Receipt of Caregiving and Fall Risk in U.S. Community-Dwelling Older Adults

Abstract

Background

Falls and fall-related injuries (FRI) are common and costly occurrences among older adults living in the community, with increased risk for those with physical and cognitive limitations. Caregivers provide support for older adults with physical functioning limitations, which are associated with fall risk.

Design

Using the 2004–2012 waves of the Health and Retirement Study, we examined whether receipt of low (0–13 weekly hours) and high levels (≥14 weekly hours) of informal care or any formal care is associated with lower risk of falls and FRIs among community-dwelling older adults. We additionally tested whether serious physical functioning (≥3 activities of daily living, ADL) or cognitive limitations moderated this relationship.

Results

Caregiving receipt categories were jointly significant in predicting NIFs (p=0.03) but not FRIs (p=0.30). High levels of informal care category (p=0.001) and formal care (p<0.001) had stronger associations with reduced fall risk relative to low levels of informal care. Among individuals with ≥3 ADLs, fall risks were reduced by 21% for those receiving high levels of informal care; additionally, FRIs were reduced by 42% and 58% for those receiving high levels of informal care and any formal care. High levels of informal care receipt were also associated with a 54% FRI risk reduction among the cognitively impaired.

Conclusions

Fall risk reductions among older adults occurred predominantly among those with significant physical and cognitive limitations. Accordingly, policy efforts involving fall prevention should target populations with increased physical functioning and cognitive limitations. They should also reduce financial barriers to informal and formal caregiving.

INTRODUCTION

Non-injurious falls (NIFs) and fall-related injuries (FRIs) (hereafter jointly referred to as “falls”) among adults ages ≥65 years old are a major cause of morbidity and injury-related mortality (1). More than one in three community-dwelling older adults experience at least one fall in a given year (2). NIFs are a strong predictor of FRIs, with 20–30% of older individuals who report a NIF also reporting an FRI (3). In turn, FRIs account for up to one-third of nursing home admissions (4). Older individuals with physical limitations of activities of daily living (ADL) and cognitive impairment have higher rates of nursing home utilization (5) and may also be at particular risk for a fall (3), suggesting that individuals with limitations are at an increased risk of losing their ability to function within the community following a FRI. Given the aging of the U.S. population and the growing prevalence of older adults with multiple physical functioning and cognitive limitations, prevention of falls is a policy priority to improve the quality of life of older adults.

Most falls among community-dwelling older adults occur in the home and are potentially attributable to the physical home environment (6). Home fall risk may be due to environmental hazards, but is also associated with risk behaviors (such as hurrying or taking the stairs without use of a handrail) resulting from unsupervised activities that occur in the home. Living alone may result in activity avoidance (7) that decreases physical functioning and increases social isolation—which may lead to greater fall risk (8).

Among the community-dwelling population, 61% of frail and 90% of disabled older adults receive help with personal self-care or household activities (9). An informal caregiver gives unpaid help to a relative or friend with an illness or disability or due to age. When informal support is not sufficient or when informal caregiving resources or other social supports are unavailable, older adults may turn to formal assistance (10). Caregivers—both formal and informal—assist care recipients with activities of daily living (ADLs) such as eating, toileting, and bathing, and instrumental activities of daily living (IADLs), such as shopping and cooking and they also provide emotional support (11). Their presence may reduce social isolation, which might mitigate fall risks (12). Additionally, they may assist recipients with modifications of the home environment (13).

While there is an extensive literature on caregiving and caregiver burden, as well as numerous examinations of falls predictors, there has been limited inquiry into the relationship between receiving care and falls. A qualitative study has examined caregiver burden associated with falls (13), but not the impact of caregiving for fall prevention from the perspective of the recipient of care. This study aims to assess fall risk reductions associated with different levels of care receipt among community-dwelling older adults with physical or cognitive limitations.

Conceptual Framework

Fall prevention may be informed by Person-Environment (P-E) Fit theory, which highlights the role of the environment in shaping older adult health outcomes (14). In P-E Fit theory, health outcomes are related to the fit between environmental demands, or “press,” and an individual’s health status, or “competence.” Whether the demands are greater than competence—or vice versa—determines adaptive responses that may impact fall risk. Environmental demands related to falls can include physical demands, such as a slippery shower floor or uneven household lighting throughout the day.

P-E Fit theory thus suggests that a caregiver can influence an older adult’s fall risk via two pathways. First, a caregiver can reduce environmental “press,” both through modification of the home environment to increase a care recipient’s familiarity and comfort in the home environment (13), and by providing support for ADL/IADL such as transfers, bathing, and medication, to increase care recipient safety during fall-risk activities. Such assistance may be particularly helpful to older adults with physical and cognitive impairment who may have difficulties assessing fall risks inherent in performing routine physical activities (15). With greater time spent providing care, caregivers may offer increased levels of supervision and guidance that can reduce fall risk opportunities in older adults’ homes (16). Second, caregivers can increase “competence” by providing stimulation—whether social or physical—that enhances functioning, awareness of environmental risks, healthy behavior, and emotional health (17). For instance, caregivers may address fears or misunderstandings of fall risks (12). Without enough stimulation, physical functioning of individuals can decrease (18).

Compared to informal caregivers, formal caregivers may improve the balance between “press” and “competence” to a greater extent. They may have more training and objectivity to improve care recipients’ fall-preventing adaptive responses to their home environment (19), while informal caregiver burden and emotional investment may result in inattention to certain needs or challenges faced by care recipients, resulting in poorer prevention (13). Compared to formal caregivers, informal caregivers may also lack adequate knowledge regarding safety and practice with assistive devices and equipment (19).

Based upon the P-E Fit model, we test three hypotheses regarding the relationship between receipt of caregiving and fall risk among community-dwelling older adults with some physical or cognitive functioning limitations. The three hypotheses are:

1. Greater levels of informal care will be associated with reduced likelihood of falls (H1);

2. Compared to all levels of informal care, receipt of formal care will be associated with a greater reduction in the likelihood of falls (H2);

3. Fall risk reductions associated with all levels of informal care and formal care receipt will be greater for more impaired individuals, i.e., those with more physical limitations and those with cognitive impairment (H3).

METHODS

The Health and Retirement Study (HRS) is a national longitudinal panel study of older Americans (≥50) with biannual survey waves. Each wave respondents are asked to re-interview. HRS re-interview response rates have typically been at or above 90% (20). The HRS also adds a new cohort of participants to the study each six years. Study inclusion criteria of community-dwelling older adults ≥65 years old with ≥1 ADL/IADL who participated in one or more adjacent survey waves (e.g., 2008–10) resulted in an analytic sample of 4,528 individuals and a total of 6,871 pooled observations (or paired person-waves, PPW) across five survey waves (2004–2012), with four person-wave periods—2004–6, 2006–8, 2008–10, 2010–12). We did not use earlier HRS waves because they lack spousal caregiving information (available beginning in 2000) or uniform measurement of physical activities (beginning in 2004). Observations were dropped for individuals in a nursing home during the interview to establish a community-dwelling sample. Each respondent contributed a maximum of one observation per PPW period, or four total PPW observations for the complete study period. There were 1,711, 1,751, 1,817, and 1,592 unique individuals in each PPW period. 2,905 respondents contributed one PPW observation, while 1,066, 394, and 163 contributed 2, 3, and 4 PPW observations, respectively.

We measured falls using a three-level categorical variable: (1) no falls (reference category), (2) “NIF” (without FRI), and (3) “FRI.” For the care receipt measure, we adapted categories and thresholds for high versus low caregiving amounts from prior research using HRS data (21) (22): (1) no formal or informal care receipt (reference category), (2) “low informal” care receipt (1–14 weekly hours) and no formal care receipt, (3) “high informal” care receipt (≥14 weekly hours) and no formal care, (4) any formal caregiving receipt (with or without informal care). We generated two variables to reflect physical and cognitive impairment: physical impairment was measured as an indicator “ADL≥3” for having three or more ADL limitations, which was the 75^th percentile status for individuals with ≥1 ADL and reflects a population with substantial long-term care needs (23).

Cognitive impairment was measured as an indicator “Cognitively Impaired” for having a score of ≤8 (range of 0–35) on HRS’ modified Telephone Interview of Cognitive Status (TICS), an alternative to the Mini-Mental State Examination (24–28). Proxy respondents rated the cognitive functioning of respondents who did not answer the TICS telephone questions. Following Suthers, Kim, and Crimmins (2003) (29), we classify these non-TICS respondents as cognitively impaired using proxy responses (see Appendix). Models with (a) all respondents and (b) only TICS respondents were separately estimated, with results from (a) presented below and from (b) presented in Tables A4–A8.

To assess fall risk differences among individuals receiving different levels and types of care, we controlled for a number of potential confounders. We adjusted for self-reported health (fair/poor, good, very good/excellent); number of chronic health conditions (an index scored from 0–5 for osteoarthritis, stroke, heart disease, high blood pressure, and diabetes) that can affect spatial orientation and stability (30–32); functional limitations (an index scored from 0–12 for reported difficulties with activities such as walking several blocks) that can reflect problems with stability during movements (31, 33, 34); eyesight (1–6, where 1=legally blind and 6=excellent) (31, 32, 35) and hearing (1–5, where 1=poor and 5=excellent) that affect the ability to notice environmental hazards (35, 36) and disability (whether a respondent received SSI or SSDI) (31, 33) that reflects difficulties with physical functioning. Most falls involve a precipitating activity (15). Accordingly, respondents were asked about vigorous physical activity (e.g., running or jogging, swimming), moderate physical activity (gardening, walking at a moderate pace), and light physical activity (vacuuming, laundry) ≥1 time per week. We also controlled for use of psychiatric medications (tranquilizers, antidepressants, or pills for nerves), (31, 35), a self-report variable with high concordance with psychiatric medication use (37).

Study Design and Model Estimation

Chi-squared and ANOVA tests were used to test for significance in bivariate analyses. For multivariate analyses, multinomial logistic (ML) models were estimated to test the association of fall risk with receipt of care adjusted for the sociodemographic, health-related, and medication use variables discussed above to predict fall risk (Model 1). Small-Hsiao tests were conducted to assess the appropriateness of ML models (38). Multiple observations were used if a respondent took part in more than one set of adjacent survey waves so we adjusted for non-independent error term using generalized estimating equation (GEE) independent standard errors (SE). In order to eliminate reverse causality (falls may have occurred prior to receipt of care), we lagged all of the predictor and control variables by using the values measured in the survey wave prior to the falls measurement.

To test H1, we used F-tests to assess whether fall risk varied according to the level of informal care received (i.e., low versus high levels of informal care received). To test H2, we used Model 1 and conducted F-tests to compare predicted fall risk reduction between informal care (low and high levels) and formal care categories. To test H3, we estimated two models, one for a moderating effect of physical limitations and one for a moderating effect of cognitive limitations. We tested the moderating effects of physical and cognitive limitations using joint significance tests of the interaction terms of the care receipt categories with (a) ADL≥3 status (Model 2 for n=33,118 individuals) and (b) cognitive impairment status (Model 3). We present estimates for the cognitive impairment moderated model (Model 3 for n=33,126 individuals) further stratified by ADL≥3 status (Model 3a for n=31,545 individuals with 0–2 ADLs and Model 3b for n=1,573 for individuals with ≥3 ADLs). Given their stratification by ADL status, Models 2 and 3a–b did not include controls correlated with ADLs—functional limitations and disability status. Models 3a–3b did not include cognitive impairment status as a control variable. To facilitate interpretation of results, we estimated marginal probabilities (MP) using Stata’s margins command. MPs indicate the absolute change in the probability of each category of a categorical response variable across the categories of the exposure variable. We used the adjusted fall risk for an individual not receiving any care as our baseline fall risk. If the baseline risk for a NIF were 34.2 and the MP for an individual with low informal care receipt were 4.0, then the NIF risk for the individual with low informal care would be [34.2+4.0=] 38.2%. The percent increase in NIF risk for that individual would be [4.0/34.2=]11.7%. We used standard F-tests and non-linear Wald tests using two-tailed significance cutoffs at p<0.05.

Limitations

There were several potential limitations. The presumed association between unmeasured (poor) health and caregiving is positive, meaning that as health worsens care receipt becomes more likely; additionally, the presumed association between (poor) health and falls is positive, meaning that as health worsens falls become more likely. Because both of these unmeasured associations involve positive relationships, the cumulative bias due to not controlling for unmeasured health confounding is positive. Thus, endogeneity in the caregiving-falls relationship due to omitted variable bias is presumed to lead to an upwardly biased estimate of the magnitude of the care receipt-falls association such that caregiving may appear to be related to a greater probability—rather than, as hypothesized, a lower probability—of falls. Reverse causality bias is another potential issue. By using lagged variables, we ensured that care receipt was assessed prior to a fall. However, given the gaps between surveys, there may be substantial lags between care receipt and the occurrence of falls. However, this should have the residual effect of reducing any observed effect of care receipt on falls. Accordingly, our estimates are likely conservative, meaning a more limited reduction in fall risk than hypothesized.

RESULTS

Sample Characteristics

The sample of older adults was 42% male and predominantly white (see Table A1). The sample was slightly older with lower educational levels, and similar gender proportion and household income levels as the U.S. older adult population (39). Informal caregivers provided most of the care reported by the recipients (Table 1). Of respondents receiving care, 89% received informal care while 11% received formal care. Of those receiving informal care, 40% received lower levels (1–13 hours) and 60% received greater levels (≥14 hours) of weekly informal care. (See Table A2 for additional descriptive statistics.)

Table 1.

Bivariate Associations: Percentages of Older Adults with ≥1ADL/IADL Difficulty Who Fell by Caregiving Receipt and Fall Status in Prior Survey Wave, 2004–12

Prior Wave	N	Overall (%)	No Fall (%)	Any Fall (%)	Non-Injurious Fall (%)	Fall-related Injury (%)
Overall	7,821	100.0	51.4	48.6	32.0	16.6
Caregiving*
Yes	1,383	20.1	40.8	59.2	36.4	22.9
No	5,488	79.9	48.4	51.6	33.7	17.9
Weekly Caregiver Hours*
0 formal and 0 informal	5,488	79.9	48.4	41.6	33.7	17.9
0 formal and 1–13 informal	491	7.2	36.7	63.3	42.6	20.8
0 formal and ≥ 14 informal	731	10.6	42.4	57.6	33.4	24.2
Any formal	161	2.3	46.0	54.0	31.1	23.0
Fall Status*
Yes	3,170	46.1	29.8	60.6	58.7	63.9
No	3,701	53.9	70.2	39.5	41.3	36.2
Fall Category*
No Fall	3,701	53.9	70.2	39.5	41.3	36.2
Non-Injurious Fall	2,134	31.1	19.9	40.9	45.3	32.9
Fall-related Injury	1,036	15.1	9.9	19.7	13.4	30.9

^*p<.05.

Note: Data are from the RAND HRS (version O), RAND “Fat Files,” and HRS Core Files, years 2004, 2006, 2008, 2010, and 2012 (survey waves 7–11). The characteristics above apply to individuals in the analytic sample, which includes 4,528 unique individuals and 6,871 observations. Caregiving and fall information in the first column (“Prior Wave”) are lagged data, i.e., data from the prior survey wave. Data in the fall outcomes columns (e.g., “No Fall,” “Any Fall”) are current data—i.e., data in from the current survey wave. Chi-squared tests were used to assess statistically significant differences in the fall outcome variable (i.e., no fall, non-injurious fall, and fall-related injury) at p<.05 (*) for categorical predictor variables. For instance, an asterisk next to “Caregiving” indicates that there is a statistically significant difference in fall outcomes (in the current survey wave) for those who receive and those who did not receive care (in the prior survey wave). ANOVA tests were used to assess statistically significant differences for continuous predictor variables. Due to rounding some rows (no fall plus fall percentage) may not add to 100% (for overall percentage).

Unadjusted Results

Among those not receiving any care, 52% of respondents reported falling, while among those receiving care, 59% reported falling (Table 1). Specifically, among those receiving low and high levels of informal care, 63% and 58% fell, respectively, while among those receiving any formal care, 54% fell. These differences across care receipt categories were statistically significant (p<0.001).

Adjusted Results: Main Model

In Model 1, relative to no care received, none of the individual categories of care receipt (lower informal, high informal, or formal care) were associated with the risk of a NIF or an FRI (Table 2). Caregiving receipt categories were jointly significant in predicting NIFs (p=0.03), but not FRIs (p=0.30). In partial support of H1 and of H2, among the categories of the care receipt variables, the high levels of informal care category (p=0.01) and the formal care category (p=0.02) had stronger associations with a reduced NIF risk, relative to low levels of informal care (Table A3). However, while H2 was supported for those with ≥3 ADLs, it was not supported for those with 0–2 ADLs.

Table 2.

Marginal Probabilities (MP) of Fall Outcomes Associated with Caregiving for Older Adults with ≥1ADL/IADL Difficulty, 2004–12 (n=6,871)

	Non-Injurious Fall		Fall-Related Injury
	MP	p	MP	p
Weekly caregiver hours (reference: no care received)
No care received (baseline fall risk)	34.2		18.9
0 formal and 1–13 informal	4.0	0.09	−0.8	0.97
0 formal and ≥14 informal	−2.4	0.23	−0.6	0.71
Any formal	−4.4	0.22	−2.7	0.30
Age (continuous, ≥65)	−0.1	0.48	0.3	<0.001**
Male (reference: female)	5.1	<0.001**	−3.6	<0.001**
Race (reference: Non-Hispanic White)
Non-Hispanic African-American	−3.3	0.04*	−4.9	<0.001**
Hispanic	−6.1	0.003*	−1.0	0.54
Other	−5.9	0.04*	−0.4	0.88
Educational level (reference: less than high school)
High school graduate/GED	−0.9	0.55	1.9	0.11
Some college	1.2	0.52	1.9	0.17
College and above	0.6	0.75	3.4	0.04*
Income ($1,000)	0.0	0.44	0.0	0.69
Wealth ($1,000)	0.0	0.68	0.0	0.75
Eyesight (1–6)	−0.3	0.56	0.4	0.35
Hearing (1–5)	0.8	0.13	−0.1	0.08
Prior fall
Non-injurious fall	22.2	<0.001**	6.3	<0.001**
Injurious fall	4.1	0.02*	22.0	<0.001**
Number of functional limitations (0–12)	0.4	0.13	0.2	0.29
Number of chronic conditions (0–6)	0.2	0.63	0.7	0.11
Vigorous physical activity	0.5	0.77	−2.5	0.12
Moderate physical activity	−0.8	0.53	0.9	0.40
Light physical activity	0.7	0.58	−1.5	0.14
Cognitive impairment	−1.6	0.36	0.5	0.73
Disabilitye	−2.6	0.35	0.3	0.90
Psychiatric medication	4.2	0.01*	5.3	<0.001**
Log pseudo-likelihood	−6,603.4
N (Person-waves)	6,871
N (Individuals)	4,528

^*=p<.05,

^**=p<0.001.

Note: Data are from the RAND HRS (version O), RAND “Fat Files,” and HRS Core Files, years 2004, 2006, 2008, 2010, and 2012 (survey waves 7–11). The model specification uses complete-case analysis. The reference category in the multinomial logit model is “no fall.” Clustering due to repeated observations was controlled for using modified “sandwich” estimators. MPs indicate the absolute change in the probability of each category of a categorical response variable across the categories of the exposure variable. An MP of 4.0 for a non-injurious fall (NIF) indicates a 4.0 percentage-point increased risk of a NIF. Baseline fall risk was assessed by estimating the predicted probability of a NIF or an FRI—adjusted for sociodemographic and health characteristics—of an individual not receiving care. A baseline NIF risk of 34.2 means that respondents who did not receive any care in the prior HRS survey wave had a 34.2% risk of experiencing an NIF in the current HRS survey wave. Therefore, for an individual with 0 formal and 1–13 informal care hours (low informal care receipt) received, the adjusted NIF risk is 34.2 + 4.0 = 38.2%. The percent increase in risk between those with low informal care receipt and those with no care received is 4.0 / 34.2 = 11.7%.

Adjusted Results: Moderating Effects of Physical and Cognitive Impairment

In Model 2, in support of H3, caregiving receipt categories were jointly significant in predicting NIFs (p=0.001) and FRIs (p<0.001) for those with serious physical functioning limitations. Care receipt was not associated with fall risk reduction among individuals with 0–2 ADLs (Table 3). However, among individuals with ≥3 ADLs, NIF risks were reduced by 7 (p=0.01) percentage points for those receiving high levels of informal care. Given the 34% baseline risk of a NIF, those translate to a 21% risk reduction. FRI risks were also reduced for individuals with ≥3 ADLs, by 10 percentage points for low (p=0.004) and high (p<0.001) levels of informal caregiving receipt and by 14 percentage points (p<0.001) for formal care recipients. Given the 24% baseline risk of an FRI, this translates to respective 42% and 58% risk reductions.

Table 3.

Marginal Probabilities (MP) of Fall Outcomes Associated with Caregiving for Older Adults by ADL Status, 2004–12 (n=33,118)

	Non-Injurious Fall		Fall-Related Injury
	MP	p	MP	p
0–2 ADLs
No care received (baseline fall risk)	23.5		9.9
0 formal and 1–13 informal	6.0	0.01*	4.7	0.003*
0 formal and ≥14 informal	3.9	0.10	3.4	0.04*
Any formal	−2.2	0.65	1.7	0.63
≥3 ADLs
No care received (baseline fall risk)	34.3		23.6
0 formal and 1–13 informal	4.5	0.25	−10.0	0.004*
0 formal and ≥14 informal	−6.8	0.01*	−10.4	<0.001**
Any formal	−7.1	0.10	−13.8	<0.001**
Log pseudo-likelihood	−25,779.5
N (Person-waves)	33,118
N (Individuals)	12,275

^*=p<.05,

^**=p<0.001.

Note: Data are from the RAND HRS (version O), RAND “Fat Files,” and HRS Core Files, years 2004, 2006, 2008, 2010, and 2012 (survey waves 7–11). The model specification uses complete-case analysis. The reference category in the multinomial logit model is “no fall.” Clustering due to repeated observations was controlled for using modified “sandwich” estimators. MPs indicate the absolute change in the probability of each category of a categorical response variable across the categories of the exposure variable. An MP of 6.0 for a non-injurious fall (NIF) indicates a 6.0 percentage-point increased risk of a NIF. Baseline fall risk was assessed by estimating the predicted probability of a NIF or an FRI—adjusted for sociodemographic and health characteristics—of an individual not receiving care. A baseline NIF risk of 23.5 means that respondents with 0–2 ADLs who did not receive any care in the prior HRS survey wave had a 23.5% risk of experiencing an NIF in the current HRS survey wave. Therefore, for an individual with 0–2 ADLs and 0 formal and 1–13 informal care hours (low informal care receipt) received, the adjusted NIF risk is 23.5 + 6.0 = 29.5%. The percent increase in risk between those with 0–2 ADLs and low informal care receipt and those with 0–2 ADLs and no care received is 6.0 / 23.5 = 25.5%.

In Model 3, in partial support of H3, caregiving receipt categories were jointly significant in predicting FRIs (p=0.04) but not NIFs (p=0.54) for those with cognitive impairment (Table 4). Care receipt was not associated with fall risk reduction among individuals not having cognitive impairment. However, among the cognitively impaired, high levels of informal care receipt were associated with an 7 percentage-point (p=0.02) risk reduction in FRI. Given the 13% baseline risk of an FRI, this translates to a 54% risk reduction. Most of the other associations of fall risk with care receipt were in the expected direction but not statistically significant due to smaller group size (n=58 cognitively impaired individuals receiving formal care).

Table 4.

Marginal Probabilities (MP) of Fall Outcomes Associated with Caregiving for Older Adults by Cognitive Impairment Status, 2004–12 (n=33,126)

	Non-Injurious Fall		Fall-Related Injury
	MP	p	MP	p
No cognitive impairment
No care received (baseline fall risk)	23.9		10.4
0 formal and 1–13 informal	3.6	0.05*	2.3	0.06
0 formal and ≥14 informal	−0.7	0.67	2.4	0.04*
Any formal	−5.9	0.06	0.7	0.75
Cognitive impairment
No care received (baseline fall risk)	23.8		12.5
0 formal and 1–13 informal	6.1	0.25	0.2	0.61
0 formal and ≥14 informal	−4.4	0.07	−3.1	0.09
Any formal	−3.9	0.39	−6.7	0.02*
Log pseudo-likelihood	−25,763.7
N (Person-waves)	33,126
N (Individuals)	12,277

^*=p<.05,

^**=p<0.001.

Note: Data are from the RAND HRS (version O), RAND “Fat Files,” and HRS Core Files, years 2004, 2006, 2008, 2010, and 2012 (survey waves 7–11). The model specification uses complete-case analysis. The reference category in the multinomial logit model is “no fall.” Clustering due to repeated observations was controlled for using modified “sandwich” estimators. MPs indicate the absolute change in the probability of each category of a categorical response variable across the categories of the exposure variable. An MP of 3.6 for a non-injurious fall (NIF) indicates a 3.6 percentage-point increased risk of a NIF. Baseline fall risk was assessed by estimating the predicted probability of a NIF or an FRI—adjusted for sociodemographic and health characteristics—of an individual not receiving care. A baseline NIF risk of 23.9 means that respondents with no cognitive impairment who did not receive any care in the prior HRS survey wave had a 23.9% risk of experiencing an NIF in the current HRS survey wave. Therefore, for an individual with no cognitive impairment and 0 formal and 1–13 informal care hours (low informal care receipt) received, the adjusted NIF risk is 23.9 + 3.6 = 27.5%. The percent increase in risk between those with no cognitive impairment and low informal care receipt and those with no cognitive impairment and no care received is 3.6 / 23.9 = 15.1%.

H3 was supported for those with cumulative physical and cognitive limitations. In models 3a and 3b stratified by physical impairment status, risk reductions among the cognitively impaired were greater among those who also had physical functioning limitations (Table 5). Among cognitively impaired individuals with ≥3 ADL difficulties, those who received high levels of informal or formal care, FRI risk was decreased by 17 and 26 percentage points, respectively. Given the 38% baseline risk of an FRI, this translates to 45% and 68% respective risk reductions. Comparatively, for cognitively impaired respondents with 0–2 ADL difficulties, care receipt was not significantly associated with FRI risk reduction.

Table 5.

Marginal Probabilities (MP) of Fall Outcomes Associated with Caregiving for Older Adults Difficulties by Cognitive Impairment (CI) and ADL Status, 2004–12

	Non-Injurious Fall		Fall-Related Injury
	MP	p	MP	p
0–2 ADLs (n=31,545)
No cognitive impairment
No care received (baseline fall risk)	23.1		9.6
0 formal and 1–13 informal	5.7	0.01*	5.1	0.002*
0 formal and ≥14 informal	5.2	0.07	−4.3	0.02*
Any formal	−1.0	0.85	0.2	0.96
Cognitive impairment
No care received (baseline fall risk)	22.6		10.6
0 formal and 1–13 informal	8.5	0.21	0.6	0.90
0 formal and ≥14 informal	0.9	0.84	0.4	0.91
Any formal	−6.9	0.50	6.0	0.57
≥3 ADLs (n=1,573)
No cognitive impairment
No care received (baseline fall risk)	37.9		30.8
0 formal and 1–13 informal	5.6	0.20	−11.6	0.001*
0 formal and ≥14 informal	−5.0	0.15	−6.3	0.04*
Any formal	−7.6	0.21	−6.6	0.19
Cognitive impairment
No care received (baseline fall risk)	38.5		38.0
0 formal and 1–13 informal	0.6	0.96	−3.8	0.68
0 formal and ≥14 informal	−8.7	0.10	−17.1	0.001*
Any formal	−3.3	0.64	−25.7	<0.001*

^*=p<.05,

^**=p<0.001.

Note: Data are from the RAND HRS (version O), RAND “Fat Files,” and HRS Core Files, years 2004, 2006, 2008, 2010, and 2012 (survey waves 7–11). The model specification uses complete-case analysis. The reference category in the multinomial logit model is “no fall.” Clustering due to repeated observations was controlled for using modified “sandwich” estimators. MPs indicate the absolute change in the probability of each category of a categorical response variable across the categories of the exposure variable. An MP of 5.7 for a non-injurious fall (NIF) indicates a 5.7 percentage-point increased risk of a NIF. Baseline fall risk was assessed by estimating the predicted probability of a NIF or an FRI—adjusted for sociodemographic and health characteristics—of an individual not receiving care. A baseline NIF risk of 23.1 means that respondents with no cognitive impairment and 0–2 ADLs who did not receive any care in the prior HRS survey wave had a 23.1% risk of experiencing an NIF in the current HRS survey wave. Therefore, for an individual with no cognitive impairment and 0–2 ADLs and 0 formal and 1–13 informal care hours (low informal care receipt) received, the adjusted NIF risk is 23.1 + 5.7 = 28.8%. The percent increase in risk between those with low informal care receipt and those with no care received is 5.7 / 23.1 = 24.7%.

DISCUSSION

Our study was the first to examine the association of fall risk with receipt of informal and formal care among community-dwelling older adults. The findings were consistent with notable fall risk reductions associated with receipt of high-level informal and formal care. However, risk reductions predominantly occurred among individuals with significant physical and cognitive limitations.

There are several important takeaways from these findings. First, caregiving fall prevention is most beneficial among older adults with physical and cognitive limitations, with the largest observed risk reductions among care recipients. Physically and cognitively impaired populations are most at risk for falls and fall injuries. Among individuals with ≥3 ADL limitations in this sample, nearly two-thirds had a fall in the prior two years, and 42% of those were FRIs, relative to one-in-five fall in the U.S. community-dwelling older adult population (3). Efforts to address fall prevention among community-dwelling older adults should focus on those with substantial limitations. Fall-prevention strategies through caregiving are expected to result in notable FRI risk reduction—up to 60% risk reductions for physically and cognitively impaired individuals, potentially avoiding the high costs of FRIs and reducing the likelihood of nursing home admission. Creating a policy landscape that increases access to informal and formal care could reduce fall risks and help maintain independence for vulnerable populations of cognitively and physically impaired older adults who are most at risk for falls and institutionalization. On the other hand, we observed fall risk increases among non-impaired care recipients. If high-level caregivers are more sensitive to care recipient needs (40) while low-level caregivers have limited fall awareness (41, 42), it may be of benefit to jointly target low-level informal caregivers and their care recipients. Early interventions such as home environmental modifications and steps to address physical and cognitive decline (6, 43) could have indirect and long-term impacts on future population fall risks while reducing the burden of fall prevention on informal caregivers.

Second, risk reduction differences between informal and formal care suggest the potential need for training and support of informal caregivers. Caregivers could be trained to help older adults better navigate the myriad services and providers that currently are part of fall prevention efforts (44). Coordination of fall-prevention activities is not seamless across these sites and providers—for instance, not all at-risk individuals receiving a falls assessment in a physician’s office receive a referral for an exercise class or home safety assessment (44). Frail and cognitively impaired older adults may particularly struggle to coordinate needed services. Caregiver education should include training and support with navigating the existing fall-prevention system for their care recipients. Educational opportunities for informal caregivers through state Medicaid agencies or primary care practices might increase their knowledge of fall risks and benefit care recipients at risk for a fall (45).

In order to encourage caregivers’ contribution to fall risk reduction efforts, policy makers might consider bolstering financial incentives for informal caregivers. In our sample, 20% of care recipients reported their informal caregivers being paid for providing help. As discussed by Reinhard et al. (2015), this could include federal or state tax credits, Social Security caregiver credits, or expansion of consumer-directed care programs—in which Medicaid provides beneficiaries with money to hire caregivers, including family members (46). Such programs can help informal caregivers provide care while avoiding loss of income and benefits due to reduced work hours while recognizing their considerable contributions.

Conclusion

This study observed reductions in fall risk among community-dwelling older adults who received high levels of informal caregiving and formal caregiving. These reductions were particularly pronounced among older individuals with physical functioning limitations and cognitive impairment. The findings suggest that policy makers should (1) target caregiving fall-prevention strategies to populations with increased functional and cognitive limitations, (2) provide training and access to educational resources regarding fall prevention activities and fall-prevention care coordination for informal caregivers, and (3) consider reducing financial barriers to provision of informal caregiving and access to formal care. An emphasis on increasing the role of informal and formal caregiving may help address the substantial burden of falls among older adults.