Differences in phenotypic and functional aging trajectories among people aging with disability and the US general population

Abstract

Background

People aging with disability (PAwD) may experience distinct life course exposures that impact their aging trajectories, yet the extent to which these trajectories differ from those of the US general population remains understudied. This study aimed to investigate the impact of long-term disability on phenotypic and functional aging trajectories, and assess the role of key demographic and social determinants on those trajectories.

Methods

Data from the 2006-2020 waves of the Health and Retirement Study were used. Phenotypic aging was measured by the metabolic dysfunction risk score (MDRS) and functional aging was operationalized through limitations in activities of daily living (ADL) and instrumental activities of daily living (IADL). Growth curve modeling was employed, adjusting for age, birth cohort, gender, race/ethnicity, and educational attainment.

Results

A total of 7234 were included in the analysis, comprising 873 PAwD and 6361 individuals from the general population. Results suggested pronounced disparities in phenotypic and functional aging trajectories between groups, with compounded vulnerabilities among marginalized subgroups. PAwD exhibited significantly higher MDRS and ADL/IADL scores, with more rapid escalation over time, suggesting premature and accelerated aging. Racial/ethnic minorities and individuals with lower education faced intersecting disadvantages in both domains. Woman aging with disability exhibited marginally greater MDRS than men. Cohort-disability interaction revealed that older PAwD cohorts experienced steeper functional decline, while younger cohorts showed accelerated metabolic dysregulation.

Conclusions

These findings highlight the imperative for inclusive public policies that address structural inequalities, particularly educational and racial disparities. Implementing preventative, disability-informed rehabilitative interventions is also essential to promote aging equity among PAwD.

Introduction

“Aging with disability” refers to individuals who experience the onset of disabilities from birth to middle age and continue to live with these long-term disabilities as they reach the later stages of life.^1,2 People aging with disability (PAwD) may be distinguished from individuals who develop disability later in life due to the aging process, and are often described as “aging into disability.” Due in part to larger demographic shifts and increased life expectancies, the population of PAwD has received greater global awareness as their numbers have steadily increased.³ Despite this recognition, there remains limited understanding of the experience of aging with disability, with research disproportionately focused on impairments associated with the aging process itself, rather than on the effects of living life with a pre-existing disability. This gap in knowledge necessitates a focused effort to understand the aging patterns of PAwD to develop specialized knowledge that meets the diverse needs to support successful aging.⁴

It is well-established that PAwD, despite heterogeneity in their conditions, share the common experience of disability impacting various aspects of their lives and aging process.³ This shared experience has prompted a concerted effort to understand the intersection of aging and disability among PAwD, with a specific focus on aging trajectories and attributes. Existing research has indicated that PAwD may have different aging experiences compared to the general population, particularly in terms of increased susceptibility to metabolic dysregulation, and functional deterioration.^5,6 However, many prior studies are constrained by cross-sectional designs, lack of comparison groups, focus on a single outcome, and small convenience samples. Furthermore, there is limited research on health and aging that specifically includes study samples of PAwD per se, highlighting the need for further investigation.^7–9 Consequently, several critical topics remain underexplored empirically, including the extent to which aging trajectories differ between PAwD and the general population, as well as the temporal dynamics of these changes.

In addition, there is a dearth of research that incorporates an intersectional lens to understand the health and aging of PAwD. As highlighted by the National Institutes on Minority Health and Health Disparities (NIMHD), there is a paucity of studies and interventions addressing intra- and inter-group differences among various minoritized populations in people with disability (PWD).^10,11 This gap encompasses various factors (eg, socioeconomic status, and racial/ethnic minority status) affecting health outcomes and the aging process. Previous research has suggested a role for intersectional factors such as gender and race/ethnicity in metabolic dysregulation in PWD.^12,13 However, much of the existing research is limited in scope, constrained by design, or pertains to the general population of PWD. To address these gaps, this study aimed to investigate the differences in aging trajectories between PAwD and the general population, analyzing how long-term disability impacts these trajectories at both phenotypic and functional levels. It also sought to examine the impact of various key sociodemographic factors on aging trajectories, highlighting potential compounded effects.

Methods

Data source

This study utilized 2006-2020 RAND Health and Retirement Study (HRS) longitudinal data, 2006-2020 RAND HRS Fat Files, and 2006-2016 Biomarker Data. The HRS, the largest ongoing panel study of the US adults aged 50 and older, has been conducted biannually since 1992 and periodically refreshed with new cohorts. Recognized in prior scoping review as one of the few US public datasets capturing disability onset, the HRS enables identification of PAwD.¹⁴ Details on the data design and methodology are available in prior publications.^15,16

Participants

Within the HRS, participants were asked whether they had any “impairments or health problems” that created limitations: (1) “in any way in activities;” (2) in “the kind or amount of paid work you can do;” and/or (3) in “the kind of amount of work you can do around the house.” Affirmative responses to any of those three items were followed by questions regarding the onset year and date of the impairment. PAwD were then identified as those with a health problem that created limitations in any of the three areas, before the age of 45, with those limitations starting at least five years before the baseline measurement point. Disability onset age was calculated by subtracting the birth year from the self-reported onset year. This approach aligns with prior research identifying these items as the most viable indicators of PAwD in the HRS, as they are the only questions capturing both the presence and the timing of disability.¹ The comparison group, referred to as the general population, comprised individuals who did not meet the criteria for inclusion as PAwD.

Measurements

Aging

This study conceptualized aging using the “Time and the Metrics of Aging” model, which views aging not merely as a sequential process of change but as a hierarchical interrelation intricately linking the temporal progression of human aging across biological, phenotypic, and functional dimensions.¹⁷ Given that the HRS included only one wave of biological aging data at the time of this study, the analysis focused solely on phenotypic and functional trajectories.

Phenotypic aging was measured using the metabolic dysfunction risk score (MDRS), which evaluates cardiometabolic health through diagnostic cut points used as key indicators in the clinical management of metabolic dysregulation.¹⁸ The MDRS is calculated through blood chemistries from dried blood spots, physical examination data, and self-reported physician diagnoses. The total score ranges from 0 to 5, calculated as follows: one point for systolic blood pressure ≥ 130 mm Hg, diastolic blood pressure ≥ 80 mm Hg, or a reported diagnosis of hypertension; one point for glycosylated hemoglobin ≥ 6.5% or a reported diagnosis of diabetes; one point for total cholesterol ≥ 240 mg/dL; one point for low high-density lipoprotein cholesterol, defined as <40 mg/dL for men and < 50mg/dL for women; and one point for waist circumference >94 cm in men or >80 cm in women. Higher scores indicate greater metabolic dysregulation. The cut points are detailed in prior literature.¹⁸

Functional aging was assessed through self-reported limitations in six activities of daily living (ADL) (ie, walking across a room, dressing, bathing, eating, getting in or out of bed, and using the toilet) and five instrumental activities of daily living (IADL) (ie, using a telephone, managing money, taking medications, shopping for groceries, and preparing hot meals) at each survey wave. These 11 items were aggregated to compute a continuous composite score, with higher scores indicating more severe limitations, ranging from 0 to 11.¹⁹

Covariates

Sociodemographic variables included age, gender, race/ethnicity, and educational attainment. Age was recorded in single-year increments, then centered around the cohort median and divided by 10. This adjustment was made to address the potential confounding between age and cohort effects due to their high correlation, and to minimize bias stemming from systematic differences in age across birth cohorts.^20–22 Gender (male [reference group] or female), race/ethnicity (non-Hispanic [NH] White [reference group], NH Black, Hispanic, or other), educational attainment (less than a high school diploma [reference group], high school diploma, some college, or a college diploma and higher) were based on self-reported responses.

Cohort effects were considered to account for heterogeneity in birth cohorts. The seven original HRS-created birth cohort classifications in the HRS data (categorical cohorts) were recoded into a continuous variable ranging from 1 to 8 in ascending order by birth year: The Asset and Health Dynamics Among the Oldest Old (AHEAD) cohort (born before 1924), the Children of the Depression (CODA) cohort (1924-1930), and the HRS cohort (1931-1941) was divided into HRS1 (1931-1936) and HRS2 (1937-1941), the War Baby (WB) cohort (born 1942-1947), the Early Baby Boomer (EBB) cohort (born 1948-1953), the Mid Baby Boomer (MBB) cohort (born 1954-1959), and the Late Baby Boomer (LBB) cohort (1960-1965). The values were then centered around the WB cohort, resulting in a range from −4 to 3, where each unit increase corresponds to approximately a 6-year later birth year.

Study population

Participants with at least two waves of valid MDRS measurements and complete covariates were included to mitigate potential selection bias that could arise from limiting the population to individuals with all three measurements, as it might exclude those who died or were unable to participate due to health conditions (eg, institutionalization). Of the 42 405 HRS respondents, 7234 met these criteria (Figure S1). The phenotypic aging analysis included 18 431 person-years across three waves (8 years) of biomarker data, while the functional aging analysis used 46 551 person-years from eight waves (14 years).

Statistical analyses

Growth curve modeling (GCM), a subset of mixed-effects models, was employed to examine the longitudinal aging trajectories at both phenotypic and functional levels. GCM accounts for within-individual clustering and unbalanced time points, allowing the inclusion of all participants regardless of the number of waves attended, thus, mitigating bias from attrition.^21,23 The GCM used in this study is structured in a two-level framework: the first level models individual trajectories, while the second level models between-individual variations through coefficients derived from the first level.

$$\begin{array}{c}\mathrm{Level}\mathrm{ }1:\mathrm{ }{y}_{ij}={\mathrm{\beta }}_{0i}+{\mathrm{\beta }}_{1i}·Ag{e}_{ij}+{\mathrm{\beta }}_{2i}·Ag{e}_{ij}^2+{\mathrm{\epsilon }}_{ij}\\ \begin{array}{c}\mathrm{Level}\mathrm{ }2:\mathrm{ }{\mathrm{\beta }}_{0i}={\mathrm{\gamma }}_{00}+{\mathrm{\gamma }}_{01}·z_i+{\mathrm{\zeta }}_{0i}\\ \begin{array}{c}{\mathrm{\beta }}_{1i}={\mathrm{\gamma }}_{10}+{\mathrm{\gamma }}_{11}·z_i+{\mathrm{\zeta }}_{1i}\\ {\mathrm{\beta }}_{2i}={\mathrm{\gamma }}_{20}+{\mathrm{\gamma }}_{21}·z_i.\end{array}\end{array}\end{array}$$

The dependent variable $y_{ij}$ represents phenotypic or functional aging for the ith individual at the jth observation. Since participants may not have the same number of observations, it is assumed that there are $n_i$ measurements for each individual ($1=j=n_i$). $\mathrm{Ag}{\mathrm{e}}_{ij}$ refers to the normalized age, calculated by subtracting cohort median from the age and divided by 10. ${\mathrm{\beta }}_{0i}$ reflects the average value of $y$ when $\mathrm{Age}=0$, that is, when the individual is at the median age of their cohort. ${\mathrm{\beta }}_{1i}$ is the linear slope, reflecting the rate of aging, with higher value indicating a faster rate of aging. ${\mathrm{\beta }}_{2i}$ captures whether the aging rate changes over time. $z_i$ represents a vector of time-invariant covariates, such as disability, birth cohort, and gender, along with their interactions. All $\mathrm{?}$ coefficients are fixed effects, while ${\mathrm{?}}_{0i}$ and ${\mathrm{?}}_{1i}$ are random effects on the trajectory curve’s intercept and slope, respectively. ${\mathrm{?}}_{ij}$ is independent random error. Both models employed the same GCM formulation. The functional aging model allowed correlated random effects, while the phenotypic model assumed independence.

Model selection was based on Akaike Information Criterion (AIC) and the Bayesian Information Criterion (BIC), with lower values indicating better model fit. For nested models, likelihood ratio tests were conducted. All statistical analyses and visualization were performed in R software (version 4.4.0) (R Foundation for Statistical Computing, Vienna, Austria) using the lme4 package.²⁴ Statistical significance was set at P < .05 (two-tailed). The HRS was approved by the Institutional Review Board at the University of Michigan.

Results

Participant demographics

Of the 7234 participants, 873 were PAwD (12%) (mean age = 62.5 [SD = 8.9]; duration of disability = 30.05 [SD = 13.81]). PAwD had fewer individuals in older cohorts (AHEAD, CODA, HRS) and more in younger cohorts (EBB, MBB, LBB) compared to the general population (P < .001). PAwD had lower educational attainment, with fewer achieving college degrees or higher compared to their counterparts (P < .05) and were more likely to belong to racial minority groups (P < .01). Regarding aging indicators, PAwD exhibited significantly higher baseline MDRS scores (P < .01) and greater ADL/IADL (P < .001; Table 1).

Table 1.

Baseline demographics of eligible participants.

	Total	PAwD	General population	P value
	(N = 7234)	(n = 873)	(n = 6361)
Age, mean (SD)***	66.7 (9.4)	62.5 (8.9)	67.3 (9.4)	<.0001
Birth cohort, n (%)***				<.0001
AHEAD	285 (3.94)	9 (1.03)	276 (4.34)
CODA	858 (11.86)	57 (6.53)	801 (12.60)
HRS	2903 (40.13)	260 (29.78)	2643 (41.56)
WB	1294 (17.89)	155 (17.75)	1139 (17.91)
EBB	1457 (20.14)	288 (32.99)	1169 (18.38)
MBB	337 (4.66)	81 (9.28)	256 (4.02)
LBB	100 (1.38)	23 (2.63)	77 (1.21)
Education, n (%)*				.014
< High school	1224 (16.92)	162 (18.56)	1062 (16.70)
High school diploma	2619 (36.20)	326 (37.34)	2293 (36.05)
Some college	1707 (23.60)	219 (25.09)	1488 (23.39)
College degree/more	1684 (23.28)	166 (19.01)	1518 (23.86)
Gender, n (%)				.136
Male	2869 (39.66)	326 (37.34)	2543 (39.98)
Female	4365 (60.34)	547 (62.66)	3818 (60.02)
Race/ethnicity, n (%)**				.007
Non-Hispanic White	5575 (77.07)	642 (73.54)	4933 (77.55)
Non-Hispanic Black	844 (11.67)	112 (12.82)	732 (11.50)
Hispanic	650 (8.99)	87 (9.97)	563 (8.85)
Other	165 (2.28)	32 (3.67)	133 (2.09)
Metabolic Dysregulation Score, mean (SD)**	2.22 (1.01)	2.31 (1.02)	2.21 (1.01)	.007
Function Limitation, mean (SD)***	0.31 (0.97)	0.67 (1.34)	0.27 (0.90)	<.0001
Disability, mean (SD)
Duration of disability		30.05 (13.81)
		Range (5-76)
		Median (28)
Onset age of disability		32.47 (10.01)
		Range (0-45)
		Median (35)

P values are represented as:

^* P < .05.

^** P < .01.

^*** P < .001.

Abbreviations: AHEAD = Asset and Health Dynamics among the Oldest Old; CODA = Children of the Depression; EBB = Early Baby Boomer; HRS = Health and Retirement Study; HS = high school; LBB = Late Baby Boomer; MBB = Mid Baby Boomer; PAwD = people aging with disability; SD = standard deviation; WB = War Babies.

Baseline participant characteristics, adjusted for sampling weights, are summarized in Table S1. Although specific values differ, the overall trends between the two groups remain consistent. The mean MDRS does not show a significant difference between groups; however, ADL/IADL scores are significantly higher in PAwD (P < .001; Table S1).

Phenotypic aging trajectories

There was a tendency for MDRS to increase with age. The estimated linear rate of increase in MDRS was approximately 0.1 points every 10 years (P < .05, Table 2), indicating a gradual worsening of metabolic dysfunction over time. However, the predicted phenotypic aging trajectory exhibited a curvilinear pattern, characterized by a negative quadratic slope, suggesting a deceleration in the rate of increase in MDRS as age advanced (Figure 1). On average, PAwD exhibited higher MDRS than the general population (P < .001). When using age-centered data, MDRS in PAwD appeared to increase more steeply than in the general population, suggesting a greater impact of aging on PAwD (Figure 1).

Table 2.

Results of the growth curve model for predicting phenotypic aging.

	Model 1			Model 2
	Coefficient	Standard error	P value	Coefficient	Standard error	P value
Fixed effects
Intercept
(Intercept)***	2.296	0.032	<.001	2.304	0.032	<.001
Cohort***	−0.035	0.007	<.001	−0.035	0.007	<.001
Education (Ref = < High school)
High school diploma***	−0.118	0.030	<.001	−0.118	0.030	<.001
Some college***	−0.216	0.033	<.001	−0.215	0.033	<.001
College degree/more***	−0.383	0.034	<.001	−0.383	0.034	<.001
Race/Ethnicity (Ref = NH White)
NH Black***	0.360	0.034	<.001	0.362	0.034	<.001
Hispanic***	0.218	0.040	<.001	0.219	0.040	<.001
Other	0.076	0.074	.300	0.076	0.074	.301
Gender (Ref = male)***	0.119	0.020	<.001	0.106	0.022	<.001
Disability (Ref = No)	0.141	0.036	<.001	0.065	0.053	.220
Disability × Cohort	0.001	0.019	.952	−0.003	0.020	.892
Disability × Race
NH Black	−0.089	0.092	.333	−0.093	0.092	.313
Hispanic	−0.056	0.103	.588	−0.061	0.103	.554
Other	0.219	0.169	.194	0.228	0.169	.177
Disability × Gender				0.122	0.063	.0529
Linear slope
Age*	0.112	0.049	.025	0.112	0.049	.025
Cohort***	0.052	0.011	<.001	0.052	0.011	<.001
Education (Ref= < High school)
High school diploma	0.027	0.048	.577	0.026	0.048	.579
Some college	0.049	0.051	.338	0.049	0.051	.343
College degree/more*	0.104	0.052	.046	0.104	0.052	.046
Race/Ethnicity (Ref = NH White)
NH Black*	0.110	0.053	.039	0.109	0.053	.039
Hispanic	0.075	0.061	.217	0.075	0.061	.217
Other	0.166	0.112	.137	0.167	0.112	.137
Gender (Ref = Male)	0.028	0.031	.367	0.029	0.031	.354
Disability (Ref = No)	0.013	0.055	.814	0.013	0.055	.818
Disability × cohort*	0.064	0.031	.039	0.064	0.031	.039
Disability × Race
Yes × NH Black	0.002	0.143	.990	0.003	0.143	.985
Yes × Hispanic	0.175	0.157	.267	0.176	0.157	.263
Yes × Other*	−0.571	0.259	.027	−0.569	0.259	.027
Quadratic slope
Age2  **	−0.100	0.037	.007	−0.100	0.037	.007
Cohort	−0.029	0.021	.161	−0.029	0.021	.151
Random effects (variance)
Intercept	0.538			0.537
Growth rate	0.163			0.159
Residual	0.389			0.390
Model fit
AIC	46 555.3			46 554.2
BIC	46 813.4			46 820.2
Deviance	46 489.3			46 486.2
Likelihood ratio test						.0529
Total number (individual)	7234			7234
Total number (person-year)	18 431			18 431

P values are represented as:

^* P < .05.

^** P < .01.

^*** P < .001.

Abbreviations: AIC = Akaike Information Criterion; BIC = Bayes Information Criteria; NH = non-Hispanic; Ref = reference group.

Figure 1.

Predicted phenotypic aging trajectories by disability status. (Top-left) Predicted phenotypic aging trajectories by disability status across age; (Top-right) predicted phenotypic aging trajectories by disability status across centered age; (Bottom-left) Predicted phenotypic aging trajectories by disability status and gender (Male); and (Bottom-right) Predicted phenotypic aging trajectories by disability status and gender (Female); MDRS = metabolic dysregulation risk score.

Building on these findings, the analysis revealed significant cohort effects in phenotypic aging. Later-born cohorts exhibited lower baseline MDRS (coefficient = −0.035, P < .001) but a more rapid escalation over time (coefficient = 0.052, P < .001). This indicates that a cohort born six years later demonstrated an aging rate approximately 0.052 points faster. Additionally, the interaction between disability status and cohort was significant for the linear slope (coefficient = 0.064, P < .05), indicating that younger cohorts of PAwD exhibit an accelerated aging rate compared to their counterparts (Figure 2).

Figure 2.

Predicted aging trajectories by disability status and cohort across age and centered age. (Top-left) Phenotypic aging trajectories by disability status and cohort; (Top-right) Phenotypic aging trajectories by disability status and cohort based on centered age; (Bottom-left) Functional aging trajectories by disability status and cohort; and (Bottom-right) Functional aging trajectories by disability status and cohort based on centered age. ADL = activities of daily living; AHEAD = Asset and Health Dynamics Among the Oldest Old; CODA = Children of the Depression; EBB = Early Baby Boomer; HRS = Health and Retirement Study; IADL = instrumental activities of daily living; LBB = Late Baby Boomer; MBB = Mid Baby Boomer; MDRS = metabolic dysregulations risk score; WB = War Babies.

There were significant differences in gender, with females having an average MDRS that was higher than males (coefficient = 0.106, P < .001; Figure S2). However, no gender differences were found in the rate of phenotypic aging. The interaction between disability and gender was marginally significant (coefficient = 0.122, P = .05), indicating a larger MDRS gap between women aging with disability and women from the general population than between men (Figure 1). There were no gender differences in the rate of aging between PAwD and their counterparts.

Racial differences in average MDRS exhibited significant variations (Figure S3), with PAwD showing higher MDRS than the general population across all groups (Figure S4). NH Black and Hispanic individuals exhibited higher MDRS than NH Whites (coefficients = 0.360 and 0.218, respectively; P < .001), and NH Blacks also showed a faster aging rate (coefficient = 0.110, P < .05). However, the interaction between disability and race/ethnicity did not result in significant differences in MDRS, indicating consistent disparities regardless of disability status.

Higher educational attainment was associated with lower average MDRS (Figure S5). Compared to those with less than a high school education, high school graduates had a lower average MDRS (coefficient = −0.118, P < .001), some college education was associated with a further reduction (coefficient = −0.216, P < .001), and a college degree or higher corresponded to the largest reduction (coefficient = −0.383, P < .001). These findings highlight the inverse relationship between educational attainment and MDRS. However, no significant differences in aging rate were observed by education. Although PAwD exhibited higher MDRS across all educational levels, the interaction between education and disability status did not yield statistically significant differences in trajectories (Figure S6). Additional exploratory models incorporating three-way interactions among disability, education, and race/ethnicity are summarized in Table S2.

Functional aging trajectories

For all participants, trajectories of functional aging reveal a pattern of progressively increasing limitations with advancing age. The estimated linear rate of increase was 0.785 per decade (P < .001), with a positive quadratic slope of 0.381 (P < .001), indicating an acceleration in functional aging over time (Table 3). On average, PAwD had higher functional limitation scores (coefficient = 0.494, P < .001), with a significantly accelerated rate of functional aging across time (coefficient = 0.268, P < .001; Figure S7).

Table 3.

Results of the growth curve model for predicting functional aging.

	Model 1			Model 2			Model 3
	Coefficient	Standard error	P value	Coefficient	standard error	P value	Coefficient	Standard error	P value
Fixed effects
Intercept
(Intercept)	0.784	0.044	<.001***	0.768	0.045	<.001***	0.771	0.046	<.001***
Cohort	−0.119	0.009	<.001***	−0.119	0.009	<.001***	−0.119	0.009	<.001***
Education (Ref = < High school)
High school diploma	−0.493	0.043	<.001***	−0.494	0.043	<.001***	−0.493	0.043	<.001***
Some college	−0.538	0.047	<.001***	−0.539	0.047	<.001***	−0.539	0.047	<.001***
College degree/more	−0.649	0.048	<.001***	−0.646	0.048	<.001***	−0.646	0.048	<.001***
Race/Ethnicity (Ref = NH White)
NH Black	0.284	0.047	<.001***	0.301	0.048	<.001***	0.301	0.048	<.001***
Hispanic	0.253	0.054	<.001***	0.242	0.056	<.001***	0.242	0.056	<.001***
Other	0.088	0.100	.377	0.049	0.104	.634	0.049	0.104	.634
Gender (Ref = male)	0.051	0.022	.021*	0.078	0.029	.007**	0.073	0.030	.015*
Disability (Ref = No)	0.494	0.048	<.001***	0.496	0.051	<.001***	0.468	0.066	<.001***
Disability × Cohort	−0.014	0.028	.603	−0.016	0.028	.562	−0.017	0.028	.529
Disability × Race
NH Black	0.029	0.100	.767	−0.109	0.132	.408	−0.110	0.132	.403
Hispanic	0.254	0.113	.024*	0.331	0.146	.023*	0.329	0.146	.024*
Other	0.414	0.184	.024*	0.626	0.242	.009**	0.629	0.242	.009**
Disability × Gender							0.044	0.068	.518
Linear slope
Age	0.785	0.068	<.001***	0.742	0.075	<.001***	0.742	0.075	<.001***
Cohort	−0.405	0.015	<.001***	−0.405	0.015	<.001***	−0.405	0.015	<.001***
Education (Ref= < High school)
High school diploma	−0.283	0.071	<.001***	−0.286	0.071	<.001***	−0.286	0.071	<.001***
Some college	−0.321	0.077	<.001***	−0.323	0.077	<.001***	−0.324	0.077	<.001***
College degree/more	−0.470	0.078	<.001***	−0.462	0.078	<.001***	−0.462	0.078	<.001***
Race/Ethnicity (Ref = NH White)
NH Black	0.090	0.073	.220	0.131	0.078	.094	0.131	0.078	.094
Hispanic	0.222	0.085	.008**	0.194	0.090	.032*	0.194	0.090	.032*
Other	0.114	0.152	.452	0.016	0.168	.920	0.016	0.168	.920
Gender (Ref = Male)				0.068	0.047	.148	0.068	0.047	.047*
Disability (Ref = No)	0.268	0.071	<.001***	0.273	0.082	<.001***	0.273	0.082	<.001***
Disability × cohort	−0.096	0.046	.037*	−0.101	0.046	.028*	−0.101	0.046	.029*
Disability × Race
Yes × NH Black				−0.347	0.214	.106	−0.346	0.214	.106
Yes × Hispanic				0.194	0.236	.409	0.195	0.236	.408
Yes × Other				0.527	0.392	.178	0.527	0.392	.178
Quadratic slope
Age2	0.381	0.025	<.001***	0.380	0.025	<.001***	0.380	0.025	<.001***
Cohort	−0.387	0.015	<.001***	−0.386	0.015	<.001***	−0.386	0.015	<.001***
Random effects (variance)
Intercept	1.248			1.247			1.247
Growth rate	2.727			2.721			2.721
Residual	0.853			0.854			0.854
Correlation between intercept and growth rate	0.76			0.76			0.76
Model fit
AIC	147 915			147 915			147 916
BIC	148 177			148 212			148 223
Deviance	147 855			147 847			147 846
Likelihood ratio test						.099			.14
Total number (individual)	7234			7234			7234
Total number (person-year)	46 551			46 551			46 551

P values are represented as:

^* P < .05.

^** P < .01.

^*** P < .001.

Abbreviations: AIC = Akaike Information Criterion; BIC = Bayes Information Criteria; NH = non-Hispanic; Ref = reference group.

Older cohorts exhibited greater differences in functional aging, with a higher baseline ADL/IADL score (coefficient = −0.119, P < .001) and a faster increase over time (coefficient = −0.405, P < .001). This indicates that older cohorts experienced more rapid progression of functional limitations compared to younger cohorts, reflecting a cohort effect on the rate of functional aging (Figure 2).

These differences were further compounded by disability status. While the interaction between cohort and disability status was not significant for the baseline intercept, it was significant for the linear slope (coefficient = −0.096, P < .05). The negative coefficient suggests that in older cohorts, PAwD experience an accelerated aging process compared to the general population, further widening the functional aging gap between these groups (Figure 2). This underscores the heightened vulnerability of older cohorts with disabilities, who face both faster functional decline and a greater overall burden of limitations.

Regarding gender, women had significantly higher ADL/IADL limitations (coefficient = 0.051, P = .021). However, the rate of aging did not significantly differ by gender. Further investigation into gender-specific interactions with education and race/ethnicity in the context of disability did not reveal significant differences (Table S3).

Racial disparities in functional aging trajectories were pronounced, particularly among PAwD (Figure S8). NH-Black individuals had significantly higher average scores than NH-Whites (coefficient = 0.284, P < .001), with Hispanic individuals also exhibiting higher scores (coefficient = 0.253, P < .001). These disparities were further exacerbated among PAwD, as Hispanic PAwD had even higher functional limitation scores than the general population (coefficient = 0.254, P = .024), and PAwD of other racial groups showed the largest gap (coefficient = 0.414, P = .024; Figure S8). This interaction underscores the compounded impact of disability on functional aging disparities, with PAwD from minority racial and ethnic groups facing the greatest challenges. However, differences in the rate of aging by race were not statistically significant.

The interaction between disability and education on functional aging was not significant. However, functional limitation scores decreased with higher educational attainment: high school graduates (coefficient = −0.493, P < .001), some college (coefficient = −0.538, P < .001), and college graduates (coefficient = −0.649, P < .001). Higher education was also associated with a slower rate of functional aging (coefficient = −0.470, P < .001). Additional interaction analyses are presented in Table S3.

Discussion

The objective of this study was to investigate and compare phenotypic and functional aging trajectories between PAwD and the general population, while also examining the social disparities inherent in these trajectories using a nationally representative longitudinal dataset of older adults in the United States. Two major findings were identified from the analyses. First, PAwD experience an unequal and accelerated aging trajectory compared to the general population. PAwD exhibit significantly higher scores in both phenotypic and functional aging from baseline, with these patterns persisting throughout the aging trajectory. Recent cohorts of PAwD demonstrate a more pronounced escalation in MDRS, highlighting the potential for accelerated phenotypic aging in younger generations. Additionally, older cohorts of PAwD exhibited a steeper decrease in ability to perform ADL/IADL over time, underscoring the accelerated functional aging they experience relative to the general population.

Second, social disparities are evident in both phenotypic and functional aging trajectories among PAwD. Women aging with disability and PAwD in non-White racial groups demonstrated more pronounced phenotypic aging trajectories. Similarly, Hispanic and non-White PAwD demonstrated more accelerated losses in ability to perform ADL/IADL (ie, functional aging) relative to their counterparts.

Unequal and accelerated aging trajectories of PAwD

Previous research has identified a heightened susceptibility to metabolic dysregulation among PWD, with prevalence rates ranging from approximately 31.5% to 43.1%,^12,25–27 although the definitions of disability and metabolic dysregulation vary across studies, complicating direct comparison. This present study demonstrates that MDRS in PAwD is significantly elevated compared to the general population, and that this pattern persists without convergence over time. Age, as an intrinsic factor influencing metabolic health, may further exacerbate these disparities. de Groot et al.²⁶ identified that a 10-year increase in age correlates with a 1.5-fold heightened risk of developing metabolic dysregulation. Similarly, Jeong and Yu¹² found the risk of metabolic syndrome is 2.14-fold higher in PWD over the age of 65 compared to PWD aged between 20- and 44-year old. Although these insights derive from cross-sectional studies, they underscore the escalating risk of metabolic dysregulation with advancing age. This study builds upon these findings by quantifying the rate of increase in MDRS, estimating a linear rise of 0.1 points every 10 years. This longitudinal analysis elucidates the progressive nature of phenotypic aging and provides a comprehensive perspective on how age impacts metabolic dysfunction in PAwD.

Furthermore, the findings of this study may suggest that PAwD experience both premature, defined as the early onset of aging-related signs and changes at a younger than expected age,²⁸ as well as accelerated aging which refers to a faster progression of the aging process, characterized by a steeper trajectory of age-related physical and cognitive decline over time.²⁹ PAwD not only exhibit higher average MDRS at baseline but also demonstrate more rapid increases over time compared to the general population, especially among recent cohorts. This finding aligns with previous research, which suggested that PAwD may experience premature, atypical, and accelerated aging.^30–32 Evidence indicates that the onset of aging symptoms in PAwD typically begins in the mid-40s, and the subsequent impact on physical deterioration results in secondary conditions.^30–32 A previous study using the same waves of HRS data on metabolic dysregulation in the general population found no age-related changes.³³ Given that their measurements included resting heart rate, high pulse pressure, and C-reactive protein in addition to MDRS, direct comparisons are limited, but this finding underscore the potential heterogeneity of phenotypic aging trajectories across different subpopulations. This study, using biometric measures, contributes to existing knowledge by identifying distinct and potentially accelerated patterns of metabolic dysregulation among PAwD. By leveraging biometric indicators, these findings offer evidence of physiological changes that may reflect underlying disruptions in homeostatic regulation by highlighting a unique pathway of phenotypic aging in this understudied population.

In terms of functional aging, the results of this study are consistent with previous research that PAwD report a significantly high prevalence of difficulties in ADL/IADL.³⁴ While such baseline disparities are understandable given the definition of disability based on work limitations, the observed acceleration in functional decline among PAwD underscores more than a static gap. It may reflect the long-term consequences of cumulative disadvantage that compounded over the course of life. These disadvantages are detrimental not only to midlife functioning but also to capacity for recovery and adaptations as PAwD grow older.³⁴ Together, this finding reinforce the critical need for early interventions and sustained policies to address the structural inequalities that accelerate decline among PAwD.

The interaction between disability and cohort on the rate of aging is also noteworthy. The observed reduction in the disparity in functional aging rates between PAwD and the general population in more recent cohorts could partially reflect changes in sociohistorical and environmental factors. For younger cohorts, advances in healthcare access, societal support, and adaptive technologies may have buffered the impact of aging on functional decline. For instance, sociohistorical developments such as the enactment of the Americans with Disabilities Act (ADA) in 1990 may have fostered the adoption of universal design principles and reduced physical barriers. These changes could have particularly benefited younger cohorts by providing greater accessibility and independence, thereby narrowing disparities in ADL/IADL limitations relative to the general population.

Importantly, cohort-specific environmental exposures may also help explain the observed differences in aging trajectories at the same chronological age. Younger cohorts, who experienced early life during periods of increasing exposure to metabolic risk factors such as sedentary lifestyles, processed foods, and environmental stressors, may exhibit faster phenotypic aging. Conversely, older cohorts show more pronounced functional decline, possibly due to the cumulative effects of life-course disadvantages and declining physical resilience over time. This distinction underscores the importance of viewing phenotypic and functional aging as interconnected but distinct processes influenced by cohort-specific environmental and social contexts. These findings highlight the need for cohort-specific interventions tailored to the unique challenges of each generation within the PAwD population. Preventive strategies targeting younger cohorts should focus on mitigating metabolic dysregulation through early identification and intervention, while older cohorts may benefit from rehabilitative programs that prioritize maintaining functional independence and mobility. Future research, particularly longitudinal and policy-focused studies, should explore how these socio-contextual factors shape health outcomes and aging trajectories among PAwD.

Social disparities in aging trajectories

This study identified disparities in aging by gender, race/ethnicity, and education attainment among PAwD. Consistent with prior research, the results reveal gender disparities, with women exhibiting higher metabolic dysregulation compared to men in older age.¹⁸ A particular noteworthy aspect is its in-depth exploration of the interaction between disability and gender. Although only marginally significant, this interaction highlights a nuanced dynamic: women aging with disability may experience a more pronounced disparity in phenotypic aging compared to women in the general population, a contrast more evident than that observed among men. This pattern implies that the metabolic dysregulation in women aging with disability is not merely additive but potentially synergistic, compounding their risk for adverse health outcomes. Prior studies indicate that women with disability are more vulnerable to metabolic syndrome compared to men with disability,¹² likely due to the cumulative burden of physiological dysregulation across multiple organ systems, particularly after menopause when the protective effects of estrogen diminish.^35,36 Additionally, physical activity is a critical factor in mitigating metabolic syndrome risk in women,³⁷ yet engagement in such activity may be more challenging for PAwD. The interplay between gender-specific physiological changes and structural barriers to physical activity may contribute to this disparity. Future research should investigate the multifactorial influences underlying this heightened vulnerability, such as hormonal changes, physiological alterations, socio-environmental factors, and the long-term disability. With an evidence-based, holistic, gender-responsive approach, healthcare can better address the unique challenges faced by women aging with disability.

Furthermore, this study reveals that PAwD in racial/ethnic minority groups (ie, Hispanic and other racial groups) experience more detrimental functional aging trajectories compared to NH-White in the general population. Persistent racial/ethnic disparities in the functional difficulty have been a long concern.^38–41 People at this intersectionality face disproportionate exposure to adverse social determinants of health, including economic instability, discrimination, and limited healthcare services.^42,43 Researchers have identified key barriers to accessing disability services among racial/ethnic minority PAwD, including a shortage of culturally competent healthcare providers and outreach personnel, language barriers, and limited awareness of policies and programs.⁴⁴ Tailored interventions can alleviate these challenges by incorporating linguistically and culturally appropriate health education and rehabilitation programs to enhance self-management. Additionally, peer support initiatives (eg, Promotora models) can improve navigation of services. At a systemic level, policies should strengthen insurance support, including home-based rehabilitation, virtual consultations, and assistive devices, to enable early detection of functional decline and promote long-term maintenance. Furthermore, improving resource dissemination and integrating cultural and disability competency training into healthcare curricula could enhance accessibility and care quality.

Across all groups, higher levels of education emerged as a protective factor for both phenotypic and functional aging trajectories, and were linked to both lower baseline values and a slower rate in functional aging. Notably, coefficients decreased progressively with increasing educational attainment, reflecting a stepwise pattern in trajectories. While this finding emphasizes the importance of education, it also highlights the reality that PAwD experience significant educational disparities that may constrain these protective benefits. Enhanced health literacy and more effective disability-related coping strategies are known mechanisms linking education to improved health outcomes in PWD.^4,5 Expanding opportunities for health literacy interventions, lifelong learning programs, and financial support for higher education may help alleviate these disparities. Above all, ensuring appropriate accommodations in educational institutions for PAwD is essential to fostering accessibility and sustaining long-term benefits.

Globally, an estimated 200 million people are aging with disability, and this number is expected to grow substantially.⁷ International initiatives, such as the Toronto Declaration on Bridging Knowledge, Policy and Practice in Aging and Disability, have acknowledged the challenges faced by PAwD.⁴⁵ Given these global demographic shifts, international collaboration remains essential. Cross-national comparative research may help identify shared challenges, set global priorities, and inform policy frameworks that promote equitable aging outcomes for PAwD across diverse socio-political contexts.

Limitations

This study has limitations that future research should address. First, it should be acknowledged that disability is a complex physiological and social construct, and any operationalization of this construct in a secondary dataset will be imperfect. We have chosen to define disability solely in terms of activity restrictions (limitations in paid work, work at home, or other activities), which is a fairly narrow and concrete definition. This approach does not account for the large range of accommodations and changes PAwD may make to function day to day, and does not address environmental factors (such as neighborhood and access barriers), social factors (such as caregiving networks) or psychological aspects (such as disability identity). All of those factors would influence the conceptualization of disability within a more comprehensive biopsychosocial model, and are not addressed here. When interpreting literature on aging and disability, it is vital to consider how “disability” is operationalized.

Related to this limitation, the lack of disability-specific data such as impairment severity and disability domain (eg, mobility vs cognitive) limits precision of interpretation. More detailed information could elucidate variations in the timing and extent of aging across subgroups, facilitate the identification of key determinants, and inform the optimal timing and targets for intervention. Similarly, although this study is grounded in the time and metrics of the aging model, it remains incomplete due to the unavailability of data on biological aging. Future studies should include biological, phenotypic, and chronological aging measures for joint trajectory analyses, enabling a deeper understanding of the impact of disability on aging.

Another limitation comes in our reliance on self-reported outcomes, which may be affected by factors such as recall bias or desirability, potentially compromising accuracy. Future studies should employ more objective data, including clinician-administrated measures, physiological assessments, and medical records, to complement these findings.

Additionally, the composition of the general population as a comparison group may present a limitation. While individuals in this group did not meet the criteria for inclusion as PAwD, it is possible that some were living with disability but did not quality under our definition (eg, due to later onset or shorter duration). As such, the comparison group may include a heterogeneous mix of people without disability and those aging into disability, which could attenuate observed differences between groups.

Finally, this study did not capture the full life course due to data limitations. Future work should link multiple datasets to examine early health changes in PWD and this may help capture detailed health changes in early adulthood following the transition of individuals with congenital disabilities or childhood disability onset, and their cumulative impact on premature or accelerated aging as they reach mid-adulthood.

Despite these limitations, this study highlights the distinct and accelerated aging trajectories of PAwD, emphasizing pronounced disparities in phenotypic and functional aging. The findings call for early preventive strategies to address metabolic dysregulation and sustained rehabilitative efforts to maintain physical function. Marginalized subgroups within PAwD face compounded vulnerabilities, necessitating targeted, disability-inclusive health promotion strategies. Future research should focus on developing and implementing age- and disability-specific guidelines to optimize health outcomes and mitigate aging-related declines.

Author contributions

S.P. contributed to study design, conceptualization, formal analysis, methodology, funding acquisition, original drafting, and manuscript review and editing. I.M. contributed to supervision, validation, and manuscript revision. L.B. contributed to supervision, validation, and manuscript review and editing. All authors reviewed and approved the final version of the manuscript.

Supplementary material

Supplementary material is available at The Journals of Gerontology, Series A: Biological Sciences and Medical Sciences online.

Funding

This research was supported by a research award from the Beta Eta at-Large Chapter of Sigma Theta Tau International Honor Society of Nursing (Sigma). The HRS data used in this research was supported by a cooperative agreement with the National Institute on Aging and the Social Security Administration. It was produced and distributed by the University of Michigan with funding from the National Institute on Aging (Grant Nos. NIA U01AG009740 and NIA R01AG073289).

Conflicts of interest

None declared.

Data availability

The data used in this study are publicly available and can be accessed through the HRS database (https://hrs.isr.umich.edu/).

Sponsor’s role

The sponsors had no involvement in the study design, data analysis, manuscript preparation, or submission decision.