Skip to main content
NCBI home page
BMC Medicine logoLink to BMC Medicine
. 2026 Jul 10;24:548. doi: 10.1186/s12916-026-05057-7

Functional decline in the final years of life: A multicohort study

Junjie Lin 1,2, Mika Kivimäki 3,4, Yue Zhang 1,2, Yaguan Zhou 1,2, Xiaolin Xu 1,2,5,✉
PMCID: PMC13637263  PMID: 42432626

Abstract

Background

Physical, cognitive, and psychological conditions are associated with an increased risk of mortality, yet few studies have characterized terminal declines in these functional domains preceding death. In this multicohort study, we examined both separate and joint trajectories of physical, cognitive, and psychological functioning during the final decade of life and assessed the role of socioeconomic status (SES).

Methods

We used data from three national cohort studies: the US Health and Retirement Study, the China Health and Retirement Longitudinal Study, and the Survey of Health, Ageing and Retirement in Europe. Physical, cognitive, and psychological functioning were assessed repeatedly using measures of activities of daily living, cognitive test performance, and depressive symptoms, respectively. Mortality information was obtained from proxy respondents or national death registries. Statistical analyses included linear mixed-effects models and group-based multi-trajectory models using a backward timescale.

Results

Among 79,305 participants (mean age 65.8 ± 10.3 years; 55.4% female), 12,135 (15.3%) died during the 10-year follow-up. Compared with survivors, decedents showed steeper and accelerated declines in physical (βdeath * time -0.354, 95% confidence interval [95% CI] -0.372 to -0.336) and cognitive (βdeath * time -0.077, 95% CI -0.092 to -0.063) functioning, and to a lesser extent in psychological functioning (βdeath * time -0.044, 95% CI -0.061 to -0.028). In joint trajectory analyses, decedents experienced varying degrees of decline in at least one functional domain, and rapid terminal decline in physical functioning was accompanied by low and worsening cognitive and psychological functioning. In addition, higher SES was associated with lower odds of membership in adverse co-decline patterns, particularly the trajectory characterized by rapidly declining physical functioning combined with low and worsening cognitive and psychological functioning (e.g., OR high SES 0.07, 95% CI 0.02 to 0.20).

Conclusions

During the final decade of life, functional loss in physical and cognitive domains was more common than decline in psychological functioning. Higher SES was less likely to be associated with more unfavorable multidomain co-decline patterns.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12916-026-05057-7.

Keywords: Terminal decline, Mortality, Activities of daily living, Psychological functioning, Cognitive functioning

Background

Advances in socioeconomic development and health have driven substantial gains in life expectancy [1]. However, a growing number of older adults are living long enough to experience physical impairment, psychological problems, and cognitive decline, imposing a substantial burden on health and social care systems [2, 3]. Perspectives on healthy longevity are mixed. The compression of morbidity hypothesis suggests that the onset of disability and chronic diseases is postponed until shortly before death [4], whereas the expansion of morbidity hypothesis argues that increasing life expectancy prolongs morbidity and worsens overall health in the final years of life [5]. Morbidity is a broad concept that extends beyond disease to include conditions that reduce health-related quality of life or increase disease burden [6, 7]. Increasingly, the compression of morbidity hypothesis has been extended from physical disability to cognitive and psychological health [8, 9], which are key components of healthy aging in older adults [10]. These hypotheses provide an important conceptual framework for considering the timing of late-life deterioration: a more nuanced understanding of health trajectories is needed to identify opportunities for timely and targeted management strategies that promote well-being and healthy longevity in later life and alleviate the burden of supportive care needs [10].

An accelerated deterioration in health-related functioning at the end of life—known as terminal decline—has been observed in physical, cognitive, and psychological functioning [11–13]. In the Whitehall II cohort study of British civil servants, pronounced deterioration in motor functioning and activities of daily living (ADL) was observed during the 10 years preceding death, particularly in the final 4 years [11], whereas a study of 853 US participants reported terminal decline in cognitive functioning over the last 8 years of life, with a marked acceleration during the final 3.5 years before death [12]. A recent study from the Rush Memory and Aging Project showed that terminal decline in psychological well-being can begin as early as 6–11 years before death [13].

Despite these advances, terminal decline is increasingly conceptualized as a multidimensional process rather than isolated domain-specific changes. For example, in the Longitudinal Aging Study Amsterdam, participants with poorer and more rapidly worsening functional limitations experienced faster declines in affective well-being [14]. However, presenting changes in one indicator according to the level or change pattern of another indicator may not fully capture multiple domains as parallel trajectories. This approach may obscure within-person multidomain heterogeneity and are less suited to identifying complex co-decline subgroups or factors associated with their membership. In addition, although socioeconomic inequalities have been documented in multidomain functional trajectories among older adults in general and in domain-specific terminal decline [15–17], it remains unclear whether these inequalities are more pronounced in terminal co-decline patterns. Identifying co-decline patterns and socioeconomic disparities could provide more informative evidence for risk stratification and for efforts to reduce health inequalities at the end of life [10, 18].

Therefore, we analyzed harmonized data from three national cohort studies involving 17 countries, with the following aims: (1) to delineate separate trajectories of physical, cognitive, and psychological functioning in the decade preceding death, compared with survivors over a comparable time period; (2) to identify the most common multidomain co-decline patterns across these functional domains in the decade preceding death; and (3) to examine the role of socioeconomic status (SES) in terminal functional decline.

Methods

Study design and participants

This population-based, cross-national study used data from three cohorts of middle-aged and older adults across 17 countries: the Health and Retirement Study (HRS), the China Health and Retirement Study (CHARLS), and the Survey of Health, Ageing, and Retirement in Europe (SHARE). These sister studies are part of the Global Aging, Health, and Policy programme and follow harmonized study protocols, enabling cross-regional comparisons [19]. All of these cohorts conduct surveys every two to three years and provide measures of physical, cognitive, and psychological functioning as well as information on mortality. Details of the three cohorts’ designs are provided elsewhere [20–22]. This study adheres to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines.

To minimize the period effect of physical, cognitive, and psychological functioning preceding death, and ensure comparability across surveys, we used data from five waves from each study (Additional file 1: Table S1): Waves 11–15 of the HRS (2012–2021), waves 1–5 of the CHARLS (2011–2020), and waves 5–9 of the SHARE (2013–2022). We included participants aged ≥ 45 years who provided baseline data on physical, cognitive, and psychological functioning, excluding those without mortality information during follow-up. Participants with measures of physical, cognitive, and psychological functioning in at least one wave were included in the analysis of trajectories preceding death (Analysis 1) (n = 79,305). A subpopulation of decedents with ≥ 3 repeated measures across these domains was further analyzed for the joint trajectories preceding death and the association between SES and trajectory membership (n = 2,985). Figure 1 displays the detailed study design.

Fig. 1.

Fig. 1

Open in a new tab

Study design and flow chart of the population selection process. A Diagram of the study design; (B) Population selection process. HRS, health and Retirement Study; CHARLS, China Health and Retirement Longitudinal Study; SHARE, Survey of Health, Ageing and Retirement in Europe

Trajectories of physical, cognitive, and psychological functioning at baseline and follow-up

In HRS and SHARE, physical functioning was assessed using a self-administered survey on difficulties with ADLs, including walking across a room, dressing, bathing, eating, getting in and out of bed, and using the toilet. In CHARLS, an additional ADL item on difficulty controlling urination and defecation was used instead of the activity of walking across a room, which was not available in this study. A summed ADL score was calculated to represent participants’ physical functioning, with higher scores denoting poorer functional status.

Cognitive functioning was assessed using cohort-specific cognitive tests, covering four domains: episodic memory (immediate and delayed word recall), orientation (date naming for month, day, year, and day of the week), executive function (serial 7’s subtraction in HRS and CHARLS, and animal verbal fluency in SHARE), and visuospatial ability (a picture-drawing test in CHARLS). Notably, beginning in 2018 (wave 4), CHARLS introduced modifications to the administration procedures and word lists of the episodic memory test to increase comparability across HRS-family studies. Therefore, we applied a weighted equipercentile equating method to adjust the word recall scores in 2018 and 2020 according to the percentile ranks observed in 2015 [23]. In SHARE, because of a small number of outliers in the animal verbal fluency test, scores greater than 45 were re-coded to 45 [24]. A summed score of cognitive tests was calculated to represent participants’ cognitive functioning, with higher scores indicating better performance.

Psychological functioning was assessed using cohort-specific scales of depressive symptoms: the 8-item version of the Center for Epidemiologic Studies Depression Scale (CESD-8) in HRS, the 10-item version of the CESD scale in CHARLS, and the 12-item EURO-D scale in SHARE. Previous work has suggested that these measures are highly correlated [25]. Higher scores on the depression scale indicate poorer psychological functioning.

To facilitate the interpretation of trajectories in physical, cognitive, and psychological functioning, the ADL and depressive scores were reverse-coded, ensuring consistency in the direction of scores across the three domains, with higher scores indicating poorer functioning. However, no clear evidence has demonstrated item-level interchangeability of ADL measures and full comparability of cognitive tests across these cohorts. These measurement differences are widely acknowledged in harmonized cross-cohort analyses [24, 26–28], and standardized approaches [29], such as within-cohort z-score transformation, are commonly used to reduce their influence [24, 27, 28]. Therefore, we further standardized the baseline distributions of the three functional domains into z-scores (i.e., mean = 0, standard deviation [SD] = 1), to improve comparability of scores across the three cohorts. In addition, scores from follow-up waves were standardized based on the baseline distribution to ensure comparability over time within each study [30]. Cohort-specific standardized z-scores were derived for each of the three functional domains, with higher scores indicating better functional performance. Further details of the measurements of physical, cognitive, and psychological functioning are provided in Additional file 1: Table S2.

Mortality

Death information for HRS participants was obtained from two primary sources: exit interviews conducted with a proxy respondent (preferably a spouse or close family member) for deceased participants, and the National Death Index. This method of death ascertainment has been validated, with more than 95% of all deaths confirmed by both sources [31, 32]. Death information for CHARLS and SHARE participants was obtained from wave files and end-of-life interviews [32]. In CHARLS, specific dates of death were available only for some of the deceased participants in waves 2 and 5. For those with an unknown date of death, the date was imputed as the midpoint between the last interview and the subsequent interview, which is considered an acceptable approach [33]. Further details on the ascertainment of death information are provided in Additional file 1: Table S2.

Covariates

Covariates assessed in this study included sociodemographic characteristics (age, sex, educational level, total household wealth, and marital status), lifestyle (current smoking status, alcohol consumption, and physical activity) [34–37], and the number of comorbidities (hypertension, diabetes, heart problem, stroke, chronic lung disease, arthritis, and cancer) at baseline [38]. These variables were included based on previous studies on terminal decline because baseline behavioral and health-related profiles may influence both functioning trajectories and mortality [11, 13]. Missing covariates were assigned to the “unknown” groups. Harmonized classifications of covariates are presented in Additional file 1: Table S2.

Statistical analysis

Baseline characteristics of the participants were summarized according to cohort, survival status, and identified trajectories. Continuous variables were summarized as mean (SD), and categorical variables were summarized as frequency (percentage). Differences between groups were compared using the analysis of variance or Chi-square test, as appropriate.

Separate trajectory analysis (Analysis 1)

Linear mixed-effects models with backward timescales were used to capture the separate trajectories of physical, cognitive, and psychological functioning during the follow-up period and the differences in these trajectories between survivors and decedents. This reverse-time approach aligns individuals on a common endpoint, allowing for the distinction between aging-related changes and the dying process, with “time 0” defined as the date of death for decedents or the end of follow-up for survivors [11]. The fixed effects included survival status, time, time2, and their interaction to account for potential linear and nonlinear trajectories. The coefficients (β) of survival status reflected the differences in physical, cognitive, and psychological functioning before time 0. The coefficients of time represented the overall annual rate of change during the follow-up, and the coefficients of the interaction term indicated the differences in these change rates between decedents and survivors (i.e., additional annual changes compared with survivors). Individuals were included as random effects in the models. Models were adjusted for age at time 0, sex, cohort study, educational level, total household wealth, marital status, current smoking status, alcohol consumption, physical activity, and the number of comorbidities.

Joint trajectory analysis (Analysis 2)

We further applied group-based multi-trajectory modeling (GBMTM) to identify the most common joint trajectories of physical, cognitive, and psychological functioning preceding death among decedents and to examine the association between SES and trajectory membership. GBMTM is an approach used to identify latent clusters of individuals who follow similar trajectories across multiple indicators of interest. The Bayesian Information Criterion (BIC close to zero), the average posterior probabilities of assignment (Ave pp > 0.7), and the sample sizes in each group were used to determine the optimal number and shapes of the trajectory groups [39]. Given that most middle-aged and older adults were retired, SES was assessed using two indicators: educational level and total household wealth, consistent with our previous research [40, 41]. Specifically, a composite SES score was calculated by summing the scores for three-category educational level and four-category total household wealth, and was then classified as low (score 0), lower-middle (scores 1 or 2), upper-middle (scores 3 or 4), and high (score 5). We also considered the 3 × 4 combinations of educational level and total household wealth. We then used multinomial logistic regression models to estimate odds ratios (ORs) and 95% confidence intervals (CIs) for the associations between SES and joint trajectory membership of physical, cognitive, and psychological functioning preceding death.

Subgroup and sensitivity analyses testing the robustness of the results

We conducted subgroup analyses to examine whether differences in functional trajectories varied by age (< 75 or ≥ 75 years at time 0) and sex. For Analysis 1, subgroup analyses were additionally conducted by SES. Several sensitivity analyses were conducted to assess the robustness of the results: (a) to ensure that the results were not biased by missing data, we imputed missing covariates using multiple imputation and generating five complete datasets, with results combined according to Rubin’s rules; (b) to obtain more stable and precise estimates of within-person trajectories, we restricted the sample to participants with ≥ 3 measurements of physical, cognitive, or psychological functioning (for separate trajectory analysis); and (c) to maintain sample representativeness consistent with the separate trajectory analysis and to mitigate potential selection bias, we conducted inverse probability weighting analysis (for joint trajectory analysis). Finally, we used random-effects models to pool the results across the three cohorts.

The “traj” command in Stata 17 MP was used for the GBMTM. All other analyses were performed using R software (version 4.5.1). A two-sided P value < 0.05 was considered statistically significant.

Results

Baseline characteristics of participants

A total of 79,305 participants were included to investigate trajectories of physical, cognitive, and psychological functioning during the ten years preceding death, of whom 8,715 (11.0%) were from the United States, 10,289 (13.0%) were from China, and 60,301 (76.0%) were from Europe (Table 1). Baseline characteristics of participants included or not included in the analysis are compared in Additional file 1: Table S3. Participants included in the analysis were older, had higher educational level and household wealth, and were more likely to be married, non-smokers, alcohol drinkers, physically active, and had fewer chronic conditions. Among the included participants, the mean (SD) age was 65.8 (10.3) years at baseline and 72.0 (10.3) years at time 0; 35,378 (44.6%) were male, and 12,135 (15.3%) had died by time 0 (Additional file 1: Table S4). Compared with survivors (n = 67,170, 84.7%), decedents (n = 12,135, 15.3%) were more likely to be older, male, less educated, have lower household wealth, be widowed, non-drinkers, physically inactive, and have more chronic diseases at baseline (Additional file 1: Table S4).

Table 1.

Characteristics of the study population by different cohort studies

Total
(n = 79305)		 HRS
(n = 8715)		 CHARLS
(n = 10289)		 SHARE
(n = 60301)
Age at baseline, mean (SD) 65.8 (10.3) 74.3 (7.8) 58.2 (9.0)  65.9 (9.9)
Age at time 0, mean (SD) 72.0 (10.3) 80.3 (7.6) 66.6 (8.7) 71.7 (10.2)
Sex, n (%)
Male 35,378 (44.6%) 3692 (42.4%) 5317 (51.7%) 26,369 (43.7%)
Female 43,927 (55.4%) 5023 (57.6%) 4972 (48.3%) 33,932 (56.3%)
Educational level, n (%)
Primary 34,024 (42.9%) 1661 (19.1%) 8980 (87.3%) 23,383 (38.8%)
Secondary 29,097 (36.7%) 5202 (59.7%) 1149 (11.2%) 22,746 (37.7%)
Tertiary 16,181 (20.4%) 1849 (21.2%) 160 (1.6%) 14,172 (23.5%)
Unknown 3 (0.0%) 3 (0.0%) 0 (0.0%) 0 (0.0%)
Total household wealth, n (%)
Q1 (lowest) 19,012 (24.0%) 2179 (25.0%) 1912 (18.6%) 14,921 (24.7%)
Q2 19,011 (24.0%) 2179 (25.0%) 1911 (18.6%) 14,921 (24.7%)
Q3 19,010 (24.0%) 2179 (25.0%) 1911 (18.6%) 14,920 (24.7%)
Q4 (highest) 19,009 (24.0%) 2178 (25.0%) 1911 (18.6%) 14,920 (24.7%)
Unknown 3263 (4.1%) 0 (0.0%) 2644 (25.7%) 619 (1.0%)
Marital status, n (%)
Married or partnered 60,421 (76.2%) 5323 (61.1%) 9211 (89.5%) 45,887 (76.1%)
Widowed 10,370 (13.1%) 2230 (25.6%) 883 (8.6%) 7257 (12.0%)
Separated or divorced or single 8430 (10.6%) 1156 (13.3%) 195 (1.9%) 7079 (11.7%)
Unknown 84 (0.1%) 6 (0.1%) 0 (0.0%) 78 (0.1%)
Current smoking status, n (%)
No 64,273 (81.0%) 7830 (89.8%) 6896 (67.0%) 49,547 (82.2%)
Yes 14,968 (18.9%) 825 (9.5%) 3393 (33.0%) 10,750 (17.8%)
Unknown 64 (0.1%) 60 (0.7%) 0 (0.0%) 4 (0.0%)
Alcohol consumption, n (%)
Less than weekly drinking 45,394 (57.2%) 5725 (65.7%) 7912 (76.9%) 31,757 (52.7%)
Weekly drinking or more 33,329 (42.0%) 2965 (34.0%) 1843 (17.9%) 28,521 (47.3%)
Unknown 582 (0.7%) 25 (0.3%) 534 (5.2%) 23 (0.0%)
Physical activity, n (%)
Physical inactive 9535 (12.0%) 1969 (22.6%) 1481 (14.4%) 6085 (10.1%)
Physical active 63,856 (80.5%) 6726 (77.2%) 2927 (28.4%) 54,203 (89.9%)
Unknown 5914 (7.5%) 20 (0.2%) 5881 (57.2%) 13 (0.0%)
Number of comorbidities, n (%)
0 23,293 (29.4%) 675 (7.7%) 4257 (41.4%) 18,361 (30.4%)
1 24,926 (31.4%) 1805 (20.7%) 3510 (34.1%) 19,611 (32.5%)
≥ 2 30,845 (38.9%) 6235 (71.5%) 2342 (22.8%) 22,268 (36.9%)
Unknown 241 (0.3%) 0 (0.0%) 180 (1.7%) 61 (0.1%)
Open in a new tab

Abbreviations: HRS, health and Retirement Study; CHARLS, China Health and Retirement Longitudinal Study; SHARE, Survey of Health, Ageing and Retirement in Europe

Separate trajectories of physical, cognitive, and psychological functioning

Figure 2 shows the separate trajectories of physical, cognitive, and psychological functioning by survival status from the backward timescale analyses. Terminal declines adjusted for age at time 0, sex, and cohort study are presented in Additional file 1: Table S5. After adjusting for other covariates, compared with survivors, decedents had lower physical (βdeath = 1.227, 95% CI = -1.262 to -1.193), cognitive (βdeath = -0.490, 95% CI = -0.519 to -0.461), and psychological (βdeath = -0.285, 95% CI = -0.317 to -0.254) functioning before time 0 (Table 2). Decedents also exhibited faster declines in physical (βdeath * time = -0.354, 95% CI = -0.372 to -0.336), cognitive (βdeath * time= -0.077, 95% CI = -0.092 to -0.063), and psychological (βdeath * time = -0.044, 95% CI = -0.061 to -0.028) functioning over time (Table 2). Moreover, the declining trends in physical (βdeath * time^2 = -0.025, 95% CI = -0.027 to -0.023), cognitive (βdeath * time^2 = -0.004, 95% CI = -0.006 to -0.002), and psychological (βdeath * time^2 = -0.002, 95% CI = -0.004 to 0.000) functioning accelerated during the ten years preceding the endpoint (Table 2). Cohort-specific results are presented in Table 2 and Additional file 1: Fig. S1, showing no significant differences in psychological functioning or its rate of change in CHARLS. Harmonized results from the meta-analysis are presented in Additional file 1: Fig. S2.

Fig. 2.

Fig. 2

Open in a new tab

Trajectories of (A) physical, (B) cognitive, and (C) psychological functioning by mortality status. Models were adjusted for age at time 0, sex, cohort study, educational level, total household wealth, marital status, current smoking status, alcohol consumption, physical activity, and the number of comorbidities at baseline

Table 2.

Differences in physical, cognitive, and psychological functioning between survivors (10 years preceding the end of follow-up) and decedents (10 years preceding death)

HRS
β (95% CI)		 CHARLS
β (95% CI)		 SHARE
β (95% CI)		 Pooled cohort
β (95% CI)
Physical functioning
Death -0.808 (-0.874, -0.743) -1.650 (-1.774, -1.527) -1.272 (-1.313, -1.232) -1.227 (-1.262, -1.193)
Time -0.043 (-0.053, -0.033) -0.072 (-0.083, -0.061) -0.047 (0.051, -0.043) -0.049 (-0.053, -0.046)
Time2 -0.002 (-0.003, 0.000) -0.004 (-0.005, -0.002) -0.003 (-0.003, -0.002) -0.002 (-0.003, -0.002)
Death × Time -0.220 (-0.252, -0.188) -0.502 (-0.570, -0.436) -0.373 (-0.394, -0.352) -0.354 (-0.372, -0.336)
Death × Time2 -0.016 (-0.019, -0.012) -0.035 (-0.043, -0.027) -0.026 (-0.029, -0.024) -0.025 (-0.027, -0.023)
Cognitive functioning
Death -0.577 (-0.636, -0.519) -0.286 (-0.385, -0.186) -0.414 (-0.450, -0.378) -0.490 (-0.519, -0.461)
Time -0.068 (-0.078, -0.059) -0.042 (-0.050, -0.033) -0.049 (-0.052, -0.046) -0.046 (-0.049, -0.043)
Time2 -0.003 (-0.004, -0.002) 0.000 (-0.001, 0.001) -0.003 (-0.003, -0.003) -0.002 (-0.002, -0.002)
Death × Time -0.093 (-0.122, -0.064) -0.088 (-0.138, -0.038) -0.041 (-0.059, -0.023) -0.077 (-0.092, -0.063)
Death × Time2 -0.004 (-0.007, 0.000) -0.008 (-0.013, -0.002) -0.001 (-0.003, 0.001) -0.004 (-0.006, -0.002)
Psychological functioning
Death -0.286 (-0.348, -0.223) 0.012 (-0.082, 0.106) -0.322 (-0.363, -0.282) -0.285 (-0.317, -0.254)
Time -0.042 (-0.052, -0.032) -0.071 (-0.079, -0.063) -0.034 (-0.039, -0.030) -0.042 (-0.045, -0.038)
Time2 -0.003 (-0.004, -0.002) -0.006 (-0.006, -0.005) -0.002 (-0.003, -0.002) -0.003 (-0.003, -0.003)
Death × Time -0.048 (-0.080, -0.016) 0.017 (-0.032, 0.066) -0.056 (-0.077, -0.035) -0.044 (-0.061, -0.028)
Death × Time2 -0.004 (-0.008, 0.000) 0.001 (-0.005, -0.007) -0.003 (-0.005, 0.000) -0.002 (-0.004, 0.000)
Open in a new tab

Models were adjusted for age at time 0, sex, cohort study, educational level, total household wealth, marital status, current smoking status, alcohol consumption, physical activity, and the number of comorbidities at baseline

Abbreviations: HRS, Health and Retirement Study; CHARLS, China Health and Retirement Longitudinal Study; SHARE, Survey of Health, Ageing and Retirement in Europe; CI, confidence interval

Subgroup analyses showed that the differences in physical (P-interaction < 0.001) and cognitive (P-interaction = 0.014) functioning before time 0 between decedents and survivors were more pronounced among females, whereas the difference in psychological functioning (P-interaction = 0.022) was more evident among males (Additional file 1: Table S6). The differences in physical functioning between decedents and survivors before time 0, as well as the rate and acceleration of change during the follow-up, were more pronounced among those aged ≥ 75 years and those with lower educational level; whereas, only before time 0, differences in cognitive and psychological functioning were more pronounced among those aged ≥ 75 and < 75 years, respectively (Additional file 1: Tables S7 and S8). Similar results were found across total household wealth groups (Additional file 1: Table S9). Sensitivity analyses supported the robustness of our findings (Additional file 1: Table S10).

Joint trajectories of physical, cognitive, and psychological functioning preceding death

A total of 2,985 decedents were included in the analysis of the joint trajectories of physical, cognitive, and psychological functioning preceding death. Decedents included in the joint trajectories analysis were younger, more likely to be men, had higher educational level and household wealth, and were more likely to be married, current smokers, alcohol drinkers, and physically active (Additional file 1: Table S11). Four distinct trajectory groups were identified (Fig. 3): “high but mildly declining physical, cognitive, and psychological functioning” (n = 1,331, 43.9%); “high and mildly declining physical and psychological functioning, combined with low and declining cognitive functioning” (n = 933, 31.3%); “high and declining physical functioning, but low and deteriorating cognitive and psychological functioning” (n = 408, 13.7%); and “rapidly declining physical functioning, combined with low and worsening cognitive and psychological functioning” (n = 333, 11.2%). The fitted parameters of the GBMTM are presented in Additional file 1: Table S12, and the baseline characteristics of the identified trajectory groups are shown in Additional file 1: Table S13.

Fig. 3.

Fig. 3

Open in a new tab

Four classes of joint trajectories of physical, cognitive, and psychological functioning in decedents. Standardized z-scores were derived from the activities of daily living score for physical functioning, the depressive symptom score for psychological functioning, and cognitive test scores for cognitive functioning, with higher values indicating better functioning across all domains

After adjustment for covariates, decedents with higher SES were less likely to experience unfavorable joint trajectories of physical, cognitive, and psychological functioning in their final years, particularly trajectories characterized by rapidly declining physical functioning combined with low and worsening cognitive and psychological functioning (OR lower−middle SES = 0.57, 95% CI = 0.39 to 0.82; OR upper−middle SES = 0.25, 95% CI = 0.16 to 0.39; OR high SES = 0.07, 95% CI = 0.02 to 0.20; and OR per SES score increase = 0.65, 95% CI = 0.58 to 0.72, Table 3). The educational gradient was stronger than that for total household wealth. For example, for the trajectory characterized by rapidly declining physical functioning combined with low and worsening cognitive and psychological functioning, the OR per educational category increase was 0.48 (95% CI = 0.38 to 0.60), whereas the OR per wealth quartile increase was 0.75 (95% CI = 0.65 to 0.86; Table 3). Most combinations of educational level and total household wealth showed similar associations with joint trajectories, with the lowest OR observed for this unfavorable trajectory among participants with tertiary education and the highest household wealth (OR = 0.06, 95% CI = 0.02 to 0.18, Additional file 1: Table S14). Cohort-specific results are shown in Additional file 1: Table S15, and harmonized meta-analysis results are shown in Additional file 1: Fig. S3. These associations were consistently observed across subpopulations stratified by sex and age (Additional file 1: Tables S16 and S17). Sensitivity analyses confirmed the robustness of our findings (Additional file 1: Table S18).

Table 3.

Association between socioeconomic status and joint trajectories of physical, cognitive, and psychological functioning in decedents

High and mildly declining physical, cognitive, and psychological functioning High and mildly declining physical and psychological functioning, combined with low and declining cognitive functioning High and declining physical functioning, but low and deteriorating cognitive and psychological functioning Rapidly declining physical functioning, combined with low and worsening cognitive and psychological functioning
Case/N OR (95% CI) Case/N OR (95% CI) Case/N OR (95% CI) Case/N OR (95% CI)
SES
Low 119/432 Ref. 162/432 Ref. 76/432 Ref. 75/432 Ref.
Lower-middle 544/1381 Ref. 448/1381 0.61 (0.46–0.81) 219/1381 0.67 (0.48–0.95) 170/1381 0.57 (0.39–0.82)
Upper-middle 470/865 Ref. 254/865 0.35 (0.26–0.48) 79/865 0.28 (0.19–0.42) 63/865 0.25 (0.16–0.39)
High 114/174 Ref. 34/174 0.18 (0.11–0.30) 22/174 0.34 (0.19–0.60) 4/174 0.07 (0.02–0.20)
Per score increase / Ref. / 0.73 (0.68–0.78) / 0.72 (0.66–0.79) / 0.65 (0.58–0.72)
Educational level
Primary 372/1254 Ref. 485/1254 Ref. 216/1254 Ref. 181/1254 Ref.
Secondary 597/1148 Ref. 319/1148 0.32 (0.26–0.41) 127/1148 0.28 (0.21–0.38) 105/1148 0.38 (0.27–0.53)
Tertiary 278/450 Ref. 94/450 0.18 (0.13–0.24) 52/450 0.30 (0.20–0.44) 26/450 0.23 (0.14–0.37)
Per category increase / Ref. / 0.42 (0.36–0.48) / 0.48 (0.40–0.59) / 0.48 (0.38–0.60)
Total household wealth / / /
Q1 (lowest) 313/828 Ref. 264/828 Ref. 129/828 Ref. 122/828 Ref.
Q2 311/792 Ref. 253/792 0.93 (0.72–1.20) 131/792 1.09 (0.80–1.49) 97/792 0.83 (0.59–1.16)
Q3 325/685 Ref. 224/685 0.86 (0.66–1.12) 79/685 0.70 (0.50-1.00) 57/685 0.51 (0.35–0.75)
Q4 (highest) 298/547 Ref. 157/547 0.79 (0.59–1.06) 56/547 0.67 (0.45–0.99) 36/547 0.46 (0.29–0.72)
Per quartile increase / Ref. / 0.93 (0.85–1.01) / 0.85 (0.76–0.96) / 0.75 (0.65–0.86)
Open in a new tab

All models were adjusted for age at time 0, sex, cohort study, marital status, current smoking status, alcohol consumption, physical activity, and the number of comorbidities at baseline; Models for educational level and total household wealth were mutually adjusted;

Bold type indicates statistical significance;

Abbreviations: SES, socioeconomic status; OR, odds ratio; CI, confidence interval

Abbreviations: ADL, activities of daily living

Abbreviations: ADL, activities of daily living

Discussion

Using longitudinal individual-level cohort data from 17 countries across North America, East Asia, and Europe, this study of repeated measures of physical, cognitive, and psychological functioning spanning the 10 years preceding death, yields three main findings. First, compared with survivors, decedents experienced a steeper and accelerated deterioration in physical, cognitive, and psychological functioning during the final 10 years of life, with the most pronounced loss occurring in ADL. Second, decedents experienced varying degrees of decline in at least one functional domain, and rapid terminal decline in physical functioning was commonly accompanied by low and worsening cognitive and psychological functioning. Compared with the trajectory characterized by high functioning with mild co-decline across the three functional domains, higher SES was associated with a lower likelihood of experiencing more unfavorable co-decline patterns, particularly the trajectory characterized by rapidly declining physical functioning combined with low and worsening cognitive and psychological functioning.

Our findings align with and extend previous research. In the Whitehall II study, decedents showed a faster decline in ADL than survivors during the final decade of life, with greater functional loss occurring during the last four years [11]. Similarly, results from the Taiwan Longitudinal Study on Aging suggested accelerated deterioration in ADL function as death approached [42]. For psychological functioning, depressive symptoms increased nonlinearly during the last 9 years of life in the Rush Memory and Aging Project, with trajectories differing from those of survivors [13]. In contrast, studies from Sweden, Denmark, and Australia have documented marked declines in psychological functioning in the final four years of life, a pattern not observed in our data [43]. Such differences may partly reflect the use of piecewise modeling that accentuates post-inflection declines [13]. Terminal decline in cognitive function is well established [44]. Findings from the United States mirror our observations, demonstrating pronounced deterioration in the years preceding death [45].

Using a backward timescale, we identified four distinct physical, cognitive, and psychological co-decline patterns among decedents. Compared with survivors, decedents showed both poorer baseline functioning and marked deterioration in at least one of these domains, consistent with previous findings on functional trajectories and mortality in European cohorts [46–48]. Although a previous study examined terminal changes in affective well-being according to the level or change pattern of physical health indicators, it was not designed to identify complex multidomain co-decline patterns while accounting for individual heterogeneity [14]. Consistent with our findings, evidence from the US HRS, using multichannel sequence analysis, identified co-pathways of physical and psychological functioning toward death and showed that lower educational level and lower income were associated with lower and worsening physical and psychological functioning [49]. Our study suggests that socioeconomic inequalities persisted into the terminal phase of life, particularly for multidomain functional decline. These findings highlight the multidimensional nature of end-of-life health inequalities and underscore the need for targeted strategies to identify vulnerable older adults and reduce disparities in terminal functional health [10].

Some subgroup differences are noteworthy. During the final decade of life, women showed greater disability than men, and older decedents exhibited a faster progression of physical disability than younger decedents–patterns consistent with findings reported among older adults in the United States [16, 50]. In contrast, men and younger decedents tended to have poorer psychological functioning in the terminal period. Previous studies similarly suggest that men experience a greater death-related psychological burden and that those who die at older ages tend to have better well-being near the end of life [43, 51]. Unlike physical and psychological functioning, terminal decline in cognitive functioning did not vary by age or sex. In addition, the association between SES and multidomain functional decline was independent of age and sex.

Some results should be interpreted with caution. Among 1,016 decedents from CHARLS, representing 85.6% of CHARLS decedents and 8.4% of all decedents, only the survey interval during which death occurred was available. The date of death was therefore imputed as the midpoint between the dates of the two adjacent follow-up waves. Given the accelerated decline in functioning near the end of life, observations reflecting steeper decline closer to death may have been shifted to earlier reverse-time points, whereas higher-functioning observations farther from death may have been shifted closer to death. This error structure would tend to smooth the terminal decline trajectory and attenuate the estimated death-by-time interaction. In addition, differences in country-specific follow-up structures within SHARE contributed to its higher exclusion rate in the joint trajectory analysis [52]. Such non-random attrition may have affected the marginal distribution of trajectory groups and the generalizability of the estimated class proportions, particularly because SHARE participants were more likely to have poorer joint functional trajectories [53]. Therefore, the estimated trajectory group proportions should be interpreted as distributions among analysis-eligible decedents rather than prevalence estimates for the original cohorts or the general population of decedents. Although the association between SES and trajectory membership was not materially changed after applying inverse probability weighting, residual selection bias cannot be fully excluded.

Several mechanisms may explain our findings. First, during the end-of-life stage, immune function typically becomes dysregulated, with excess proinflammatory cytokines causing systemic damage to muscles, the cardiovascular system, and neurons, thereby accelerating declines in physical and cognitive functioning and impairing psychological well-being [54]. Second, in both the separate and joint trajectory analyses, we observed a more pronounced terminal decline in physical functioning. One possible explanation is that self-reported ADL reflects not only physical capacity but also psychosocial and other health-related components [55]. For example, results from the Whitehall II study showed that, compared with survivors, decedents experienced a steeper decline in self-reported physical functioning in the final years of life, whereas declines in objective physical functioning were less pronounced [11]. Third, social orientation and cultural differences may shape trajectories of psychological decline. For example, evidence suggests that older Chinese adults may express depressive symptoms through physical complaints, which could partly account for the less pronounced decline in psychological function and the steeper terminal decline in physical function in CHARLS, compared with US HRS and European SHARE [51, 56].

This study addresses several research priorities highlighted in prior studies and has important theoretical and practical implications [18, 44]. Although life expectancy has increased, longer survival does not necessarily imply uniform preservation of functioning across all health domains. While it is impossible to predict when an individual will begin the “countdown” to the end of life, most people will inevitably experience terminal decline and require comprehensive care [1]. For policymakers, efforts should be made to integrate ADL, depression, and cognitive screening into essential public health services and medical insurance coverage, as recommended in the World Health Organization’s Integrated Care for Older People guidelines, making medical care more accessible and affordable for the oldest-old [10]. Healthcare providers should provide integrated care to ensure that health services can effectively meet the diverse needs of older adults, promote the adoption of a comprehensive, person-centered approach, and enhance their physical, mental, and social well-being [1, 10]. Particular attention should be paid to socioeconomically disadvantaged older adults, who may face greater barriers to timely detection and management of terminal functional deterioration, including lower health literacy, limited affordability and access to rehabilitation services [57, 58]. These barriers may delay the recognition and management of ADL limitations, cognitive impairment, and depressive symptoms, thereby increasing the likelihood that decline in one domain progresses into multidimensional co-decline.

Using the terminal decline framework provides a clear understanding of how declines unfold across domains by capturing within-person changes [11, 13, 43]. Joint trajectory modeling further distinguishes more adverse co-decline patterns and provides additional insight into the role of socioeconomic inequality. However, several limitations should also be noted. First, physical and psychological functioning were self-reported, which may have introduced bias. Second, the two- to three-year follow-up intervals limited our ability to identify finer-grained declines, particularly in the final months of life. Third, between-cohort differences (e.g., mean age, the instruments used to assess functioning, and exclusion percentages) should be taken into account. Although we used cohort-specific z-scores to improve cross-cohort comparability, adjusted for both age and cohort, and additionally conducted a meta-analysis, cohort-specific interpretations may still be more appropriate in certain contexts. Fourth, the midpoint imputation of death dates for some participants in CHARLS may introduce additional measurement noise. Fifth, the long follow-up period introduced loss-to-follow-up bias, and survival status after the final follow-up was unknown for survivors, possibly leading to misclassification of deaths occurring shortly thereafter. Sixth, we could not account for acute events. Last, because cause of death data were unavailable, we could not examine terminal decline by specific causes. Future research could elucidate the interrelationships among terminal declines across different functional domains and further identify main drivers of co-decline patterns.

Conclusions

In summary, physical, cognitive, and psychological functioning showed both independent and concurrent declines during the final decade of life. In addition, socioeconomic inequalities were observed in multidomain co-decline patterns. Early screening to identify overall functional impairment, and strategies to reduce accelerated functional decline, should start before old age. Identifying co-decline patterns, particularly among socioeconomically disadvantaged groups, may help inform targeted strategies to promote healthy longevity and reduce end-of-life health inequalities.

Supplementary Information

Below is the link to the electronic supplementary material.

12916_2026_5057_MOESM1_ESM.docx (1.2MB, docx)

Supplementary Material 1: Additional file 1. Tables S1-S14, Figures S1-S3

Acknowledgements

We thank all team members who contributed to the data collection and data cleaning of HRS, CHARLS, and SHARE. We thank all the volunteers who participated in the study. This analysis used data or information from the harmonized HRS, CHARLS, and SHARE datasets and codebooks developed by the Gateway to Global Aging Data and supported by the National Institute on Aging.

Abbreviations

ADL

Activities of daily living

CESD

Center for Epidemiologic Studies Depression Scale

CHARLS

China Health and Retirement Study

CI

Confidence interval

GBMTM

Group-based multi-trajectory modeling

HR

Hazard ratio

HRS

Health and Retirement Study

OR

Odds ratio

SD

Standard deviation

SES

Socioeconomic status

SHARE

Survey of Health, Ageing, and Retirement in Europe

Author contributions

XX contributed the study conceptualization and supervised the whole project. JL made the analysis plan and conducted the statistical analyses. JL, YZ1, YZ2 and XX validated data and statistical analyses. JL contributed to visualization and wrote the initial draft of the manuscript. MK contributed the study conceptualization and provided valuable revision advice on writing and statistical methods. The corresponding author attests that all listed authors meet authorship criteria and that no others meeting the criteria have been omitted. XX and JL have accessed and verified the data, and XX were responsible for the decision to submit the manuscript. All authors read and approved the final manuscript.

Funding

XX was supported by Zhejiang University, the Fundamental Research Funds for the Central Universities, the Key Laboratory of Intelligent Preventive Medicine of Zhejiang Province, the China Medical Board (No. 21–416), and the National Natural Science Foundation of China (No. 72474197). MK was supported by the Wellcome Trust (221854/Z/20/Z), the UK Medical Research Council (MR/Y014154/1), and the Research Council of Finland (350426).

Data availability

The original data for this study are available on their respective websites: the Health and Retirement Study (HRS; https://hrs.isr.umich.edu), the China Health and Retirement Longitudinal Study (CHARLS; https://charls.pku.edu.cn/en/), and the Survey of Health, Ageing and Retirement in Europe (SHARE; https://share-eric.eu).

Declarations

Consent for publication

Not applicable. This study used de-identified data from the publicly available database, which contains no personally identifiable information. Therefore, individual patient consent for publication was not required.

Competing interests

The authors declare no competing interests.

Ethics approval and informed consent

The HRS was approved by the institute for social research and Survey Research Center at the University of Michigan (NIA U01AG009740). The CHARLS was approved by the Ethical Review Committee at Peking University (IRB00001052-11015). The SHARE was approved by the Ethics Committee of the University of Mannheim.

Footnotes

Publisher’s note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

References

  • 1.World Health Organization. Ageing: Global population. 2025. https://www.who.int/news-room/questions-and-answers/item/population-ageing
  • 2.Christensen K, Doblhammer G, Rau R, Vaupel JW. Ageing populations: the challenges ahead. Lancet. 2009;374:1196–208. 10.1016/S0140-6736(09)61460-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3.Maresova P, Javanmardi E, Barakovic S, Barakovic Husic J, Tomsone S, Krejcar O, et al. Consequences of chronic diseases and other limitations associated with old age – a scoping review. BMC Public Health. 2019;19:1431. 10.1186/s12889-019-7762-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4.Fries JF, Aging. Natural Death, and the Compression of Morbidity. N Engl J Med. 1980;303:130–5. 10.1056/NEJM198007173030304. [DOI] [PubMed] [Google Scholar]
  • 5.Gruenberg EM. The Failures of Success. Milbank Mem Fund Q Health Soc. 1977;55:3. 10.2307/3349592. [DOI] [PubMed] [Google Scholar]
  • 6.Fries JF, Bruce B, Chakravarty E. Compression of Morbidity 1980–2011: A Focused Review of Paradigms and Progress. J Aging Res. 2011;2011:1–10. 10.4061/2011/261702. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7.Beltrán-Sánchez H, Razak F, Subramanian SV. Going beyond the disability-based morbidity definition in the compression of morbidity framework. Glob Health Action. 2014;7:24766. 10.3402/gha.v7.24766. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Langa KM, Larson EB, Karlawish JH, Cutler DM, Kabeto MU, Kim SY, et al. Trends in the prevalence and mortality of cognitive impairment in the United States: Is there evidence of a compression of cognitive morbidity? Alzheimers Dement. 2008;4:134–44. 10.1016/j.jalz.2008.01.001. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Beller J, Regidor E, Lostao L, Miething A, Kröger C, Safieddine B, et al. Decline of depressive symptoms in Europe: differential trends across the lifespan. Soc Psychiatry Psychiatr Epidemiol. 2021;56:1249–62. 10.1007/s00127-020-01979-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.World Health Organization. Integrated care for older people (ICOPE): guidance for person-centred assessment and pathways in primary care, 2nd ed. 2024. https://www.who.int/publications/i/item/9789240103726
  • 11.Landré B, Fayosse A, Ben Hassen C, Machado-Fragua MD, Dumurgier J, Kivimaki M, et al. Terminal decline in objective and self-reported measures of motor function before death: 10 year follow-up of Whitehall II cohort study. BMJ. 2021;n1743. 10.1136/bmj.n1743. [DOI] [PMC free article] [PubMed]
  • 12.Wilson RS, Beck TL, Bienias JL, Bennett DA. Terminal Cognitive Decline: Accelerated Loss of Cognition in the Last Years of Life. Psychosom Med. 2007;69:131–7. 10.1097/PSY.0b013e31803130ae. [DOI] [PubMed] [Google Scholar]
  • 13.Wang J, Guo J, Dove A, Zhang X, Yue J, Bennett DA, et al. Trajectories of psychological and social well-being preceding death. BMJ Ment Health. 2025;28:e301967. 10.1136/bmjment-2025-301967. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Schilling OK, Deeg DJH, Huisman M. Affective well-being in the last years of life: The role of health decline. Psychol Aging. 2018;33:739–53. 10.1037/pag0000279. [DOI] [PubMed] [Google Scholar]
  • 15.Xu X, Liang J, Bennett JM, Botoseneanu A, Allore HG. Socioeconomic Stratification and Multidimensional Health Trajectories: Evidence of Convergence in Later Old Age. J Gerontol B Psychol Sci Soc Sci. 2015;70:661–71. 10.1093/geronb/gbu095. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.Smith AK, Walter LC, Miao Y, Boscardin WJ, Covinsky KE. Disability During the Last Two Years of Life. JAMA Intern Med. 2013;173:1506. 10.1001/jamainternmed.2013.8738. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Terrera GM, Minett T, Brayne C, Matthews FE. Education associated with a delayed onset of terminal decline. Age Ageing. 2014;43:26–31. 10.1093/ageing/aft150. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Gerstorf D, Ram N. Inquiry Into Terminal Decline: Five Objectives for Future Study. Gerontologist. 2013;53:727–37. 10.1093/geront/gnt046. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Lee J, Phillips D, Wilkens J, Gateway to Global Aging Data Team, Aging H. J Gerontol Ser B. 2021;76 Supplement1:S5–16. 10.1093/geronb/gbab050. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Sonnega A, Faul JD, Ofstedal MB, Langa KM, Phillips JWR, Weir DR. Cohort Profile: the Health and Retirement Study (HRS). Int J Epidemiol. 2014;43:576–85. 10.1093/ije/dyu067. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21.Zhao Y, Hu Y, Smith JP, Strauss J, Yang G. Cohort profile: the China Health and Retirement Longitudinal Study (CHARLS). Int J Epidemiol. 2014;43:61–8. 10.1093/ije/dys203. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Börsch-Supan A, Brandt M, Hunkler C, Kneip T, Korbmacher J, Malter F, et al. Data Resource Profile: the Survey of Health, Ageing and Retirement in Europe (SHARE). Int J Epidemiol. 2013;42:992–1001. 10.1093/ije/dyt088. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Wu Y, Zhang YS, Kobayashi LC, Mayeda ER, Gross AL. How to assess cognitive decline when test administration changes across study waves? Harmonizing cognitive scores across waves in the China Health and Retirement Longitudinal Study. J Alzheimers Dis Rep. 2024;8:1661–9. 10.1177/25424823241302759. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Huang Y, Chen H, Gao M, Lv X, Pang T, Rong S, et al. Self- and interviewer-reported cognitive problems in relation to cognitive decline and dementia: results from two prospective studies. BMC Med. 2024;22:23. 10.1186/s12916-023-03147-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.Courtin E, Knapp M, Grundy E, Avendano-Pabon M. Are different measures of depressive symptoms in old age comparable? An analysis of the CES-D and Euro-D scales in 13 countries. Int J Methods Psychiatr Res. 2015;24:287–304. 10.1002/mpr.1489. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26.Lu X, Yao Y, Jin Y. Digital exclusion and functional dependence in older people: Findings from five longitudinal cohort studies. eClinicalMedicine. 2022;54:101708. 10.1016/j.eclinm.2022.101708. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27.Zhang Y, Zhou Y, Kivimäki M, Holt-Lunstad J, Carrillo-Larco RM, Xu X. Cardiometabolic multimorbidity, social activity, and joint trajectories of physical disability, depressive symptom, and cognitive function in mid-to-late life: a multicohort study. BMC Med. 2025;23:577. 10.1186/s12916-025-04400-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28.Luo S, Chen W, Hu W, Wang HHX, Li J, Guo VY, et al. Parental Education, Own Education, and Cognitive Function in Middle-Aged and Older Adults. JAMA Netw Open. 2025;8:e2513036. 10.1001/jamanetworkopen.2025.13036. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.Fortier I, Raina P, Van Den Heuvel ER, Griffith LE, Craig C, Saliba M, et al. Maelstrom Research guidelines for rigorous retrospective data harmonization. Int J Epidemiol. 2016;dyw075. 10.1093/ije/dyw075. [DOI] [PMC free article] [PubMed]
  • 30.Guo M, Wu Y, Gross AL, Karvonen-Gutierrez C, Kobayashi LC. Age at menopause and cognitive function and decline among middle-aged and older women in the China Health and Retirement Longitudinal Study, 2011–2018. Alzheimers Dement J Alzheimers Assoc. 2025;21:e14580. 10.1002/alz.14580. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31.Weir DR. VALIDATING MORTALITY ASCERTAINMENT IN THE HEALTH AND RETIREMENT STUDY. 2016.
  • 32.Machado S, Kyriopoulos I, Orav EJ, Papanicolas I. Association between Wealth and Mortality in the United States and Europe. N Engl J Med. 2025;392:1310–9. 10.1056/NEJMsa2408259. [DOI] [PubMed] [Google Scholar]
  • 33.Murayama H, Liang J, Bennett JM, Shaw BA, Botoseneanu A, Kobayashi E, et al. Trajectories of Body Mass Index and Their Associations With Mortality Among Older Japanese: Do They Differ From Those of Western Populations? Am J Epidemiol. 2015;182:597–605. 10.1093/aje/kwv107. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34.Cunningham C, O’ Sullivan R, Caserotti P, Tully MA. Consequences of physical inactivity in older adults: A systematic review of reviews and meta-analyses. Scand J Med Sci Sports. 2020;30:816–27. 10.1111/sms.13616. [DOI] [PubMed] [Google Scholar]
  • 35.Jia J, Zhao T, Liu Z, Liang Y, Li F, Li Y, et al. Association between healthy lifestyle and memory decline in older adults: 10 year, population based, prospective cohort study. BMJ. 2023;380:e072691. 10.1136/bmj-2022-072691. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36.Xu C, Cao Z, Huang X, Wang X. Associations of healthy lifestyle with depression and post-depression dementia: A prospective cohort study. J Affect Disord. 2023;327:87–92. 10.1016/j.jad.2023.01.111. [DOI] [PubMed] [Google Scholar]
  • 37.Zhang Y-B, Chen C, Pan X-F, Guo J, Li Y, Franco OH, et al. Associations of healthy lifestyle and socioeconomic status with mortality and incident cardiovascular disease: two prospective cohort studies. BMJ. 2021;n604. 10.1136/bmj.n604. [DOI] [PMC free article] [PubMed]
  • 38.Kadambi S, Abdallah M, Loh KP. Multimorbidity, Function, and Cognition in Aging. Clin Geriatr Med. 2020;36:569–84. 10.1016/j.cger.2020.06.002. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39.Nagin DS, Jones BL, Passos VL, Tremblay RE. Group-based multi-trajectory modeling. Stat Methods Med Res. 2018;27:2015–23. 10.1177/0962280216673085. [DOI] [PubMed] [Google Scholar]
  • 40.Wang D, Dai X, Mishra SR, Lim CCW, Carrillo-Larco RM, Gakidou E, et al. Association between socioeconomic status and health behaviour change before and after non-communicable disease diagnoses: a multicohort study. Lancet Public Health. 2022;7:e670–82. 10.1016/S2468-2667(22)00157-8. [DOI] [PubMed] [Google Scholar]
  • 41.Ni Y, Zhou Y, Kivimäki M, Cai Y, Carrillo-Larco RM, Xu X, et al. Socioeconomic inequalities in physical, psychological, and cognitive multimorbidity in middle-aged and older adults in 33 countries: a cross-sectional study. Lancet Healthy Longev. 2023;4:e618–28. 10.1016/S2666-7568(23)00195-2. [DOI] [PubMed] [Google Scholar]
  • 42.Chiu C-J, Li M-L, Chang C-M, Wu C-H, Tan MP. Disability trajectories prior to death for ten leading causes of death among middle-aged and older adults in Taiwan. BMC Geriatr. 2021;21:420. 10.1186/s12877-021-02300-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 43.Petkus AJ, Reynolds CA, Catts VS, Christensen K, Finkel D, Nygaard M, et al. Terminal Increases in Depressive Symptoms in a Multinational Twin Consortium. Psychol Sci. 2025;36:637–55. 10.1177/09567976251351022. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 44.Cohen-Mansfield J, Skornick-Bouchbinder M, Brill S. Trajectories of End of Life: A Systematic Review. J Gerontol Ser B. 2018;73:564–72. 10.1093/geronb/gbx093. [DOI] [PubMed] [Google Scholar]
  • 45.Wilson RS, Yu L, Leurgans SE, Bennett DA, Boyle PA. Proportion of cognitive loss attributable to terminal decline. Neurology. 2020;94. 10.1212/WNL.0000000000008671. [DOI] [PMC free article] [PubMed]
  • 46.Gardeniers MKM, Van Groenou MIB, Meijboom EJ, Huisman M. Three-year trajectories in functional limitations and cognitive decline among Dutch 75 + year olds, using nine-month intervals. BMC Geriatr. 2022;22:89. 10.1186/s12877-021-02720-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 47.Saeed Mirza S, Ikram MA, Freak-Poli R, Hofman A, Rizopoulos D, Tiemeier H. 12 Year Trajectories of Depressive Symptoms in Community-Dwelling Older Adults and the Subsequent Risk of Death Over 13 Years. J Gerontol Ser A. 2018;73:820–7. 10.1093/gerona/glx215. [DOI] [PubMed] [Google Scholar]
  • 48.Lobo E, De La Cámara C, Gracia-García P, Saz P, López-Antón R, Lobo A. Cognitive trajectories in older adults and associated mortality and predictors. Soc Psychiatry Psychiatr Epidemiol. 2025. 10.1007/s00127-025-02862-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 49.Raab M, Fasang A, Hess M. Pathways to death: The co-occurrence of physical and mental health in the last years of life. Demogr Res. 2018;38:1619–34. 10.4054/DemRes.2018.38.53. [DOI] [Google Scholar]
  • 50.Guralnik JM, LaCroix AZ, Branch LG, Kasl SV, Wallace RB. Morbidity and disability in older persons in the years prior to death. Am J Public Health. 1991;81:443–7. 10.2105/ajph.81.4.443. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 51.Gerstorf D, Hoppmann CA, Löckenhoff CE, Infurna FJ, Schupp J, Wagner GG, et al. Terminal decline in well-being: The role of social orientation. Psychol Aging. 2016;31:149–65. 10.1037/pag0000072. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 52.SHARE Waves overview. https://share-eric.eu/data/data-documentation/waves-overview/. Accessed 9 May 2026.
  • 53.Haviland AM, Jones BL, Nagin DS. Group-based Trajectory Modeling Extended to Account for Nonrandom Participant Attrition. Sociol Methods Res. 2011;40:367–90. 10.1177/0049124111400041. [DOI] [Google Scholar]
  • 54.Taylor JA, Greenhaff PL, Bartlett DB, Jackson TA, Duggal NA, Lord JM. Multisystem physiological perspective of human frailty and its modulation by physical activity. Physiol Rev. 2023;103:1137–91. 10.1152/physrev.00037.2021. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 55.Kok AAL, Henstra MJ, van der Velde N, Rhebergen D, van Schoor NM. Psychosocial and Health-Related Factors Associated With Discordance Between 13-Year Trajectories of Self-Reported Functional Limitations and Performance-Based Physical Functioning in Old Age. J Aging Health. 2020;32:1084–97. 10.1177/0898264319884404. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 56.Krause N, Liang J. Second Place (tie) 1991IPA Research Awards in Psychogeriatrics. Int Psychogeriatr. 1992;4:185–202. 10.1017/S1041610292001261. [DOI] [Google Scholar]
  • 57.McMaughan DJ, Oloruntoba O, Smith ML. Socioeconomic Status and Access to Healthcare: Interrelated Drivers for Healthy Aging. Front Public Health. 2020;8:231. 10.3389/fpubh.2020.00231. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 58.Ma T, Meng H, Ye Z, Jia C, Sun M, Liu D. Health Literacy Mediates the Association Between Socioeconomic Status and Productive Aging Among Elderly Chinese Adults in a Newly Urbanized Community. Front Public Health. 2021;9:647230. 10.3389/fpubh.2021.647230. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

12916_2026_5057_MOESM1_ESM.docx (1.2MB, docx)

Supplementary Material 1: Additional file 1. Tables S1-S14, Figures S1-S3

Data Availability Statement

The original data for this study are available on their respective websites: the Health and Retirement Study (HRS; https://hrs.isr.umich.edu), the China Health and Retirement Longitudinal Study (CHARLS; https://charls.pku.edu.cn/en/), and the Survey of Health, Ageing and Retirement in Europe (SHARE; https://share-eric.eu).

Articles from BMC Medicine are provided here courtesy of BMC

RESOURCES