Racial and Ethnic Differences in Caregiving Networks and Unmet Care Needs among People Living with Cognitive Impairment

Abstract

Unmet needs are common among people living with cognitive impairment (PLwCI) and are linked to adverse outcomes. While caregiving networks (CGNs) are crucial in supporting care, little is known regarding how CGNs relate to unmet needs for PLwCI across racial and ethnic groups. Using data from the 2023 National Health and Aging Trends Study, we adapted the latent profiles of CGN sizes and composition among Black, Latinx, and White PLwCI and assessed their associations with unmet care needs. Zero-inflated negative binomial models showed that White PLwCI in “Spousal CGNs” had more unmet needs than those in “Spouse and other CGNs.” Black PLwCI who relied on “Children and Others CGNs” experienced fewer unmet needs. Among Latinx CGNs, “Friends and others CGNs” CGNs showed fewer unmet needs. These findings underscore the importance of addressing within-group variation in CGNs to inform equitable, culturally responsive caregiving policies.

Introduction

Alzheimer’s disease and related dementias (ADRD) are the 5^th leading cause of death among individuals aged 65 and older, affecting over six million Americans. Nearly 70% of those with dementia live within the community and depend on unpaid care from over 16.1 million family members and friends (Alzheimer’s Association, 2024; Kasper et al., 2015; Wolff et al., 2016). As dementia progresses, persons living with cognitive impairment (PLwCI) require increased assistance with activities of daily living (ADLs) —ranging from self-care (e.g., personal hygiene, eating, bathing, dressing), mobility (e.g., going outside), medical management (e.g., doctor visits, taking medications), and household activities (e.g., shopping, meal preparation, paying bills) (Spillman et al., 2020; Wolff et al., 2016). Most PlwCI receive unpaid support from a large and complex network typically formed by spouses, adult children, extended family and friends, rather than a single primary caregiver (Kasper et al., 2015; Wolff et al., 2016). As the aging population grows and the demand for long-term care increases, the care-related burden on friends and family members is expected to rise (Redfoot et al., 2013; Spillman & Pezzin, 2000). This can lead individuals to rely on paid care services or leave care recipients at risk of having their needs unmet (Abrahamson et al., 2017).

Nationally, 25–33% of older adults who need assistance with ADLs report unmet care needs, meaning that basic activities, such as getting dressed and bathing, are not adequately supported (Abrahamson et al., 2017; Freedman & Spillman, 2014; Potter, 2019). Unmet care needs increase the risk of falls and injuries, health complications, and lower quality of life in older adults (Beach et al., 2020; Gaugler et al., 2005; Hayes et al., 2016; He et al., 2015). Among PLwCI, these risks are amplified, with unmet needs contributing to emergency department visits, hospitalizations, nursing home placement, and mortality (Beach et al., 2020; Black et al., 2013; Gaugler et al., 2005). While prior research has identified individual-level risk factors for unmet needs—such as being unmarried, older, lower income or poverty, minority status, and living alone without helpers (Abrahamson et al., 2017; Rahman et al., 2022). PLwCI without any helpers may be the most vulnerable. Yet, little is known about unmet needs among PLwCI who do have helpers, despite the fact that most individuals with cognitive impairment rely on at least some informal support. Among PLwCI, it remains unclear how the size and configuration of CGNs relate to unmet needs, or whether larger and more diverse networks confer protective advantages Moreover, few studies have examined how these associations may vary by race/ethnicity. To address these gaps, we focus on PLwCI with at least one helper and investigate how CGN configurations shape unmet needs and whether these patterns differ across racial/ethnic groups.

Complexity and Collaboration in Diverse Caregiving Networks

Racial and ethnic disparities further shape dementia risk and caregiving dynamics. Black and Latinx older adults experience earlier onset and higher prevalence of cognitive impairment, than their White counterparts (Alzheimer’s Association, 2024; Beach et al., 2020; Mehta et al., 2008). They are more likely to rely on unpaid familial caregivers and less likely to access paid help than their White peers (Lai et al., 2025). Black and Latinx older adults often have larger and more complex caregiving networks (CGNs) and more ADL-related care needs as well as caregivers that report more frequent engagement and time spent providing, in comparison to White caregivers. care (Rote & Moon, 2020). Despite growing recognition that caregiving is a collaborative, network-based effort, most studies have focused primarily on ADL-related needs and unmet care needs only in primary caregivers, without addressing racial/ethnic differences in the CGNs complexity or the adequacy of care provided.

CGNs—comprising of spouses, children, extended family members, and friends—vary in structure and function across racial and ethnic groups. The task-specific and hierarchical compensatory models highlight how care responsibilities and tasks (e.g., personal care, housework, mobility assistance, and transportation) (Ellis et al., 2023; Spillman et al., 2020) are shaped by relationship proximity (e.g., long-term commitment, living arrangement), availability (i.e., whether someone has the time, capacity, or geographic proximity to help), and role expectations (e.g., social roles) (Litwak et al., 1990). Typically, spousal caregivers take on intensive, ongoing tasks, such as personal care and housework, while adult children, especially middle-to-older-aged daughters, and other family members step in when spouses are unavailable (Wolff et al., 2018). Friends and neighbors generally offer social support and intermittent or ad hoc support such as transportation (Spillman et al., 2020). Thus, larger CGNs and more diverse CGNs may foster increased collaboration and buffer caregivers’ strain (Ellis et al., 2023; Hu et al., 2023).

Racial and Ethnic Differences in the CGNs for PLwCI

CGNs are shaped by demographic factors such as age, gender, and race/ethnicity, as well as by the availability of potential caregivers within the family and social network (Dilworth-Anderson et al., 2020; Fuller et al., 2020). Prior studies show that Black and Latinx caregivers are more likely than White caregivers to provide care for family and friends with cognitive or age-related impairments (Badana et al., 2019; Dilworth-Anderson et al., 2020; Fabius et al., 2020; Martinez et al., 2022). Black older adults, in particular, tend to receive care from adult children, extended family members, or friends rather than spouses partially due to lower marital rates and stronger ties to “fictive” kin (Broman, 2005; Choi et al., 2021; Ellis et al., 2023; Fabius et al., 2020). These caregiving arrangements are deeply rooted in cultural values. In Black communities, kinship reciprocity, familial obligation, and intergenerational reliance reinforce a strong preference for home-based care over formal care, such as nursing homes(Dilworth-Anderson et al., 2020; Nkimbeng & Parker, 2021). Similarly, Latinx caregiving practices reflect familismo —emphasizing family obligation, mutual support, and prioritization of elders (Chavez-Dueñas et al., 2020; Jaldin et al., 2023; Mendez-Luck et al., 2016). Thus, Latinx PLwCI frequently rely on children, spouses, and extended family for care, with women often taking on the caregiving role and formal care viewed less favorably (Chavez-Dueñas et al., 2020; Jaldin et al., 2023; Mendez-Luck et al., 2016).

Pervious research suggests that reliance on spousal support is more common among White PLwCI than their non-White counterparts (Lai et al., 2025). In contrast, the majority of Black PLwCI are supported by “children-and-others CGNs,” often without spousal involvement, and Latinx PLwCI also show strong reliance on family-based networks, primarily “children-mainly CGNs” and “children-and-others CGNs” (Lai et al., 2025). These caregiving patterns reflect cultural norms, but also stem from structural inequities, such as limited access to healthcare, language barriers, and mistrust of formal services, which reinforce reliance on unpaid caregiving in Black and Latinx communities (Armstrong et al., 2007; Cothran et al., 2015; Dilworth-Anderson et al., 2020; Samson et al., 2016).

Racial and Ethnic Differences in Unmet Need

Nearly half (44.3%) of older adults in the U.S. reported at least one ADL-related unmet need including self-care, mobility, household activities, or transportation in the past month, even among those without dementia (Beach et al., 2020; Black et al., 2013). Among PLwCI, findings are mixed, with some studies suggesting fewer unmet needs for those with severe cognitive impairment, likely due to greater visibility and support (Black et al., 2013; Miranda-Castillo et al., 2010) (Black et al., 2013), while others point to more unmet care needs when neuropsychiatric symptoms (e.g., agitation, psychosis, and sleep disturbances) are present (Black et al., 2019). The absence of spousal or child caregivers is also associated with a higher risk of unmet needs, especially among unmarried individuals (Abrahamson et al., 2017; Patterson et al., 2022). Black older adults report more unmet needs than their White peers, likely due to having higher care needs (Berridge & Mor, 2018; Black et al., 2019; Lin & Liu, 2024), but it remains unclear whether their reliance on children and extended kin alleviates or intensifies unmet needs.

Black caregivers, relative to White, often express positive attitudes toward caregiving yet face challenges coordinating care across large networks (Badana et al., 2019; Black et al., 2013; Chavez-Dueñas et al., 2020; Sörensen et al., 2002). Meanwhile, Black caregivers report higher level of strain when managing high care demands (Cothran et al., 2015; Fabius et al., 2020; Samson et al., 2016). Likewise, Latinx PLwCI often live longer with dementia than their White counterparts extending the caregiving period, yet are known to underutilize formal services such as nursing homes or paid caregivers due to cultural and linguistic barriers (Martinez et al., 2022; Nkimbeng & Parker, 2021). Despite the growing focus on racial caregiving disparities, research on CGNs characteristics remains limited. Most studies treat race as a control variable or use White caregivers as the benchmark obscuring within-group variation and reinforcing deficit-based comparisons (Cohen et al., 2019; Cothran et al., 2015; Ellis et al., 2023). While Black and Latinx families are often characterized as having broad CGNs, some individuals still experience limited access to comprehensive CGNs. Understanding both between- and within-group differences in CGNs is essential for identifying whether larger or more diverse caregiving networks buffer or contribute to unmet care needs (Choi et al., 2021; Ellis et al., 2023; Fabius et al., 2020).

To address these gaps, our study examines the association between CGNs’ characteristics and unmet care needs by race and ethnicity. Using data from the 2023 wave of the National Health and Aging Trends Study (NHATS), we address two key questions: 1) whether there are racial and ethnic differences in unmet care needs among PLwCI and 2) whether CGN characteristics—defined by network size and composition of caregiving networks—are associated with unmet care needs within Black, Latinx, and White older adults living with possible or probable dementia. By focusing on within-group differences in CGNs, our study aims to identify which caregiving structures are most protective or limiting in meeting the care needs of PLwCI, providing insight to inform culturally responsive and equitable dementia care.

Methods

Data and Sample

Data is drawn from the 2023 wave of NHATS (Round 13), a nationally representative survey of Medicare beneficiaries aged 65 years and older. NHATS collects detailed information on the older adults’ health, daily functioning, caregiving needs, and unmet care needs (Freedman & Spillman 2014). The 2023 data set and its previous wave oversampled Latinx older adults. Our analytic sample was restricted to community-dwelling adults (i.e., not residing in nursing homes or receiving care from facility staff) who met the NHATS criteria for possible or probable dementia, had at least one helper, and reported at least one care need as the presence of a care need is necessary to assess unmet needs.

Race/ethnicity was categorized as non-Hispanic White (labeled as White), non-Hispanic Black (labeled as Black), and Hispanic/Latinx (labeled as Latinx). Other groups were excluded from the analysis due to the small sample size (N=241). The final sample consisted of 2,257 older adults: 25.40% White, 28.44% Black, and 36.15% Latinx.

Measures

Independent Variables

Care Network Profiles.

We conducted a Latent Profile Analysis (LPA) to identify CGNs types for each racial/ethnic group based on: 1) network size— defined as the total number of helpers assisting with self-care/mobility, household, medical, and transportation tasks, and 2) the proportion of helpers who were spouse, adult children, family members, or friends (Blinded for review). Separate LPAs were conducted by race/ethnicity using Mplus 8.3 with robust maximum likelihood (MLR) (Muthén & Muthén, 2018), allowing for group-specific caregiving patterns that may be obscured in pooled models (Morin et al., 2016). Model fit statistics are available in Appendix A.

Panel one of Table 1 shows that four CGNs were identified among White PLwCI: 1) spouse and others (6.89%); 2) Spousal CGN (13.15%); 3) balanced CGN (4.62%; a balanced mix of spouse, children family and friends), and 4) children and others CGN (75.34%). Among Black PLwCI, three CGN clusters were observed: 1) Spouse and others (7.32%); 2) spouse CGN (5.92%); 3) children and others CGN (86.76%). For Latinx PLwCI, four CGNs were found: 1) Spouse and others CGN (9.56%); 2) spousal CGN (9.93%), 3) friends and others CGN (15.81%); 4) children mainly CGN (64.71%). Given that spouses and children are typically the primary caregiver for PLwCI and provide the majority of caregiving hours, we used the “spouse and other” group—a CGN making up with spouse, children, family and friends—as reference groups for comparison.

Table 1.

Characteristics of Older Adults Living with Cognitive Impairment by Race/ethnicity, National Health and Aging Trends Study 2023 (unweighted N=2,257)

	White	(N=799)	Black	(N=1,014)	Latinx	(N=326)
Variables	M/%	S.D.	M/%	S.D.	M/%	S.D.
Probable dementia (%)	50.45c		52.27c		60.85ab
Possible dementia (%)	49.55c		47.73c		39.15ab
CGN	3.12b	1.76	3.46ac	2.10	3.14b	1.72
Proportion of CGN (%)
Spouse	18.76b		9.96ac		15.34b
Adult children	39.52c		43.00		45.68a
Family	18.08b		25.64ac		20.99b
Friends	23.64c		21.30		17.99a
Unmet care needs	1.64	1.54	1.71+	1.28	1.75+	1.42
Number of care needs	6.2		6.52		6.33
Self-care and mobility	1.58	1.89	1.70c	2.32	1.48b	1.52
Medical	1.30c	0.97	1.33c	1.30	1.57ab	1.57
House	2.01b	1.25	2.24ac	1.51	2.09b	1.39
Transportation	1.31bc	0.78	1.25ac	0.75	1.19ab	0.78
Male (%)	43.57b		32.35a		44.27
Female (%)	56.42b		67.65a		55.73
Age	84.44bc	7.60	82.63a	7.92	82.62a	7.38
Immigrant (%)	3.99c		3.75c		32.45ab
Education (%)
Less than high school	29.68bc		58.55ac		70.02ab
High school	33.82bc		22.96ac		14.56ab
College and above	36.50bc		18.49a		15.42a

Note. Differences by race/ethnicity were examined using Pearson’s chi-square test for categorical variables and t-test for continuous variables.

CGN=caregiving network

S.D. = standard deviation

^aSignificantly different from White PLwCI at the p<0.05 level.

^bSignificantly different from Black PLwCI at the p<0.05 level.

^cSignificantly different from Latinx PLwCI at the p<0.05 level.

⁺Significantly different from White PLwCI at the p<0.10 level.

Care Assistance.

Care assistance was measured as the number of ADL-related needs for care across four domains: self-care/mobility, household activities, medical care, and transportation (Spillman et al., 2020). Self-care and mobility activities include basic personal tasks such as bathing, dressing, eating, toileting, getting out of bed, navigating indoors, and going outside (range from 0 to 7). Household activities encompass four practical tasks for daily living including laundry, shopping, preparing hot meals, and managing bills/banking (range from 0 to 4). Medical care assistance refers to help with medical appointments and medications management (range from 0 to 2). Lastly, transportation assistance involves going places outside of the home (range from 0 to 2).

Dementia Status.

Dementia status in NHATS is classified using three primary sources: self- or proxy-reported physician diagnoses (self or proxy report), AD8 Dementia Screening scores and a cognitive test assessing memory, orientation, and executive function. When cognitive tests are incomplete, AD8 scores and proxy-reported diagnoses are used for classification. Overall, cognitive status is categorized as no dementia, possible dementia, and probable dementia—reflecting increasing severity. The definitions of probable and possible dementia demonstrate fair sensitivity (71.8%) for identifying individuals diagnosed with dementia or cognitive impairment no dementia (CIND; see Kasper et al., 2013 for details on full classification methodology). Only individuals with possible or probable dementia were included.

Dependent Variable

Unmet Care Needs.

Respondents who reported any difficulty with four self-care tasks (i.e., bathing, dressing, eating, and toileting), or three mobility tasks (i.e., going outside, getting around inside, and getting out of bed) were asked whether they had experienced any adverse consequences in the past month due to lack of help with these seven task (Spillman et al., 2020). A summary score (0–7) was created to indicate the severity of unmet care needs, with higher scores reflecting more unmet needs, following the previous studies (Xiang et al., 2018).

Control Variables

We adjusted for demographic characteristics of PLwCI, including age (centered in years), gender (0=male, 1=female), and living arrangement (0=living alone, 1=living with spouse only, 2=living with spouse and with others, 3= living with others only). Socioeconomic variables included educational attainment (0=less than high school, 1=high school degree, 2=some college, 3= bachelor’s degree or higher), and log-transformed household income (NHATS imputed values). Health status was captured by a count of nine self-reported chronic conditions (e.g., heart disease, diabetes, high blood pressure/hypertension), resulting in a range from 0 to 9.

Statistical Analyses

We first conducted descriptive analyses using Pearson’s chi-squared tests and t-tests to compare sample characteristics across Whites, Blacks, and Latinx PLwCI (see Table 1). We then estimated zero-inflated negative binomial (ZINB) models to examine associations between CGN profiles and the counts of unmet care needs. ZINB was selected due to the overdispersion and a higher frequency of zero values in the outcome distribution. All models were stratified by race/ethnicity adjusting for covariates. Survey design and sampling weights were applied for all analyses. Results are presented as incidence rate ratios (IRRs) with 95% confidence intervals.

Results

Descriptive Statistics

Table 1 presents PLwCI demographic characteristics by race/ethnicity. About 68% of Black PLwCI were female, and one-third of Latinx PLwCI were foreign-born. White PLwCI were older, had a higher educational attainment, and greater income than their Black and Latinx counterparts. Nearly half of White (49.55%) and Black (47.73%) participants had possible dementia, while about 39.15% of Latinx participants were classified as having possible dementia. Black and Latinx PLwCI reported a similar number of unmet needs, marginally greater (M=1.71–1.75, p=0.07), than their White peers (M=1.64). Additionally, Black PLwCI reported more care needs related to self-care and mobility (M=1.70, p<0.05) and household activities (M=2.24, p<0.05) than White and Latinx PLwCI. While Latinx reported more medical needs (M =1.57, p<0.05), they had fewer needs for self-care and mobility (M=1.48, p<0.05) and transportation (M =1.19, p<0.05) than White and Black PLwCI. On the other hand, White PLwCI reported more transportation needs (M=1.31, p<0.05) than non-White PLwCI. (Table 1 about here). Appendix B presents the characteristics of each CGN by race and ethnicity.

Multivariate Models: Care Network Factors and Unmet Needs

Table 3 presents the results of zero-inflated negative binomial regression models predicting incidence rate ratios (IRR) of unmet needs among White, Black, and Latinx PLwCI. Among White PLwCI, those in “Spousal CGNs” had an IRR reflecting 2.52 times higher unmet care needs compared to those in “Spousal and other CGNs” (IRR=2.52, p<.05). No significant differences in unmet needs were observed for other CGNs types, including “Balanced CGNs” and “Children and others CGNs,” relative to the “spouse and other CGNs.” In the zero-inflation portion of the model, no statistically significant associations were found, indicating that all CGN groups had similar odds of being in the always-zero group—individuals who never experienced unmet needs.

Table 3.

Zero-inflation Negative Binomial Regression Models on CGNs and Unmet Needs

White (N=799)	IRR	SE	Black (N=642)	IRR	SE	Latinx (N=816)	IRR	SE
CGN (ref. Spouse and others CGN)			CGN (ref. Spouse and others CGN)			CGN (ref. Spouse and others CGN)
Balanced CGN	1.15	(0.38)	Children & others CGN	0.22**	(0.11)	Spousal CGN	1.38	(0.34)
Spousal CGN	2.52*	(1.13)	Spousal CGN	1.08	(0.42)	Non-family CGNs	0.43**	(0.12)
Children & others CGN	1.85	(0.78)				Children mainly CGN	0.69	(0.16)
Possible dementia (ref. probable dementia)	1.03	(0.12)	Possible dementia (ref. probable dementia)	1.11	(0.17)	Possible dementia (ref. probable dementia)	1.05	(0.13)
Number of tasks			Number of tasks
Household	1.08	(0.06)	Household	0.92*	(0.04)	Household	0.82***	(0.04)
Medical assistance	1.05	(0.07)	Medical assistance	1.01	(0.08)	Medical assistance	0.91	(0.07)
Self-care and Mobility	1.26***	(0.05)	Self-care and Mobility	1.56***	(0.10)	Self-care and Mobility	1.64***	(0.06)
Transportation	1.09	(0.08)	Transportation	1.01	(0.09)	Transportation	1.01	(0.09)
Number of chronic diseases	1.17***	(0.04)	Number of chronic diseases	1.08	(0.06)	Number of chronic diseases	1.15***	(0.04)
Living arrangement (ref. living alone)			Living arrangement (ref. living alone)			Living arrangement (ref. living alone)
With spouse only	1.77+	(0.57)	With spouse only	0.29**	(0.11)	With spouse only	0.31***	(0.08)
With spouse and with others	1.53	(0.44)	With spouse and with others	0.16***	(0.06)	With spouse and with others	0.36***	(0.09)
With others only	1.07	(0.15)	With others only	0.54***	(0.10)	With others only	0.37***	(0.07)
Female	0.99	(0.12)	Female	1.36	(0.26)	Female	1.13	(0.16)
Age (centered)	0.99	(0.01)	Age (centered)	0.98**	(0.01)	Age (centered)	1.00	(0.01)
Education (ref. less H.S.)			Education (ref. less H.S.)			Education (ref. less H.S.)
High school	1.13	(0.18)	High school	1.29	(0.22)	High school	0.65**	(0.11)
Some college	0.90	(0.16)	Some college	1.43	(0.32)	Some college	0.84	(0.20)
College and above	1.35+	(0.22)	College and above	1.08	(0.22)	College and above	0.59**	(0.12)
Log income	0.84*	(0.06)	Log income	0.92	(0.08)	Log income	0.81**	(0.06)
Inflation model	OR			OR			OR
Balanced CGN	0.00	(0.00)	Children & others CGN	1.20	(4.85)	Spousal CGN	3.10*	(1.55)
Spousal CGN	5.36	(4.44)	Spousal CGN	2.70		(5.14)	0.01***	(0.01)
Children & others CGN	0.00	(0.03)				Children mainly CGN	0.77	(0.58)
Constant	0.56	(0.48)	Constant	3.58	(3.25)	Constant	11.38**	(9.54)
Observations	949		Observations	745		Observations	959

Note. se in parentheses.

^***p<0.001

^**p<0.01

^*p<0.05

⁺p<0.10

For Black PLwCI, those in “Children and others CGNs”—which comprised 87% of Black PLwCI—were less likely to have unmet care needs compared to those in “Spousal and other CGNs” (IRR= 0.22, p<0.01). There were no significant differences in unmet care needs between spousal CGNs and “spouse and other CGNs.” Also, there are no statistically significant associations found in the zero-inflation models, suggesting that all CGN groups had similar odds of never experiencing unmet needs. Because the distribution of CGN types among Black PLwCI was highly skewed, with most classified into the Children & Others profile, we conducted sensitivity checks with rotated reference groups. These analyses confirmed the robustness of our findings, indicating that the lower unmet needs observed in the Children & Others network type reflect a substantive within-group pattern rather than an artifact of unbalanced group sizes.

Among Latinx PLwCI, those in the “Friends and other CGNs” had a significantly lower IRR of unmet care needs compared to those in “Spouse and other CGNs” (IRR=0.43, p<.01). However, the zero-inflation portion of the model revealed that individuals in “friends and others” had lower odd of being in the always-zero groups (OR=.01, p<.001), suggesting that nearly all members of these groups and were at some risk of experiencing unmet needs. In addition, Latinx PLwCI in a spousal CGNs have 3.1 times higher odds of belonging to the structural zero class (OR=3.10, p<.05) groups. Thus, Latinx PLwCI cared for only by a spouse were more likely than those in “spousal and other” to experience no unmet needs at all, though this group did not show a significantly lower incidence rate among those at risk.

However, our findings indicate that Black and Latinx PLwCI reported slightly higher unmet needs compared to their White peers (Table 1: Black mean = 1.71, Latinx mean = 1.75, White mean = 1.64). Among White PLwCI, those reporting greater needs for self-care and mobility had a higher IRR of unmet care needs (IRR=1.26, p<.001). Among Black PLwCI, greater need for self-care and mobility were associated with a higher rate of unmet care needs (IRR=1.56, p<.001), while a greater need for household assistance was associated with a lower rate of unmet needs (IRR=0.92, p<.05). Additionally, among Black PLwCI, living with others —whether with spouse only, with spouse and other, or with others only—was associated with lower IRR of unmet care needs compared with living alone (IRRs=0.16–0.54, ps<.01). Similarly, among Latinx PLwCI, greater self-care and mobility needs were linked to a higher rate of unmet care needs (IRR=1.64, p<.001), whereas a greater need for household assistance was associated with a lower rate of unmet needs (IRR=0.82, p<.001). Compared with living alone, living with others (e.g., including with a spouse only, with a spouse and others, with others only) was associated with lower IRR of unmet care needs among Latinx PLwCI (IRRs=0.31–0.37, p<.001).

Sensitivity Analysis: The Role of Dementia Status

To rigorously assess the robustness of our main findings, sensitivity analyses (see Table S1-1) were conducted by restricting the sample only to individuals with probable dementia (model 2) and with individuals with probable dementia only (model 3). This restriction led to substantively different results for the “Children & Others” caregiving network, particularly between White and Black PLwCI. For White older adults, the “Children & Others CGN” was associated with an increase in unmet need in the probable dementia subsample, a finding that contrasts with the results from the primary analysis. While “Spousal CGN” show no significant differences compared with the “Spouse and Other CGNs (reference group).” This association suggests that for White older adults, the “Children & Others CGN” is linked to a higher likelihood of unmet needs when the care demands are certainly greater due to more advanced dementia.

However, for Black PLwCI, the initial protective association of the “Children & Others CGN” observed in the primary analysis disappeared when examining only those with probable dementia. This implies that the familial support inherent in this network configuration, which is beneficial in earlier or less certain stages of cognitive impairment, is insufficient to prevent unmet needs once the disease reaches a more advanced stage.

Discussion

Unmet needs are a known risk factor for adverse health outcomes among community-dwelling older adults and are particularly prevalent among PLwCI (Beach et al., 2020; Freedman & Spillman, 2014; Potter, 2019). While facing a higher risk of dementia, Black and Latinx older adults often have larger CGNs than their White peers (Alzheimer’s Association, 2024; Cothran et al., 2015). Using data from the 2023 wave of NHATS, we examined how CGN characteristics relate to unmet needs, particularly among racially and ethnically diverse older adults with possible or probable dementia. Our findings show that PLwCI with self-care and mobility limitations were more likely to experience unmet needs, regardless of race/ethnicity. However, Black and Latinx PLwCI needing help with household tasks were less likely to report unmet needs, possibly reflecting cultural caregiving norms. Notably, PLwCI with possible dementia had the highest risk of unmet needs. Our findings further demonstrate within-group variation in CGNs, with these differences linked to unmet care needs among White, Black, and Latinx CGNs.

Our descriptive findings show Black and Latinx PLwCI reported similar or marginally higher levels of unmet needs than their White peers (p=0.07), despite Black PLwCI reporting greater overall care needs than White PLwCI (except for transportation). This contrasts with previous research suggesting higher unmet needs among Black older adults rerceiving ADL assistance (Berridge & Mor, 2018; Lin & Liu, 2024; Moon et al., 2019). Moreover, Black PLwCI were more likely to receive help with self-care and mobility, and household activities, likely reflecting both the higher rates of chronic health conditions and cultural expectations for family involvement in daily tasks. (Dilworth-Anderson et al., 2020; Nkimbeng & Parker, 2021). Latinx PLwCI, however, reported a greater need for medical assistance and fewer transportation needs, possibly due to cultural and linguistic barriers that hinder medication management, communication with healthcare providers, and navigation of the healthcare system (Martinez et al., 2022; Nkimbeng & Parker, 2021). White PLwCI reported higher care needs for transportation than their non-White counterparts. Although transportation is vital for social engagement and access to care, it generally requires less intensive caregiving effort and more logistical support than self-care, mobility, or household assistance.

Second, White PLwCI who relied solely on a spouse reported 2.5 times more unmet care needs than those CGN that include a spouse and additional family members or friends. About 13% of White PLwCI had a spouse as their solo caregiver —a pattern rooted in proximity and traditional marital caregiving norms (Wolff et al., 2018). However, caregiving responsibilities shared between a spouse and other caregivers—such as adult children, extended family, or friends—brings additional resources, time, and attention, which helps reduce unmet needs (Herrera et al., 2013). In contrast, spousal caregivers who are the sole source of care may face physical, emotional, or informational limitations that increase care gaps when support is not shared (Chen et al., 2021; Potter, 2019). It may be that CGNs that include other members help ease this burden, ensure care continuity, and increase access to formal supports like paid caregivers or healthcare services among White PLwCI, which further supplement informal caregiving efforts and reduce unmet needs. These findings align with the task-specific model, which suggest that care responsibilities are distributed based on the availability, capability, and relational closeness of network members (Fuller et al., 2020). A larger network with multiple caregivers can be more beneficial for older adults needing diverse assistance (Choi et al., 2021).

As for Black PLwCI, those receiving care from CGNs composed of children, family, and friends —but no a spouse—had a lower likelihood of unmet needs compared to those with spousal and others CGNs. This “Children and others” CGN accounted for nearly 87% of Black PLwCI networks. Consistent with prior studies showing that Black older adults depend more on non-spousal caregivers (e.g., adult children, extended family members, fictive kins or friends) (Broman, 2005; Choi et al., 2021; Ellis et al., 2023; Fabius et al., 2020), our finding extend this knowledge by demonstrating that these networks are not only more common but also more effective at preventing unmet care needs. This protective effect may be rooted in longstanding cultural values in Black communities. Kinship reciprocity foster ongoing exchanges of emotional and instrumental support among relatives and fictive kin, creating flexible and overlapping caregiving roles that minimize care gaps (Armstrong et al., 2007; Cothran et al., 2015; Dilworth-Anderson et al., 2020; Samson et al., 2016). A strong familial obligation and moral expectation to care for aging family members encourages monitoring and responsiveness to older adults’ well-being and needs (Badana et al., 2019; Black et al., 2013; Chavez-Dueñas et al., 2020; Sörensen et al., 2002), while intergenerational reliance distributes caregiving across adult children, grandchildren, and extended family members—broadening the pool of available support and care (Cothran et al., 2015; Fabius et al., 2020; Samson et al., 2016). Additionally, a cultural preference for home-based care and a general reluctance toward institutionalization helps maintain close physical and emotional proximity between caregivers and care recipients, enabling timely responses (Armstrong et al., 2007; Cothran et al., 2015; Dilworth-Anderson et al., 2020; Samson et al., 2016). Together, these cultural values cultivate resilient, responsive CGNs that help reduce unmet care needs for Black PLwCI.

Among Latinx PLwCI, those supported by “Friends and Others” CGNs, which make up 16% of the groups, had significantly fewer unmet needs than those in “Spouse and Others” CGNs. However, the zero-inflated model showed that PLwCI in “Friends and Others” CGNs were less likely to be in the structural-zero group (OR=0.01, p< .001), indicating that they still faced some unmet needs. This suggests that while the “Friends and Others” CGNs may not shield care recipients entirely from the possibility of unmet needs, they may be more effective at managing and reducing them once they arise. Several cultural and structural factors may help explain this finding. In Latinx communities, caregiving often extends beyond the nuclear family to include friends, neighbors, and fictive kin (Jaldin et al., 2023; Martinez et al., 2022). Collectivist values and traditions such as compadrazgo—which formalize caregiving ties through ritual and social bonds—further strengthen these non-spousal networks. These broader caregiving webs bring diverse resources and perspectives, allowing for flexible and responsive care, even if not entirely protective against unmet needs. By contrast, spousal CGNs may offer more stable and continuous care, thereby increasing the likelihood of zero unmet needs. However, spouses—except in Black older adult dyads—may face physical or emotional limitations that hinder their ability to meet all care needs. Thus, while spousal CGNs may offer protection against any unmet needs, friend and family-based CGNs may be more adaptable and effective in managing more care needs.

Finally, our sensitivity analyses also suggest that vulnerabilities exist for “Children & Others CGN” under high demand. For White older adults the ”Children & Others” CGN may become strained or structurally inadequate as the demands increase with dementia, suggesting such networks may be less equipped than the “Spouse and Other CGNs” to meet increasing care demands in the context of dementia. In contrast, the ”Spousal CGN” showed no significant difference compared to the reference group, indicating that spouse-inclusive networks may possess more robust capabilities or access to adequate resources, allowing them to maintain service provision as needs intensify. However, for Black older adults, the protective association of the ”Children & Others CGN” observed in the primary analysis was eliminated in the probable dementia subsample. This implies that while the familial support inherent in this network is beneficial in earlier or less certain stages of cognitive impairment, it is ultimately insufficient to prevent unmet needs once the disease reaches a more advanced or confirmed stage. This finding highlights a systemic challenge where supportive networks, while protective initially, may lack the formal resources or capacity to fully absorb the substantial and chronic burdens associated with probable dementia, underscoring the necessity of targeted, culturally sensitive formal support that scales with dementia progression.

Limitations & Future Directions

This study has several limitations. First, its cross-sectional design prevents causal inference between the CGN characteristics, care needs, and unmet care needs. While this study provides a snapshot of CGN profiles and unmet needs at a single point in time, caregiving dynamics shift over time due to progression of dementia, changes in family structures, and access to formal care. Future research should adopt a longitudinal approach to capture this evolution. Second, we did not examine the underlying causes of unmet needs, which may stem from limited resources, caregiver capacity, or system-level barriers due to the limitations in the data. Future research efforts should aim to explore these pathways to better understand how unmet needs develop among PLwCI. Third, caregiver experiences such as burden or satisfaction were not assessed, though these likely interact with CGN dynamics and care outcomes, especially in underrepresented groups. Future research should include Asian and other racial/ethnic groups and explore regional differences to broaden insights into caregiving disparities. Moreover, although sampling weights were applied to address selection bias, the findings cannot be generalized to the broader population, because the analyses was limited to individuals living with cognitive impairment who had at least one helper. Also, dementia status was classified using a validated survey-based algorithm with moderate sensitivity/specificity (71.8%), which may introduce some individuals classified as having cognitive impairment may not meet formal diagnostic criteria for dementia. Future research would benefit from the use of clinical diagnoses. Finally, the distribution of CGNs among Black PLwCI was highly skewed, with most belonging to the Children & Others profile. While sensitivity analyses with rotated reference groups confirmed the robustness of the results, these checks cannot fully eliminate concerns about the unbalanced distribution among Black PLwCI.

Despite these limitations, our study highlights the need for culturally responsive policy solutions. For White older adults, expanding support for spousal caregivers is essential—through respite services, caregiver training, and initiatives that facilitate greater involvement of adult children or other family members in CGNs. For Black PLwCI, policies should recognize and support extended CGNs by broadening eligibility for informal caregiver recognition to include adult children, grandchildren, and fictive kin who are often the main sources of care. Expanding eligibility would ensure that these non-spousal caregivers have access to financial assistance, respite care, and training programs that may be restricted to spouses or legal guardians. Culturally tailored programs—such as caregiver education, intergenerational housing assistance, and support for navigating the health care system—can further strengthen these extended caregiving networks. Similarly, for Latinx PLwCI, policy frameworks should redefine “family caregivers” to include close friends and kin-like ties, while also addressing linguistic and cultural barriers through bilingual resources and community outreach, particularly for those needing medical assistance. Across all groups, moving beyond one-size-fits-all caregiving models and investing in culturally grounded supports will be critical to reducing unmet needs and promoting equity in dementia care.

Conclusion

This study demonstrates how CGN characteristics shape unmet care needs among Black, Latinx, and White PLwCI. White PLwCI relying solely on a spouse reported significantly more unmet needs, highlighting the limitations of spousal-only caregiving. Black PLwCI cared for by adult children, extended family, or friends had the fewest unmet needs, possibly reflecting culturally rooted practices of kinship reciprocity and intergenerational support. Latinx PLwCI in “Friends and Others” networks also experienced fewer unmet needs, possibly due to collectivist caregiving norms and strong social ties beyond the nuclear family. These findings highlight the need for policies that recognize and strengthen diverse, culturally grounded CGNs to advance equity in dementia care.

Table 2.

Latent Profiles Mean and Characteristic of CGN by Race and ethnicity

Panel 1. White PLwCI (N=799)
Profiles	2_Spousal	1_Balanced	3_Spouse & Other	4_Children & Others
N=	105	37	55	602
%	13.15	4.62	6.89	75.34
Mean CGN size	1.000	3.432	2.000	3.005	P<.001
Proportion of spouse	1.000	0.305	0.500	0.000	P<.05
Proportion of children	0.000	0.356	0.309	0.494	P<.01
Proportion of family	0.000	0.140	0.109	0.230	P<.001
Proportion of friends	0.000	0.199	0.082	0.276	P<.01
Panel 2. Black PLwCI (N=642)
Profiles	3_Spousal	1_Spouse & others	2_Children & others
N=	38	47	557
%	5.92	7.32	86.76
Mean CGN size	1.000	2.58	3.04		P<.05
Proportion of spouse	1.000	0.417	0.001		P<.001
Proportion of children	0.000	0.319	0.487		P<.05
Proportion of family	0.000	0.167	0.294		P<.05
Proportion of friends	0.000	0.098	0.217		P<.05
Panel 3. Latinx PLwCI (N=816)
Profiles	1_Spousal	2_Spouse & others	3_Non-Family	4_Children mainly
N=	81	78	129	528
%	9.93	9.56	15.81	64.71
Mean CGN size	1.000	2.719	3.413	2.946
Proportion of spouse	1.000	0.399	0.000	0.001
Proportion of children	0.000	0.375	0.274	0.646
Proportion of family	0.000	0.104	0.117	0.298
Proportion of friends	0.000	0.122	0.609	0.054