The Natural History of Disability and Caregiving Before and After Long-Term Care Entry

Key Points

Question  What is the natural history of disability and caregiving of persons entering long-term care facilities?

Findings  Prior to entry, almost half of nursing home (NH) entrants and one-tenth of assisted living (AL) entrants experienced a month or more of severe disability and almost all entrants had a home caregiver, but only one-third had a paid caregiver. Home caregiving was significant: the average NH and AL entrant received 27 and 18 hours of help per week; before they moved, respectively; within 2 years of entry, severe disability was present in two-thirds of AL residents, comparable to 88% in NH residents.

Meaning  This cohort study found that persons often enter NHs and ALs after months of severe disability and escalating help at home, usually from unpaid caregivers; AL residents move when less disabled, but approach levels of disability comparable to NH entrants within 2 years.

Importance  Many older persons move into long-term care facilities (LTCFs) due to disability and insufficient home caregiving options. However, the extent of disability and caregiving provided around the time of entry is unknown.

Objective  To quantitatively describe disability and caregiving before and after LTCF entry, comparing nursing home (NH), assisted living (AL), and independent living (IL) entrants.

Design, Setting, and Participants  A longitudinal cohort study using prospectively collected annual data from the National Health and Aging Trends Study from 2011 to 2020 including participants in the continental US. Overall, 932 community-dwelling Medicare beneficiaries entering LTCF from 2011 to 2019 were included. Entry into LTCF was set as t = 0, and participant interviews from 4 years before and 2 years after were used.

Main Outcomes and Measures  Prevalence of severe disability (severe difficulty or dependence in ≥3 activities of daily living), prevalence of caregivers, and median weekly caregiving hours per entrant, using weighted mixed-effects regression against time as linear spline.

Results  At entry, mean (SD) age was 84 (8.4) years, 609 (64%, all percentages survey weighted) were women, 143 (6%) were Black, 29 (3%) were Hispanic, 30 (4%) were other (other race and ethnicity included American Indian, Asian, Native Hawaiian, and other), and 497 (49%) had dementia. 349 (34%) entered NH, 426 (45%) entered AL, and 157 (21%) entered IL. Overall, NH and AL entry were preceded by months of severe disability and escalating caregiving. Before entry, 49% (95% CI, 29%-68%) of NH entrants and 10% (95% CI, 3%-24%) of AL entrants had severe disability. Most (>97%) had at least a caregiver, but only one-third (NH, 33%; 95% CI, 20%-50%; AL, 33%; 95% CI, 24%-44%) had a paid caregiver. Median care was 27 hours weekly (95% CI, 18-40) in NH entrants and 18 (95% CI, 14-24) in AL entrants. On NH and AL entry, severe disability rose to 89% (95% CI, 82%-94%) and 28% (95% CI, 16%-44%) on NH and AL entry and was 66% (95% CI, 55%-75%) 2 years after entry in AL residents. Few IL entrants (<2%) had severe disability and their median care remained less than 7 hours weekly before and after entry.

Conclusions  This study found that persons often enter NHs and ALs after months of severe disability and substantial help at home, usually from unpaid caregivers. Assisted living residents move when less disabled, but approach levels of disability similar to NH entrants within 2 years. Data may help clinicians understand when home supports approach a breaking point.

One-third of older US persons move into a long-term care facility (LTCF) during their lifetime.¹ Almost all would rather “age in place” (eg, stay in their own home if possible),^2,3 but worsening health and increasing disability can overwhelm care available at home.^4-7 As 1 caregiver of a recent LTCF entrant described trying to stay at home: “it was a continual downward turn… knowing he was struggling in the home environment… seeing him having trouble going to the bathroom.”⁸ Older persons may then move to obtain more care and reduce reliance on caregivers based on the availability and affordability of options.^9,10 Moving vs staying at home has clinical implications. Unmet needs in community-dwelling older adults are associated with increased health care spending, and the inability to live at home safely is a leading cause of emergency department visits among persons older than 65 years.^11-13

The move into an LTCF thus often represents a clinically relevant breaking point when disability and at-home supports become mismatched.¹⁴ Yet, we have little epidemiological data on disability and how much time family, friends, and hired caregivers spend trying to help older persons live at home before they cannot do it any longer.^15,16 Without such data, it is hard to grasp what is happening to function and home support to make aging in place untenable for so many.

Moreover, moving into an LTCF does not guarantee disability stabilizes or caregiving stops. For example, assisted living (AL) is traditionally seen as a setting for persons requiring only modest assistance, but studies suggest AL increasingly cares for persons similar to NH residents from decades ago, and AL often requests residents hire supplemental support if needs are high.^17,18 Whether this happens immediately on AL admission or develops over time is unknown. Independent living (IL) is also important to consider. Residents of IL facilities often move in anticipation of, rather than due to, functional decline, but this has not been objectively shown.

We therefore used data from a nationally representative study of Medicare beneficiaries to longitudinally examine function and caregiving before and after persons entered LTCFs, and compared findings by NH, AL, or IL entry.

We used the National Health and Aging Trends Study (NHATS). Started in 2011, the NHATS is a nationally representative cohort of Medicare beneficiaries in the continental US aged 65 years and older followed via in-person interviews with participants, their proxies, and/or facility staff.¹⁹ The cohort was refreshed in 2015. The NHATS interviews participants annually, prospectively gathering data on where participants live and their health, function, and social supports. The repeated interviews make it possible to observe trends in outcomes over time. The NHATS is sponsored by the National Institute on Aging (grant number NIA U01AG032947) through a cooperative agreement with the Johns Hopkins Bloomberg School of Public Health and approval from their institutional review board.

From the 2011 and 2015 cohorts of NHATS, we constructed an analytic cohort of persons who moved from a community-dwelling residence into an LTCF between 2011 and 2019 (Figure 1). We included entrants who were community-dwelling during an interview and then lived in an NH, AL, or IL in the next. Residence was determined as follows: interviewers asked staff of multiunit buildings if the place was an NH. Otherwise, ALs and ILs were multiunit buildings offering meals or assistance with activities of daily living. Some facilities offered multiple levels of care; we thus defined residence based on the specific area where the participant lived and services they used. Assisted living residents lived in assisted living areas or received assistance with medications, bathing, or dressing. Independent living residents lived in independent living areas and received no such assistance.²⁰

The LTCF entrants were asked the month and year they moved. We set this month as t = 0 and used the month and year of interviews to identify when interviews occurred relative to LTCF entry. For example, if someone entered AL in July of 2014 and had interviews in August of 2013, August of 2014, and July of 2015, those interviews supplied data for months −11, 1, and 12. This allowed longitudinal analyses per person before and after entry.

We included decedents if proxies reported the participant lived in a facility in their last month of life (eg, persons who died within the first year of entry) because decedents still provided data during the preentry period. We considered only the first transition into an LTCF, excluding uncommon subsequent transitions back to community living or between settings.²¹ The NHATS asked LTCF residents if they expected their stay would be temporary (eg, stays for postacute care, rehabilitation, respite). We excluded these stays and hospice stays.²² We excluded entry from 2020 onward due to the influence of the COVID-19 pandemic on preferences for LTCFs (eFigure 1 in Supplement 1).²³

Our primary outcomes of interest were (1) disability and (2) caregiving for each participant at each interview from 2011 to 2020.

We defined disability dichotomously as severely disabled or not. To match state thresholds for NH coverage,²⁴ we defined severe disability as assistance or severe difficulty in 3 or more activities of daily living (ADLs²⁵; getting out of bed, getting around inside, bathing, dressing, toileting, and eating). Unlike other national surveys, NHATS measures assistance (eg, “Do you get help?”) for ADLs rather than dependence (eg, “Do you need help?”). We interpreted assistance as disability because in the NHATS, ADL assistance is associated with markedly higher mortality (6 times the hazard compared with fully able persons).²⁶ We also considered severe difficulty in each ADL to be disability because not all persons who need help get help. Severe difficulty was defined by participants indicating they had a lot of difficulty when independently completing an ADL, allowing for assistive devices (eg, walkers for getting around). Persons who never completed the ADL also had severe disability.

We determined if participants had any caregiver, any paid caregiver, and the hours of help received from all caregivers weekly. In the NHATS, participants indicated assistance for instrumental ADLs (IADLS; getting around outside, transportation, doing laundry, getting groceries, preparing meals, banking, medications, making medical decisions, taking people to appointments) and ADLs in the past month and were asked who helped. Any helper was considered a caregiver, prompting additional questions about whether the caregiver was paid (by any source, including family and friends, and paid any amount), the hours per day they helped, and days they helped in the last month or week. We used these data to identify the prevalence of any caregiver, any paid caregiver, and care hours received per participant per week; participants with no helpers had 0 care hours. The NHATS does not gather data on caregiving from staff persons at the participants’ residence. Thus, caregiving after LTCF entry reflects only supplemental care from nonstaff and excludes hours provided by the facility.

The theory by Litwak and Longino⁹ on late-life migration posits that people move into facilities due to increasing disability from chronic conditions and insufficient family support. Bradley and Andersen¹⁰ additionally highlighted the importance of psychosocial factors such as attitudes and norms in mediating actual long-term care use. Baseline characteristics thus included demographics such as age, gender (male/female, investigator observed or by self-report), race and ethnicity (Black, Hispanic, White, and other [American Indian, Asian, Native Hawaiian, and other] by self-report), census region, metropolitan status, education, marital status, income (as quintiles),²⁷ Medicaid status, homeownership, and prior living arrangement. Health measures included self-rated health, possible/probable dementia (per validated algorithm),²⁸ a count of chronic conditions out of 6 (heart disease, arthritis, diabetes, lung disease, stroke, cancer), and physical performance (using the NHATS Short Physical Performance Battery; SPPB).²⁹ We disaggregated and also included impairments (severe difficulty or dependence) in each ADL as baseline characteristics.

We knew the year persons moved but not always the month; unknown months in 178 participants (19%) were singly imputed as we have done elsewhere. In 2319 of 11237 caregivers identified across all years (21%), we knew someone helped but not the number of care hours supplied. We used the NHATS method to impute how long caregivers likely spent per activity.³⁰ Baseline SPPB scores were missing in 155 participants (17%). All other missingness in outcomes and baseline characteristics was less than 3%.

We ran analyses from September 7, 2021, to October 7, 2023. We described baseline characteristics for the overall cohort and by NH, AL, and IL entry by reporting raw numbers and weighted percentages using survey weights from each participants’ enrollment year.³¹ We compared NH, AL, and IL entrants using weighted bivariate linear regression for continuous characteristics and χ² tests for categorical characteristics. We visualized the data in 6-month intervals (eg, grouping interviews with t ≥ −6 months and <0 months from LTCF entry) using bar graphs and box plots to plan our analysis (eFigure 2 and 3 in Supplement 1); the IL cohort was small and month-to-month variability obscured underlying trends. We thus used methods to longitudinally model prevalence of severe disability, prevalence of any caregiver and any paid caregivers, and the median weekly caregiving hours from all caregivers.

To smooth trends, we modeled time as a spline. Splines allowed us to describe temporal trends without making assumptions about the relationship between outcomes and time. Splines accommodated the possibility that outcomes increased, decreased, or stayed unchanged across several time intervals. Intervals were prespecified by “knots” —points in time when the trend could change. We additionally prespecified discontinuity (eg, allowed for a step change in prevalence) at t = 0 because other research has described how sudden health events can catalyze the transition into LTCF.^4,5,32,33 We used back-selection to choose our final spline model, starting with a complex cubic spline with 7 knots at yearly intervals and simplifying step-wise for parsimony, using likelihood ratio tests and visualizations to inform model selection. Our final model used linear splines with knots at t = −48, −12, and 0 months for all outcomes (eTable and eFigure 4 in Supplement 1).

To reduce the influence of outliers on weekly caregiving hours, we log-transformed continuous outcomes to normalize their distribution and reported back-transformed values (eFigure 5 in Supplement 1). Back-transformed values approximate a sample’s median, an approach often used for modeling other skewed data such as cost or length of stay.³⁴

To correct for repeated measures in individuals, we used linear and logistic mixed-effects regression, treating spline-time as a fixed effect and participants as a random effect, such that individuals had their own intercepts relating outcomes to spline-time. We used t tests to determine if trends within spline intervals were nonzero. Our visualizations and figures used predicted point estimates and 95% confidence intervals from the fixed-effects portion of these models.

All analyses were conducted using StataSE (version 16.1, StataCorp) and incorporated multilevel time-varying weights to account for the NHATS survey design, correcting for NHATS oversampling of older persons and Black non-Hispanic persons, sampling cluster effects, and survey attrition. The only exception was likelihood ratio tests comparing spline models, which excluded survey weights. We fit models on full data, truncating results to 4 years before entry and 2 years after entry to abide by the NHATS Data Use Agreement, which permits presenting results only for cells or strata with a number greater than 10. A 2-sided P value of .05 or less was statistically significant for analyses.

A dichotomous definition of severe disability may not be sensitive to changes in the severity of disability over time. We therefore explored the prevalence of mild disability (≥1 ADL disabilities) and how median function evolved on a validated 28-point scale.²⁶ We also examined median care hours supplied by unpaid caregivers to confirm if most care was unpaid. Our a priori decision to combine cohorts differed from NHATS technical guidance and assumed that function and caregiving are more correlated with LTCF entry than any demographic differences produced by imprecisely combining survey weights. To validate our analytic approach, we reran analyses (1) unweighted and (2) using only the 2011 NHATS cohort to explore the effects of survey design.¹⁹

The cohort included 932 new LTCF entrants, with 349 entering NH, 426 entering AL, and 157 entering IL. Participants supplied a total of 3237 interviews before entry and 2667 interviews after entry. The mean (SD) follow-up time was 36.9 (26.5) months before entry and 27.9 (23.7) months after entry. Weighted, mean (SD) age was 84 (8.4) years, 609 (64%, weighted) were women, 143 (6%) were Black, 29 (3%) were Hispanic, 30 (4%) had other race and ethnicity, and 497 (49%) had dementia (Table). Overall, IL entrants were younger, healthier, had less disability, and were more likely to live in an urban setting than NH and AL entrants. The AL and IL entrants were more likely to be White non-Hispanic, have more education, have higher incomes, and own their previous home than NH entrants. Overall, IL entrants were more likely to be married, and AL entrants were more likely to be widowed. The most common disabilities prior to entry were with dressing (326 [34%]), bathing (292 [27%]), and getting around inside (229 [23%]).

More than 4 years before entry and 2 years after, almost all IL entrants (≤1%) remained without severe disability. Almost all IL entrants (≥87%) had a caregiver, but prevalence of a paid caregiver stayed at 10% or less and median care hours stayed below 7 per week (Figure 2, Figure 3, and Figure 4 and eTables 6-7 in Supplement 1).

For AL and NH entrants, the prevalence of severe disability rose slowly over 4 years pre-LTCF entry and accelerated the year prior, with accompanying increases in caregiving hours (P<.05, Figure 2, Figure 3, Figure 4, and eTables 6-7 in Supplement 1). Disability and caregiving were greater for NH vs AL entrants. The modeled prevalence of severe disability was 49% (95% CI, 29%-68%) in NH entrants and 10% (95% CI, 3%-24%) in AL entrants before entry. Almost all (≥98%) had a caregiver, but only 33% (95% CI, 20%-50%) of NH entrants and 33% (95% CI, 24%-44%) of AL entrants received paid caregiving before entry. Median caregiving hours per week per entrant was 27 (95% CI, 18-40) among NH entrants and 18 (95% CI, 14-24) among AL entrants prior to entry.

Overall, NH and AL entry were marked by stepwise increases in the prevalence of severe disability and decreases in caregivers, paid caregivers, and nonfacility caregiving hours (P<.001, eTable 7 in Supplement 1). The prevalence of severe disability rose to 89% (95% CI, 82%-94%; P < .001) in new NH residents and 28% (95% CI, 16%-44%; P < .001) in new AL residents on entry. Although the prevalence of severe disability did not change after entry (P = .62) in NH residents, it rose (P <.001) to 66% (95% CI, 55%-75%) in AL residents after 2 years. Median weekly caregiving provided by nonfacility staff decreased to 10 hours (95% CI, 8-12) in NH residents and 8 hours (95% CI, 7-10) in AL residents on entry.

In sensitivity analyses, mild disability was present years before entry; 2 years prior, 32% (95% CI, 24%-41%) of NH entrants and 13% (95% CI, 9%-18%) of AL entrants had severe difficulty or assistance in at least 1 ADL. Median functional score similarly demonstrated increases before entry (eFigure 8 in Supplement 1). Care hours were supplied predominantly by unpaid caregivers except immediately prior to entry; median weekly caregiving hours from unpaid caregivers was similar to median total hours (eTable 9 in Supplement 1). Our base analysis produced similar or conservative estimates of function and caregiving over time vs unweighted analyses and analyses of the 2011 NHATS cohort only (eFigures 10 and 11 in Supplement 1).

In this study of the natural history of disability and caregiving around the time of LTCF entry in the US, we showed NH and AL entry were preceded by months of severe disability and escalating home care, most often by unpaid caregivers. Overall, NH entrants were more likely to be Black or Hispanic and have lower socioeconomic status, with higher levels of disability and more home care prior to entry compared with AL or IL entrants. Despite differences in disability at time of entry, the prevalence of severe disability in AL residents was 66% within 2 years of moving—similar to the 89% in NH residents.

Our findings are consistent with qualitative work describing NH as a last resort for family caregivers trying to avoid crises but feeling strained by increasing care needs and exhaustion.^8,33 The average NH entrant received 27 hours of caregiving per week from caregiver(s) before they moved. At a median cost of $27 per hour (and up to $36 per hour in some states), wholly relieving these hours with paid caregivers would have a market cost of $3134 per month.³⁵ This is not easily sustained without wealth: this figure exceeds the median income ($2200 per month) for persons aged 65 years and older, and only 11% of older persons carry long-term care insurance that might cover such services.³⁶ State Medicaid programs cover some paid home caregiving based on financial eligibility,³⁷ but we found only one-third of NH entrants had any paid caregiver. Thus, NH entry may not be what older persons want, but it happens when the alternative is another day at home potentially undressed, unbathed, and unable to get around, but for 27 hours of caregiving per week or more from unpaid persons.^16,38 Under these circumstances, new stressors—as suggested by the sudden rise in disability and decline in caregivers on entry—then catalyze NH entry.

Overall, AL entrants, who are more likely to be White and more affluent, can afford home and residential care options at earlier stages of disability. Still, they were not far from NH entrants in disability and caregiving. Notably, two-thirds of AL residents were severely disabled within 2 years of entry, underscoring similarities in the physical and mental health issues that arise in AL and NH populations and the calls for AL facilities to be accountable for health as NHs are.^18,39 The reduced preference for AL by Black communities and poor access to local high-quality LTCF options may additionally contribute to racial disparities in AL access.^40,41 Independent living residents—also disproportionately White—maintained low levels of disability, severe disability, and caregiving for years before and after moving.

When present, caregivers continued to be involved after older persons entered LTCFs. The average resident received approximately 8 to 10 hours of care a week from nonfacility staff, a reduction of 10 or more hours per week for continuing caregivers. Other research has reported similar findings; after moving, caregivers change roles to provide more managerial support (eg, attending medical appointments) instead of direct care.⁴²

Our study is applicable to the care of older persons in several ways. Patients and caregivers wish health care professionals provided better anticipatory guidance around disability and long-term care planning during later stages of life, with specific requests to explain when more care is needed and how life for entrants and caregivers changes with LTCF entry.^43-45 Health care professionals also admit to lacking evidence to make unbiased and evidence-based recommendations for care settings for older persons.^16,17,46 Our study provides this evidence in terms salient to both entrants (vis-a-vis function) and caregivers (vis-a-vis hours). We hope these findings help clinicians better understand “who lives where” to inform LTCF decision-making and estimate when home supports are critically strained.⁴⁷

Our illustration of how home- and facility-based care are exchanged may also inform policy. Recent efforts advocate for home- and community-based services (HCBS) to help older adults age in place. With only one-third of NH and AL entrants having any paid caregiver, programs may be underutilized. However, our data also estimate how much HCBS may be needed to facilitate aging in place with dignity. If HCBS programs cannot support 27 or more hours of care per week per person, we must acknowledge LTCFs provide substantial relief for caregivers and ensure NHs and ALs are staffed to support and maintain quality of life in severely disabled residents.⁴⁸

This study has several limitations. We assumed assistance with ADLs signified disability. We imputed one-fifth of caregivers’ hours. We could not determine the precise breakdown of paid vs unpaid hours; the NHATS only indicated if caregivers were paid or unpaid and not whether specific hours were paid or unpaid. The combination of the 2011 and 2015 cohorts in NHATS meant this study was not strictly nationally representative.³¹ Annual interviews cannot observe transient fluctuations in outcomes (eg, sudden rises in caregiving hours right before entry). Our analysis could not determine causality between LTCF entry and outcomes (eg, widowhood could precipitate LTCF entry), thus we did not investigate if specific health, social, or financial events precipitated LTCF entry though this may be a valuable future study.

Persons often enter NHs and ALs after months of severe disability and more than 27 and 18 hours of caregiving per week, respectively, usually from unpaid caregivers. Assisted living residents move when less disabled, but approach levels of disability comparable to NH residents within 2 years. Independent living entrants remain mostly independent of ADLs before and after entry. Data may help clinicians understand when home supports approach a breaking point and appraise long-term care options.

Accepted for Publication: August 23, 2023.

Published Online: November 6, 2023. doi:10.1001/jamainternmed.2023.5427

Corresponding Author: Kenneth Lam, MD, MAS, University of Colorado Anschutz Medical Campus, 12631 E 17th Avenue, L15-8101, Aurora, CO 80045 (kenneth.lam@cuanschutz.edu).

Author Contributions: Dr Lam had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Lam, Levy, Matlock, Covinsky.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Lam.

Critical review of the manuscript for important intellectual content: All authors.

Statistical analysis: Lam, Cenzer.

Obtained funding: Lam, Covinsky.

Administrative, technical, or material support: Levy, Covinsky.

Supervision: Levy, Matlock, Smith, Covinsky.

Conflict of Interest Disclosures: Dr Lam reported grants from National Institute on Aging (R03 AG074038 and P30 AG044281) and grants from National Center for Advancing Translational Sciences (KL2 TR001870) during the conduct of the study. Dr Covinsky reported grants from National Institute on Aging (P30 AG044281, P01 AG066605) during the conduct of the study. Dr Smith reported grants from National Institute on Aging K24 AG068312, P01 AG066605) during the conduct of the study. No other disclosures were reported.

Disclaimer: Dr Covinsky is an Associate Editor at JAMA Internal Medicine but was not involved in the editorial review or decision on this manuscript.

Meeting Presentation: This work was presented as a poster at the American Geriatrics SocietyAnnual Meeting on May 12, 2022; Orlando, Florida.

Data Sharing Statement: See Supplement 2.