for older women, both married and with other marital statuses, whose moving intention was not associated with moving behavior. Overall, these studies are in line with the results of this analysis in that older adults without caregivers in the community are less likely to be able to age in place, while married older adults, particularly married men, are able to rely on spousal caregiving to remain in the community.

Limitations

Elder Orphan and At Risk grouping is based on the life status of the older adult during Wave One and kept through the duration of the analysis. This does not consider changes in status that may have occurred over the course of the follow up period. These changes, including the death of a spouse or the presentation of a caregiver, could change both elder orphan status and the health trajectory of the older adult. Although this information is available in subsequent waves of the NHATS survey, they would represent a time varying covariate. As previous studies advise against the use of time varying covariates in competing risk models, the decision was made to forgo including changes in elder orphan status as a time varying covariate in order to perform the competing risk analysis of residential care facility residence with mortality as the competing risk (Austin et al., 2020; Poguntke et al., 2018). While out of scope for this paper, the trajectory to elder orphanhood can have differing profiles, and could affect risk of residential care facility placement over time. Further studies to examine these circumstances are warranted.

Another important limitation is the decision to assess time to residence in a residential care facility with and without nursing care available together. While the experience of residence in facilities with and without nursing care can be different for older adults, both types of facilities provide assistance with caregiving tasks that are included in the calculation to determine if an older adult has caregiving needs and/or receives caregiving. Further studies to assess the contributing factors for residence in a facility with or without nursing care available may be valuable.

A third limitation is based in the previous categorization of elder orphans, particularly for ADL status and the effects of income. The definition of elder orphans in this study relies on a previously established categorization of elder orphan status. Those who did not need ADL assistance and did not have caregivers were categorized separately from those who needed ADL assistance and did not have caregivers. This separation could lead to discrepancies as those who were not categorized as elder orphans begin to need assistance with ADLs. Additionally, friends and families who were paid for caregiving tasks were categorized as paid caregiver, regardless of whether they would have offered unpaid caregiving if needed. Finally, the influence of income may warrant further consideration in any eventual move to a residential care facility. Traditionally, higher education level tracks with higher income level, which may guide whether an older adult utilizes paid caregiving resources or voluntarily resides in a residential care facility. The increased risk of residential care facility residence for older adults with a college education or higher align with evidence that more years of education in men is associated with increased intention to move to a skilled nursing facility (Lu et al., 2020). Further studies to clarify the relationship between income, paid and unpaid caregiving, and residential care facility residence would be beneficial.

Contribution to the Literature

While the literature is rich in examples of the impact of informal caregiving on those who give and receive care, there is a lack of focused study on the outcomes of those who may need caregiving but do not have friends or family resources to provide it. Studies examining outcomes for elder orphans adds to this gap in the literature by providing context and outcomes for older adults who definitionally do not have friends or family to act as caregivers or surrogates. This study determined that there is evidence of higher risk of residential care facility residence for elder orphans and decreased risk of residential care facility residence for those At Risk for becoming an elder orphan. Further studies are needed to investigate if this reduction in risk carries over to provide some protection upon the death of a spouse.

What This Paper Adds:

• This paper adds to a nascent literature studying Elder Orphans and their health and social outcomes through assessing their time to mortality or placement in a residential care facility.

Applications of Study Findings:

• The outcome that older adults living in the community who do not have caregivers are at increased risk for residential care facility placement can be applied to focus policies and resources for this population to begin preparations to age according to their wishes.

• The outcome that older adults living in the community who only receive caregiving from their spouse are at decreased risk for residential care facility placement can be applied to focus policies and resources on spousal caregiving relationships to maintain the status of both the caregiver and care recipient.