Abstract
Background
Cognitive impairment (CI) and physical disability (PD) frequently co-occur in older adults and are both significant risk factors for depressive symptoms. However, their combined impact on the risk and reversibility across different depressive symptom states remains unclear.
Methods
Utilizing data from five waves (2011–2020) of the China Health and Retirement Longitudinal Study (CHARLS), we included 8,473 adults aged 60 and above. A continuous-time multi-state Markov model was used to analyze transition probabilities and intensities among states of no (NDS), mild (MDS), and severe depressive symptoms (SDS), and to examine the effects of CI and PD on these transitions. Cox proportional hazards models were employed to estimate hazard ratios (HRs) for the associations of CI and PD with the risk and reversibility of depressive symptoms. Restricted cubic splines (RCS) were further applied to assess dose–response relationships.
Results
Among 8,473 older participants (mean age 67.91 ± 6.59 years; 49.3% female), 14.0% had CI, 17.5% had PD, and 6.5% had both at baseline. Multi-state Markov models showed dynamic transitions of depressive symptoms, with reversion generally exceeding progression. Both CI and PD were associated with increased progression risk and reduced likelihood of reversion, with PD showing stronger associations and a lower likelihood of symptom reversal. Their co-occurrence showed the strongest effects (NDS → MDS: HR = 1.76, 95% CI: 1.42–2.18; MDS → SDS: HR = 1.79, 95% CI: 1.30–2.46), alongside reduced reversion probabilities (MDS → NDS: HR = 0.82, 95% CI: 0.66–0.92; SDS → MDS: HR = 0.74, 95% CI: 0.57–0.91). Additive interactions were observed across all transition pathways (all P < 0.05). Findings were consistent in prospective analyses, and restricted cubic spline analyses supported dose–response relationships.
Conclusion
This study highlights the dynamic nature of depressive symptoms in older Chinese adults. CI and PD were independently associated with a higher likelihood of symptom progression and jointly associated with a lower probability of reversion. These findings suggest that approaches focusing on cognitive and physical function may hold promise for informing strategies to modify depressive symptom trajectories.
Supplementary Information
The online version contains supplementary material available at https://doi.org/10.1186/s12877-026-08065-7.
Keywords: Depressive symptoms, Cognitive impairment, Physical disability, Older adults, Multi-state Markov model
Introduction
Depression is a prevalent mental disorder characterized by persistent low mood and loss of interest, severely impairing quality of life [1]. Globally, it affects approximately 280 million people, with 5.7% of cases occurring among adults aged 60 and above (WHO, 2023). China, with its rapidly aging population [2], faces a particularly pressing challenge: its elderly population is projected to surpass the combined total of all developed countries by 2030 [3]. This demographic shift underscores the critical importance of addressing mental health in older adults as a key public health priority.
With the increasing focus on frailty syndrome in aging, the co-occurrence of physical disability (PD) and cognitive impairment (CI) is often regarded as an intertwined syndrome in older adults [4]. These conditions are interrelated [5] and may share common pathophysiological mechanisms, such as reduced vitamin D levels [6, 7], metabolic syndrome [8, 9], and chronic systemic inflammation [10, 11]. PD and CI signify a loss of independence in basic activities of daily living (ADL) and a decline in cognitive functions, progressively undermining an individual’s capacity for autonomous decision-making and social engagement. It is estimated that 45 million older adults in China were living with disability or cognitive decline in 2021, a number expected to grow steadily in the coming decades [12]. This expanding population not only places structural pressure on long-term care systems, healthcare resource allocation, and social security funds [13] but is also closely linked to profound psychological consequences [14, 15]. The erosion of autonomy and self-worth significantly elevates the risk of anxiety, depression, and other mood disorders, making mental health intervention an urgent public health priority.
Although numerous studies have identified PD and CI as key risk factors for depressive symptoms in older adults, significant gaps remain in understanding their combined effect on the longitudinal course of depression. Depressive symptomatology is dynamic, involving both progression and reversion over time [16, 17]. However, existing evidence relies predominantly on cross-sectional designs, capturing static associations at single time points rather than long-term trajectories. There is also a notable lack of large-scale longitudinal studies that quantify transitions between symptom states and pay sufficient attention to the potential for reversibility. Consequently, how the co-occurrence of PD and CI influences the dynamics of depressive symptoms—including the magnitude, pattern, and critical junctures of these transitions—remains unclear, limiting the development of targeted interventions.
Therefore, this study employs a nationally representative Chinese cohort to quantify transition probabilities between depressive symptom states, to assess the individual and combined associations of PD and CI with both the onset and reversibility of these symptoms, and to examine how these conditions jointly shape longitudinal symptom changes. By doing so, it aims to provide an evidence base for precise interventions that may help reduce adverse health outcomes and alleviate caregiving burdens in an aging society.
Method
Data sources and sample selection
Data were obtained from five waves (2011, 2013, 2015, 2018, and 2020) of the China Health and Retirement Longitudinal Study (CHARLS). CHARLS is a nationally representative longitudinal survey of the middle-aged and elderly population (aged ≥ 45 years) in China, conducted by the National School of Development at Peking University. It employs a multistage probability sampling strategy and has completed five waves of data collection to date. All participants provided informed consent. Participants were included based on the following criteria: (1) aged 60 years or older at the 2011 baseline; (2) completed at least two follow-up surveys; (3) provided responses to all 10 items of the Centre for Epidemiologic Studies Depression Scale (CES-D 10); and (4) completed assessments for cognitive function and ADL. Missing values in covariates (chronic diseases, sleep duration, and alcohol consumption) were handled using multiple imputation by chained equations (MICE) with 20 imputed datasets and 10 iterations each [18]. The imputation model included all covariates listed in Table 1, the outcome variables (depressive symptom states at each wave), and survival status (death). Results were pooled using Rubin’s rules. The final analytical sample comprised 31,923 observations from 8,473 unique participants. To assess potential selection bias due to missing data or insufficient follow-up, we compared the baseline characteristics of included participants (n = 8473) with those aged ≥ 60 years who were excluded from the analytic sample (n = 2499). A detailed flowchart of the sample selection process is presented in Fig. 1.
Table 1.
Baseline characteristics of the 8,473 older adults
| Variables | Total(N = 8473) | NDS(N = 5116) | MDS(N = 2710) | SDS(N = 647) | P value |
|---|---|---|---|---|---|
| Age, mean (SD) | 67.91 ± 6.59 | 67.79 ± 6.52 | 68.00 ± 6.65 | 68.40 ± 6.83 | 0.046 |
| Sex, n (%) | < 0.001 | ||||
| Male | 4294 (50.7) | 2893 (56.5) | 1182 (43.6) | 219 (33.8) | |
| Female | 4179 (49.3) | 2223 (43.5) | 1528 (56.4) | 428 (66.2) | |
| Chronic diseases, n (%) | < 0.001 | ||||
| No | 1976 (23.3) | 1412 (27.6) | 483 (17.8) | 81 (12.5) | |
| Yes | 5859 (69.2) | 3235 (63.2) | 2098 (77.4) | 526 (81.3) | |
| Unknown | 638 (7.5) | 469 (9.2) | 129 (4.8) | 40 (6.2) | |
| Educational level, n (%) | < 0.001 | ||||
| Illiterate | 4695 (55.4) | 2505 (49.0) | 1707 (63.0) | 483 (74.7) | |
| Primary school | 2140 (25.3) | 1375 (26.9) | 655 (24.2) | 110 (17.0) | |
| Middle school | 1036 (12.2) | 748 (14.6) | 242 (8.9) | 46 (7.1) | |
| ≥ High school | 602 (7.1) | 488 (9.5) | 106 (3.9) | 8 (1.2) | |
| Marital status, n (%) | < 0.001 | ||||
| Unmarried | 1716 (20.3) | 881 (17.2) | 635 (23.4) | 200 (30.9) | |
| Married | 6757 (79.7) | 4235 (82.8) | 2075 (76.6) | 447 (69.1) | |
| Residence, n (%) | < 0.001 | ||||
| Urban | 3518 (41.5) | 2417 (47.2) | 946 (34.9) | 155 (24.0) | |
| Rural | 4955 (58.5) | 2699 (52.8) | 1764 (65.1) | 492 (76.0) | |
| Sleep duration, n (%) | < 0.001 | ||||
| < 6 | 2754 (32.5) | 1178 (23.0) | 1207 (44.6) | 369 (57.0) | |
| 6–9 | 4491 (53.0) | 3132 (61.2) | 1168 (43.1) | 191 (29.5) | |
| > 9 | 519 (6.1) | 334 (6.5) | 158 (5.8) | 27 (4.2) | |
| Unknown | 709 (8.4) | 472 (9.3) | 177 (6.5) | 60 (9.3) | |
| Smoking, n (%) | < 0.001 | ||||
| No | 4849 (57.2) | 2809 (54.9) | 1619 (59.7) | 421 (65.1) | |
| Yes | 3624 (42.8) | 2307 (45.1) | 1091 (40.3) | 226 (34.9) | |
| Alcohol consumption, n (%) | < 0.001 | ||||
| No | 5666 (66.9) | 3254 (63.6) | 1930 (71.2) | 482 (74.5) | |
| Yes | 2593 (30.6) | 1720 (33.6) | 726 (26.8) | 147 (22.7) | |
| Unknown | 214 (2.5) | 142 (2.8) | 54 (2.0) | 18 (2.8) | |
| Cognitive Function and Physical Disability, n (%) | < 0.001 | ||||
| Healthy | 5255 (62.0) | 3707 (72.5) | 1349 (49.8) | 199 (30.8) | |
| CI | 1184 (14.0) | 659 (12.9) | 438 (16.2) | 87 (13.4) | |
| PD | 1480 (17.5) | 586 (11.5) | 639 (23.6) | 255 (39.4) | |
| CI + PD | 554 (6.5) | 164 (3.1) | 284 (10.4) | 106 (16.4) | |
NDS No depressive symptoms, MDS Mild depressive symptoms, SDS Severe depressive symptoms, CI Cognitive Impairment, PD Physical Disability
Fig. 1.
Flowchart of participant selection
Measurement of depressive symptoms
Depressive symptoms were assessed using the 10-item Centre for Epidemiologic Studies Depression Scale (CES-D 10). This scale is a validated self-report instrument for measuring depressive symptomatology in older adult populations [19]. Each item is scored from 0 to 3, with items 5 and 8 reverse-scored. The total score ranges from 0 to 30, with higher scores indicating greater symptom severity. A total score of ≥ 10 was used to define the presence of depressive symptoms. To allow for more granular analysis and in line with the operational definitions used in prior studies [20–22], depressive symptom severity was further categorized as follows: no depressive symptoms (NDS, score < 10), mild depressive symptoms (MDS, 10 ≤ score < 20), and severe depressive symptoms (SDS, score ≥ 20). The internal consistency of the CES-D 10 in our sample was high, with Cronbach's alpha coefficients ranging from 0.805 to 0.815 across survey waves.
Cognitive impairment and physical disability at baseline and follow‑up
Cognitive function was assessed using three tasks: episodic memory, evaluated by a word recall test; and non-memory cognition, evaluated by the Telephone Interview for Cognitive Status (TICS-10) and a drawing test. A global cognitive score (range 0–21) was calculated as the sum of the episodic memory and non-memory cognition scores, with higher scores indicating better function. Participants aged 60 and above were stratified into 5‑year age groups, and CI was defined as a global cognitive score more than one standard deviation below the mean of the corresponding age group. Cognitive function was reassessed at each survey wave, and the same age‑stratified definition was applied at every wave.
PD was assessed based on self-reported difficulty in performing six basic activities of daily living (ADL) [23]: dressing, bathing, eating, transferring (e.g., getting in or out of bed), toileting, and continence (controlling urination and defecation). An ADL difficulty score was summed across these items, with a higher total score indicating greater disability. For analytical purposes, participants were classified as having a PD if they reported needing help with one or more of these ADL items. PD was reassessed at each survey wave using the same ADL questions.
Based on the repeated assessments of CI and PD at each wave, we constructed a time‑varying four‑category variable (healthy, CI only, PD only, both) and updated it for every participant at each survey wave (2011–2020).
Baseline covariates
Covariates included socio-demographic and health-related characteristics. Socio-demographic factors comprised: age (continuous, in years), sex (male/female), marital status (married vs. unmarried [including separated, divorced, widowed, or never married]), educational attainment (categorised as illiterate, primary school, lower secondary school, or upper secondary school and above), and residence (rural/urban). Health-related factors included: current smoking status (yes/no), current drinking status (yes/no), the presence of at least one of 14 physician-diagnosed chronic conditions (yes/no), and average nightly sleep duration in the past month, categorised as short (< 6 h), normal (6–9 h), or long (> 9 h). All covariates were measured at baseline and treated as time‑fixed covariates.
Statistical analysis
All analyses were conducted using R software (version 4.4.3). First, a preliminary description of the data characteristics was provided, with participants categorized into three groups based on baseline depressive symptom status. Categorical variables are presented as numbers (percentages), and continuous variables as mean ± standard deviation. Baseline characteristics were compared across the three depressive symptom groups (NDS, MDS, SDS) using ANOVA for continuous variables and chi‑square tests for categorical variables. The analysis proceeded in three steps.
First, a continuous-time multi-state Markov model was selected to describe transitions between states of depressive symptoms (NDS, MDS, SDS) and death at the individual level, estimating transition intensities (instantaneous risk of moving from one state to another), probabilities, and average sojourn times (expected duration spent in a given state) [24, 25]. States were labeled as: 1 = NDS, 2 = MDS, 3 = SDS, and 4 = Death. Each participant began in one state at baseline and could remain or transition, with death as an absorbing state. Model fit was assessed by comparing observed and theoretical state frequencies over time [26, 27]. Given the low frequency of direct transitions between NDS and SDS, and considering that such observed transitions are likely to reflect two-step changes occurring within follow-up intervals rather than true single-step transitions, we adopted a progressive model based on previous studies [28, 29]. Direct transitions between non-adjacent states were therefore excluded to reduce the number of parameters and improve the stability of model estimation (Fig. 2). To incorporate the dynamic nature of CI and PD, the time‑varying four‑category variable (healthy, CI only, PD only, both) was included as a covariate in the model. For each transition interval between consecutive survey waves, the exposure status at the beginning of the interval was used to predict the transition intensities to other depressive states during that interval. This model was then used to estimate the strength and direction of the associations between CI, PD, and transitions between depressive symptom states, quantifying each factor’s association with specific transition pathways. To assess the interaction between CI and PD, we included CI, PD, and their product term (CI × PD) in a separate multi-state Markov model. Multiplicative interaction was evaluated using the interaction term, and additive interaction was assessed by the Relative Excess Risk due to Interaction (RERI), which quantifies departure from additivity on the risk scale. RERI > 0, = 0, and < 0 indicate positive interaction (synergism), no interaction (pure additivity), and negative interaction (antagonism), respectively. Confidence intervals for RERI were estimated using the delta method [30].
Fig. 2.
Schematic of the model transition pathways and intensities. Note: NDS = no depressive symptoms;MDS = mild depressive symptoms; SDS = severe depressive symptoms
Second, to mitigate potential reverse causality and examine prospective associations, we constructed separate prospective cohorts for each transition of interest (NDS → MDS, MDS → SDS, MDS → NDS, and SDS → MDS). Cox proportional hazards models were used to estimate transition-specific hazard ratios (HRs). For each transition, individuals entered the risk set upon reaching the corresponding starting state and were followed until the occurrence of the target transition, loss to follow-up, death, or the end of the study period, whichever came first. Individuals who experienced a transition were allowed to re-enter the risk set if they subsequently returned to the same starting state, using a counting process approach to account for recurrent transitions. The time-varying four-category exposure variable was updated at the beginning of each risk interval, defined as the wave at which an individual entered the starting state. Death was treated as a censoring event in the primary analysis. The proportional hazards assumption was assessed using Schoenfeld residuals and log–log plots, with no evidence of violation.
These two modeling approaches serve complementary purposes. The multi-state Markov model characterizes the dynamic patterns of depressive symptom transitions, capturing progression and reversion over time, whereas Cox models quantify the associations of CI and PD with specific transitions using hazard ratios, providing interpretable estimates of relative risk. Together, the Markov model captures the overall transition process of depressive symptoms and allows assessment of how CI and PD relate to this process at a system level, while Cox models quantify their effects on specific transitions in clinically interpretable terms.
Finally, restricted cubic splines (RCS) with four knots were used to explore the continuous relationships of cognitive function and physical disability with transition risks. Knots were placed at the 5th, 35th, 65th, and 95th percentiles of each variable's distribution, with the median set as the reference value. Both continuous variables were winsorized at the 1 st and 99th percentiles for robustness [31]. Subgroup analyses were conducted to consider potential interactions with covariates. Results from these models are presented as HRs with 95% confidence intervals (CIs).
Several sensitivity analyses were conducted: (1) To reflect clinical practice (e.g., direct transition from NDS to SDS), we relaxed the constraints of the multi-state Markov model by allowing non-zero transition intensities between state 1 and state 3 (i.e., q13, q31 ≠ 0) [32]. Nested models were compared via the likelihood ratio test (LRT), and model fit was assessed using the Akaike Information Criterion (AIC). One-year transition probability matrices were further compared to verify the robustness of the primary model’s pathway assumptions. (2) All core analyses were repeated using a complete-case sample (excluding observations with any missing covariates).
A two-tailed P-value < 0.05 was considered statistically significant.
Result
Baseline characteristics
The baseline characteristics of the 8,473 participants are summarized in Table 1. The mean age was 67.91 years, and 50.68% (n = 4,294) were men. At baseline, 60.38% (n = 5,116) of participants were classified as having NDS. Among those with depressive symptoms (39.62%, n = 3,357), the majority had MDS (31.98%, n = 2,710), while 7.64% (n = 647) had SDS.
A comparison of baseline characteristics between included and excluded participants is presented in Table S1. The two groups were generally comparable across most variables (all P > 0.05), although excluded participants had a higher proportion of illiteracy and slightly higher alcohol consumption (both P < 0.05).
Frequency, intensity, probability, and average sojourn time of transitions between depressive symptom states
Among the 31,923 records, transitions from NDS to MDS occurred 2,919 times, from MDS to NDS 2,826 times, from MDS to SDS 760 times, and from SDS to MDS 685 times. Excluding remaining in the same state, the transition from NDS to MDS was the most frequent. Details are provided in Table 2.
Table 2.
Transition frequencies, intensities, and 1-, 3-, 5-, and 10-year transition probabilities between depressive symptom states
| Transition State | Transition Frequency | Transition intensity | Transition probability | |||
|---|---|---|---|---|---|---|
| 1 year | 3 year | 5 year | 10 year | |||
| NDS → NDS | 10,738 | - | 0.846 | 0.682 | 0.599 | 0.496 |
| NDS → MDS | 2919 | 0.176 (0.168,0.183) | 0.125 | 0.217 | 0.238 | 0.227 |
| NDS → SDS | 367 | 0 | 0.008 | 0.035 | 0.051 | 0.058 |
| NDS → Death | 787 | 0.020 (0.018,0.023) | 0.021 | 0.066 | 0.111 | 0.219 |
| MDS → NDS | 2826 | 0.363 (0.347,0.379) | 0.258 | 0.447 | 0.492 | 0.468 |
| MDS → MDS | 2885 | - | 0.624 | 0.362 | 0.286 | 0.226 |
| MDS → SDS | 760 | 0.150 (0.139,0.163) | 0.087 | 0.104 | 0.086 | 0.062 |
| MDS → Death | 489 | 0.032 (0.027,0.038) | 0.031 | 0.087 | 0.136 | 0.245 |
| SDS → NDS | 310 | 0 | 0.064 | 0.267 | 0.385 | 0.434 |
| SDS → MDS | 685 | 0.541 (0.499,0.585) | 0.314 | 0.374 | 0.311 | 0.222 |
| SDS → SDS | 514 | - | 0.576 | 0.242 | 0.133 | 0.064 |
| SDS → Death | 163 | 0.052 (0.041,0.067) | 0.047 | 0.117 | 0.171 | 0.279 |
NDS No depressive symptoms, MDS Mild depressive symptoms, SDS Severe depressive symptoms
Using a multi-state Markov model, the estimated transition intensities were as follows: from NDS to MDS, 0.176; from MDS to NDS, 0.363 (2.06 times the intensity of NDS to MDS); from MDS to SDS, 0.150; and from SDS to MDS, 0.541 (3.61 times the intensity of MDS to SDS). The intensity of reversion from MDS to NDS was 2.42 times that of progression from MDS to SDS. Transition probabilities over 1-, 3-, 5-, and 10-year intervals were also estimated (see Table 2). Overall, the probability of individuals remaining in their original state gradually decreased over time, with the likelihood of symptomatic reversion exceeding that of deterioration. Furthermore, the average sojourn times in the NDS, MDS, and SDS states were 5.1 years, 1.83 years, and 1.69 years, respectively.
Model fit
Figure 3 presents the goodness-of-fit plot, comparing the observed and model-predicted prevalence of each state over time. The predicted trajectories closely align with the observed data, demonstrating good fit of the multi-state Markov model.
Fig. 3.
Goodness-of-fit of the model: observed versus predicted state prevalence
Associations of CI and PD with transition intensities
Table 3 presents the associations between cognitive impairment, physical disability, and transition intensities across depressive symptom states based on multi-state Markov models. Compared with healthy individuals, participants with either CI or PD showed higher risks of transition to more severe depressive states. Participants with CI alone showed increased risks of worsening from NDS to MDS (HR = 1.22) and from MDS to SDS (HR = 1.42). Participants with PD alone showed elevated risks for both transitions (NDS → MDS: HR = 1.50; MDS → SDS: HR = 1.47) and lower likelihood of symptom reversion. The co-occurrence of CI and PD was associated with the highest risks of symptom deterioration (NDS → MDS: HR = 1.76; MDS → SDS: HR = 1.79) and reduced likelihood of reversion (MDS → NDS: HR = 0.82; SDS → MDS: HR = 0.74). Additive interaction between CI and PD was observed across all transition pathways, whereas no evidence of multiplicative interaction was found (see Table S2).
Table 3.
Hazard ratios (HR,95%CI) for transitions between depressive symptom states, associated with cognitive impairment and physical disability, based on the multi-state Markov model
| Cognitive Function and Physical Disability | Worsening | Reversion | ||
|---|---|---|---|---|
| NDS → MDS | MDS → SDS | MDS → NDS | SDS → MDS | |
| HR(95%CI) | HR(95%CI) | HR(95%CI) | HR(95%CI) | |
| Model 0 | ||||
| Healthy | 1.00 | 1.00 | 1.00 | 1.00 |
| CI | 1.55 (1.36,1.75) | 1.58 (1.31,1.92) | 1.04 (0.91,1.18) | 0.89 (0.79,1.17) |
| PD | 1.73 (1.53,1.95) | 1.65 (1.23,2.16) | 0.84 (0.74,0.94) | 0.86 (0.70,0.95) |
| CI + PD | 2.42 (1.97,2.97) | 2.01 (1.52,2.66) | 0.82 (0.70,0.96) | 0.85 (0.68,0.97) |
| Model 1 | ||||
| Healthy | 1.00 | 1.00 | 1.00 | 1.00 |
| CI | 1.25 (1.09,1.43) | 1.48 (1.10,2.01) | 1.04 (0.97,1.20) | 0.88 (0.77,1.25) |
| PD | 1.55 (1.37,1.76) | 1.51 (1.23,1.83) | 0.85 (0.73,0.98) | 0.89 (0.72,1.20) |
| CI + PD | 1.86 (1.50,2.31) | 1.79 (1.33,2.41) | 0.85 (0.69,0.93) | 0.74 (0.54,0.93) |
| Model 2 | ||||
| Healthy | 1.00 | 1.00 | 1.00 | 1.00 |
| CI | 1.22 (1.06,1.41) | 1.42 (1.16,1.74) | 1.04 (0.97,1.20) | 0.88 (0.78,1.25) |
| PD | 1.50 (1.10,2.06) | 1.47 (1.29,1.67) | 0.84 (0.74,0.95) | 0.95 (0.78,1.16) |
| CI + PD | 1.76 (1.42,2.18) | 1.79 (1.30,2.46) | 0.82 (0.66,0.92) | 0.74 (0.57,0.91) |
Model 0: unadjusted; Model 1: adjusted for sex, age, educational level, marital status, and residence; Model 2: additionally adjusted for smoking, alcohol consumption, sleep duration, and chronic diseases based on Model 1
CI Cognitive impairment, PD Physical disability, NDS No depressive symptoms, MDS Mild depressive symptoms, SDS Severe depressive symptoms
Estimates in bold are statistically significant (P < 0.05)
Association of CI and PD with the risk and reversibility of depressive symptoms
Results from Cox proportional hazards regression models showed that (Table 4), compared to healthy individuals, cognitive impairment, physical disability, and their combination were associated with an increased risk of developing MDS by 23% (HR = 1.23), 22% (HR = 1.22), and 75% (HR = 1.75), respectively. These conditions also increased the risk of progression from MDS to SDS by 12% (HR = 1.12), 27% (HR = 1.27), and 46% (HR = 1.46), respectively. Furthermore, compared to healthy individuals, physical disability alone and the combination of cognitive impairment with physical disability were associated with a reduced probability of transitioning to a state of NDS, with reductions of 28% (HR = 0.72) and 36% (HR = 0.64), respectively. The combination of cognitive impairment and physical disability was also associated with a 10% reduction (HR = 0.90) in the probability of reversion from SDS to MDS.
Table 4.
Associations with the risk and reversibility of depressive symptom states, estimated from Cox proportional hazards models
| Cognitive Function and Physical Disability | Worsening | Reversion | ||
|---|---|---|---|---|
| NDS → MDS | MDS → SDS | MDS → NDS | SDS → MDS | |
| HR(95%CI) | HR(95%CI) | HR(95%CI) | HR(95%CI) | |
| Healthy | 1.00 | 1.00 | 1.00 | 1.00 |
| CI | 1.23 (1.06,1.41) | 1.12 (1.01,1.44) | 0.86 (0.75,1.03) | 1.05 (0.77,1.43) |
| PD | 1.22 (1.04,1.44) | 1.27 (1.02,1.59) | 0.72 (0.63,0.82) | 0.85 (0.66,1.11) |
| CI + PD | 1.75 (1.39,2.20) | 1.46 (1.11,1.92) | 0.64 (0.53,0.77) | 0.90 (0.80,0.97) |
CI Cognitive impairment, PD Physical disability, NDS No depressive symptoms, MDS Mild depressive symptoms, SDS Severe depressive symptoms
Models were adjusted for sex, age, educational level, marital status, residence, smoking, alcohol consumption, sleep duration, and chronic diseases
Estimates in bold are statistically significant (P < 0.05)
RCS plot and subgroup analysis
Multivariable RCS curves showed dose–response relationships between cognitive function, PD, and depressive symptom transitions (Fig. 4). For NDS → MDS, both cognitive function and PD showed significant non-linear associations (both P < 0.05). For MDS → SDS, PD showed a borderline non-linear association (P = 0.054), whereas cognitive function showed a significant non-linear association (P < 0.05). For MDS → NDS reversion, both cognitive function and PD showed significant non-linear associations (both P < 0.05). For SDS → MDS reversion, PD showed a significant overall association (P-overall = 0.015) with no evidence of non-linearity (P-nonlinear = 0.325), indicating an approximately linear inverse association. Cognitive function also showed a significant overall association (P-overall < 0.001), with evidence of non-linearity (P-nonlinear < 0.05). The threshold level of cognitive function associated with a protective effect was lower for SDS → MDS than for MDS → NDS.
Fig. 4.
Relationship between the continuous levels of cognitive function and physical disability and the development and reversal of depressive symptoms. A Relationship with the transition from NDS to MDS. B Relationship with the transition from MDS to SDS. C Relationship with the reversion from MDS to NDS. D Relationship with the reversion from SDS to MDS. HR: Hazard ratio adjusted for age, sex, educational level, marital status, residence, chronic diseases, sleep duration, smoking, and alcohol consumption
Subgroup analyses (Figures S1-S4) explored potential effect modification by demographic and behavioral factors. The most consistent patterns were observed for age and smoking, which modified the associations of cognitive function with depressive symptom transitions (interaction P < 0.05). Other subgroup findings were less consistent and should be interpreted as exploratory.
Sensitivity analysis
In the sensitivity analysis relaxing the progressive assumption of the multi-state Markov model, the likelihood ratio test showed no significant difference between the main and alternative models (G = 0.25, df = 2, p > 0.05). The AIC values of the main model (43,718.88) and the sensitivity model (43,721.07) were very similar, indicating that relaxing the progressive assumption did not meaningfully affect model fit. The one-year transition probability matrices of the two models were nearly identical (Table S3). These results indicate that the restriction of model pathway assumptions did not substantially affect the main study conclusions.
After excluding observations with any missing covariates, the associations between cognitive function, physical disability, and transitions between depressive symptom states were consistent with the main results (Tables S4-S5), further supporting the robustness of our findings.
Discussion
This is among the first studies to examine associations of CI and PD with the risk and reversibility of depressive symptom transitions. We found that transition intensities and sojourn times differed markedly between symptom states, with PD showing a stronger association than CI. Notably, CI and PD co-occurrence was linked to the highest risk of symptom deterioration and the lowest likelihood of reversion. Clear dose–response relationships were also observed between cognitive function, PD, and transition hazards.
Prior research indicates that depressive symptom progression and reversion are dynamic processes [33, 34]. Our multi-state analysis confirms this pattern. Reversion intensity from MDS to NDS was 2.06 times that of progression from NDS to MDS; from SDS to MDS was 3.61 times that of MDS → SDS; and from MDS to NDS was 2.42 times that of MDS → SDS. Mean sojourn times were 5.1 years (NDS), 1.83 years (MDS), and 1.69 years (SDS). The probability of remaining in a symptomatic state—particularly SDS—declined over time. Together, these findings highlight distinct dynamic patterns across symptom severity, suggesting an unstable pathophysiological state [35] and supporting the need for staged, targeted interventions.
Our findings suggest that older adults with PD face a higher risk of depressive symptom progression and a lower likelihood of reversion compared to those with CI. This may be partly explained by intact cognitive function allowing a clearer awareness of lost independence and social isolation, thereby exacerbating psychological distress [36]. Furthermore, PD in older adults is often precipitated by chronic conditions such as stroke, diabetes, and osteoarthritis [37, 38]. These diseases not only limit mobility but are also associated with persistent issues like chronic pain (disrupting sleep [39]) and symptoms such as fatigue and anorexia [40]. The confluence of this physiological burden and the psychological stress of disability likely worsens mental health outcomes. In contrast, CI may be associated with reduced insight into one’s own functional decline and changing social roles [41]. Additionally, the symptomatic overlap between CI and depression (e.g., apathy, loss of interest) can mask depressive disorders [42]. This dual effect—impaired self-awareness and symptomatic overlap—may lead to the underdetection and delayed intervention of depression in this population.
Our findings suggest that the co-occurrence of CI and PD is associated with a higher risk of depressive symptom progression and a lower likelihood of reversion compared with either condition alone. Physiologically, PD often involves chronic stress, pain, and sleep disruption, which may activate systemic inflammatory responses [43, 44], while CI is frequently linked to neurotransmitter dysregulation [45–47]. These pathways may share common mechanisms—for example, inflammatory mediators can cross the blood–brain barrier [48] and affect mood-regulating systems [49]—potentially contributing to the observed associations when both conditions are present. However, the underlying mechanisms of this joint effect warrant further investigation. Overall, this stepwise pattern suggests that the cumulative burden of functional impairments may be more clinically relevant than a specific interaction effect between CI and PD in explaining adverse depressive outcomes.
Furthermore, mobility limitations due to PD and communication barriers due to CI can lead to difficulties in accessing social support, which is a key protective factor against late-life depression and cognitive decline[50]. The lack of social support significantly exacerbates depressive symptoms in this population [51, 52]. Conversely, adequate social support may buffer the negative effects of cognitive impairment and physical disability on mental health by promoting health behaviors, providing practical help, and enhancing psychological resilience. In this study, only marital status was included as a covariate; however, marital status alone does not fully capture the multidimensional nature of social support (including emotional support, instrumental assistance, social network size, and perceived loneliness) [53]. Therefore, future studies should incorporate multidimensional assessments of social support to elucidate its moderating role in the transitions between depressive symptom states.
Additionally, previous studies suggest that depressive symptoms may accelerate cognitive decline and functional disability through pathways such as unhealthy health behaviors, chronic stress, and neuroendocrine dysregulation [54, 55]. Conversely, cognitive impairment and physical disability may increase vulnerability to depressive symptoms through loss of independence, reduced social engagement, and increased psychological burden, indicating a bidirectional relationship. In this study, we used exposure status at the beginning of each transition interval to help establish temporal ordering. However, reverse causation cannot be fully excluded, as subclinical or unmeasured depressive symptoms may precede and contribute to subsequent functional decline. Therefore, the observed associations are more likely to reflect a dynamic and reciprocal process rather than a strictly unidirectional effect. Future studies employing cross-lagged panel models or more fine-grained longitudinal designs with shorter follow-up intervals are warranted to better disentangle the reciprocal relationships and clarify the directionality among these domains.
This study reveals a dose–response relationship between cognitive function, physical disability, and depressive symptom transitions: greater impairment is associated with higher depression risk and lower likelihood of recovery. Physical disability not only independently increases depressive symptom risk but may also compound the effects of cognitive decline by restricting social participation, increasing dependency, and reducing personal control, thereby confining individuals to a depressive state [51, 56]. Notably, even among individuals with SDS, higher cognitive function may provide cognitive reserve to support engagement in psychological interventions and rehabilitation, increasing the likelihood of reversion to MDS and facilitating eventual recovery.
Subgroup analyses showed that age and smoking moderated the association between cognitive function and depressive symptom transitions. In addition, a combined moderating effect of smoking and physical disability was observed for the reversion from MDS to NDS. These findings suggest that the pathway linking cognitive function to depressive symptoms may be embedded in a complex regulatory network of demographic and behavioral factors, implying that depression prevention and treatment strategies should adopt a multidimensional approach. For high-risk individuals (e.g., those with concurrent cognitive decline, smoking, and physical disability), integrated interventions simultaneously targeting cognitive function, smoking cessation, and physical function may be more effective than single-domain strategies.
In summary, this study highlights cognitive function and physical disability not merely as static risk factors, but as dynamic determinants that influence the entire course of late-life depression. This implies that prevention and management strategies should concurrently address both domains. For older adults with existing physical disability, integrated care should include routine cognitive screening and early intervention to disrupt the ‘disability-cognitive decline-depression’ cycle. For the general elderly population, public health initiatives that jointly promote cognitive training and physical activity may represent a valuable strategy for primary prevention and supporting recovery from depression.
Strengths and limitations
This study uses CHARLS data from China and applies a multi-model approach to address the limitations of traditional analytical methods. By incorporating the dynamic nature of depressive symptoms, it describes the associations of physical disability and cognitive impairment with the risk and reversibility of depressive symptoms in older Chinese adults, and further examines the independent and combined effects of these conditions on transitions between different depressive symptom states, holding important clinical implications. However, this study also has several limitations. First, depressive symptom status was defined using the CES-D-10 scale, which may involve measurement error and is not equivalent to a clinical diagnosis, potentially leading to instability in the classification of symptom states. Second, beyond measurement bias, our broad binary definitions of CI and PD may introduce heterogeneity that could partly explain the observed differential associations; nevertheless, RCS analyses using continuous scores showed dose‑response relationships, partially alleviating this concern. Third, the multi-state Markov model assumes that transition risks depend only on the current state and does not account for potential duration dependence. However, differences in sojourn times may reflect variation in transition intensities rather than violation of this assumption, and the interval-censored nature of the CHARLS data further limits the feasibility of applying semi-Markov models. Fourth, because depressive symptoms, CI, and PD were all assessed at discrete survey waves approximately 2‑3 years apart, we assumed that these states remained constant within each interval. This may not capture short‑term fluctuations or multiple transitions within an interval, potentially leading to underestimation of the true associations and transition frequencies. Fifth, our assessment of social support was limited to marital status, which does not capture its multidimensional nature (e.g., emotional support, instrumental assistance, social network size, and perceived loneliness). Additionally, despite adjusting for multiple covariates, residual confounding may persist due to unmeasured factors such as medication use, nutritional status, and healthcare access, which are inherent limitations of observational studies. Sixth, although multiple imputation was used for missing covariates, participants with missing key variables or insufficient follow-up were excluded, which may introduce selection bias and informative attrition. If non-random, this could underestimate transitions to worse states and overestimate recovery, thereby biasing transition probabilities and hazard ratios. Finally, although CHARLS is nationally representative, sampling weights and complex survey design factors were not incorporated in this study; therefore, the generalizability of the findings to the national population should be interpreted with caution.
Conclusion
Among older Chinese adults, depressive symptoms demonstrate distinct dynamic patterns across severity states. The co-occurrence of cognitive impairment and physical disability is associated with the highest likelihood of progression and the lowest probability of reversion, with physical disability showing a stronger association than cognitive impairment. These findings highlight the important role of cognitive and physical function in shaping long-term depressive symptom trajectories and suggest the potential value of incorporating both domains into integrated intervention frameworks.
Supplementary Information
Acknowledgements
The authors are grateful to all individuals for participating in this study.
Abbreviations
- CI
Cognitive Impairment
- PD
Physical Disability
- NDS
No Depressive Symptoms
- MDS
Mild Depressive Symptoms
- SDS
Severe Depressive Symptoms
- ADL
Activities of Daily Living
- CHARLS
China Health and Retirement Longitudinal Study
- CES-D 10
Centre for Epidemiologic Studies Depression Scale 10-item
- TICS-10
Telephone Interview for Cognitive Status 10-item
- RCS
Restricted Cubic Splines
- AIC
Akaike Information Criterion
- LRT
Likelihood Ratio Test
- RERI
Relative Excess Risk due to Interaction
Authors’ contributions
RZ designed and drafted the manuscript. RZ, CJ, ZG, XS, GC and WT conducted data collection and analysis. WT and RZ have made revisions to the important content of the manuscript. WT provided assistance in revising the manuscript format. All authors have read and approved the manuscript.
Funding
This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.
Data availability
All of our research data are available from the CHARLS(http://charls.pku.edu.cn) datasets.
Declarations
Ethics approval and consent to participate
The CHARLS study was approved by the Ethical Review Committee of Peking University (IRB00001052-11015).
Consent for publication
Not applicable.
Competing interests
The authors declare that they have no competing interests.
Footnotes
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
References
- 1.Kendler KS. The origin of our modern concept of depression-the history of melancholia from 1780–1880: a review. JAMA Psychiat. 2020;77(8):863–8. 10.1001/jamapsychiatry.2019.4709. [DOI] [PubMed] [Google Scholar]
- 2.Xia CC, Lin B. International experiences in addressing population aging and implications for China’s population policy. Soc Sci J. 2023;45(5):148–57. [Google Scholar]
- 3.Li XL. Challenges and countermeasures of accelerated population aging in China. New Finance. 2023;8:4-7,35. [Google Scholar]
- 4.Liu X, Wei YN, Liu J, Wang JT. Factors influencing comorbidity of physical dysfunction and cognitive impairment in older adults. Chin Gen Pract. 2024;27(26):3281–8. [Google Scholar]
- 5.Mello BH, Lenardt MH, Moraes DC, et al. Cognitive impairment and physical frailty in older adults in secondary health care. Rev Esc Enferm USP. 2021;55:e03687. 10.1590/S1980-220x2019029803687. [DOI] [PubMed] [Google Scholar]
- 6.Zeng XK, Shen SS, Chu JJ, He T, Cheng L, Chen XJ. Relationship of serum vitamin D level on geriatric syndromes and physical performance impairment in elderly hypertensive patients. J Geriatr Cardiol. 2016;13(6):537–45. 10.11909/j.issn.1671-5411.2016.06.013. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Laughlin GA, Kritz-Silverstein D, Bergstrom J, et al. Vitamin D insufficiency and cognitive function trajectories in older adults: the Rancho Bernardo study. J Alzheimers Dis. 2017;58(3):871–83. 10.3233/JAD-161295. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Penninx BW, Nicklas BJ, Newman AB, et al. Metabolic syndrome and physical decline in older persons: results from the Health, Aging and Body Composition Study. J Gerontol A Biol Sci Med Sci. 2009;64(1):96–102. 10.1093/gerona/gln005. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Yaffe K, Weston AL, Blackwell T, Krueger KA. The metabolic syndrome and development of cognitive impairment among older women. Arch Neurol. 2009;66(3):324–8. 10.1001/archneurol.2008.566. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Soysal P, Stubbs B, Lucato P, et al. Inflammation and frailty in the elderly: a systematic review and meta-analysis. Ageing Res Rev. 2016;31:1–8. 10.1016/j.arr.2016.08.006. [DOI] [PubMed] [Google Scholar]
- 11.Ozawa M, Shipley M, Kivimaki M, Singh-Manoux A, Brunner EJ. Dietary pattern, inflammation and cognitive decline: the Whitehall II prospective cohort study. Clin Nutr. 2017;36(2):506–12. 10.1016/j.clnu.2016.01.013. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Ansah JP, Chiu CT, Wei-Yan AC, Min TLS, Matchar DB. Trends in functional disability and cognitive impairment among the older adult in China up to 2060: estimates from a dynamic multi-state population model. BMC Geriatr. 2021;21(1):380. 10.1186/s12877-021-02309-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Hu HW, Li Y, Yu XD. The social support system for family care of disabled and dementia elderly: historical evolution, practical logic, and optimization paths. Soc Secur Stud. 2025;5:3–16. [Google Scholar]
- 14.Wang J, Luo N, Sun Y, et al. Exploring the reciprocal relationship between activities of daily living disability and depressive symptoms among middle-aged and older Chinese people: a four-wave, cross-lagged model. BMC Public Health. 2023;23(1):1180. 10.1186/s12889-023-16100-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Bora E, Harrison BJ, Yücel M, Pantelis C. Cognitive impairment in euthymic major depressive disorder: a meta-analysis. Psychol Med. 2013;43(10):2017–26. 10.1017/S0033291712002085. [DOI] [PubMed] [Google Scholar]
- 16.Shen G, Yang X, Zou Y, et al. Dynamic changes in network structure of depressive symptoms: a two-year naturalistic follow-up study. BMC Psychiatry. 2025;25(1):676. 10.1186/s12888-025-07124-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Kessing LV. Epidemiology of subtypes of depression. Acta Psychiatr Scand Suppl. 2007;433:85–9. 10.1111/j.1600-0447.2007.00966.x. [DOI] [PubMed] [Google Scholar]
- 18.Welch C, Bartlett J, Petersen I. Application of multiple imputation using the two-fold fully conditional specification algorithm in longitudinal clinical data. Stata J. 2014;14(2):418–31. [PMC free article] [PubMed] [Google Scholar]
- 19.Irwin M, Artin KH, Oxman MN. Screening for depression in the older adult: criterion validity of the 10-item Center for Epidemiological Studies Depression Scale (CES-D). Arch Intern Med. 1999;159(15):1701–4. 10.1001/archinte.159.15.1701. [DOI] [PubMed] [Google Scholar]
- 20.Zhou L, Ma X, Wang W. Relationship between cognitive performance and depressive symptoms in Chinese older adults: the China Health and Retirement Longitudinal Study (CHARLS). J Affect Disord. 2021;281:454–8. 10.1016/j.jad.2020.12.059. [DOI] [PubMed] [Google Scholar]
- 21.Wang JJ, Zhang ZY, Jia YM, Chen O. Relationship between grip strength and the progression and recovery of depressive states in middle-aged and elderly individuals with mild to moderate depression. Mod Prev Med. 2023;50(10):1884–8. 10.20043/j.cnki.MPM.202211403. [DOI] [Google Scholar]
- 22.Yu S, Chu Y. Joint effects of sleep duration and depressive symptoms on the risk of cardiovascular disease and mediating role of depressive symptoms in middle-aged and older Chinese adults: a nationwide prospective cohort study. Gen Hosp Psychiatry. 2025;98:14–22. 10.1016/j.genhosppsych.2025.11.011. [DOI] [PubMed] [Google Scholar]
- 23.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(1):577. 10.1186/s12916-025-04400-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Le-Rademacher JG, Therneau TM, Ou FS. The utility of multistate models: a flexible framework for time-to-event data. Curr Epidemiol Rep. 2022;9(3):183–9. 10.1007/s40471-022-00291-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Titman AC, Putter H. General tests of the Markov property in multi-state models. Biostatistics. 2022;23(2):380–96. 10.1093/biostatistics/kxaa030. [DOI] [PubMed] [Google Scholar]
- 26.Luo YX, Zhou XH, Heng T, et al. Bidirectional transitions of sarcopenia states in older adults: the longitudinal evidence from CHARLS. J Cachexia Sarcopenia Muscle. 2024;15(5):1915–29. 10.1002/jcsm.13541. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.RouziNiyazi AZ, Shan M, Hu XM, He XY, Chen J, Ni MJ. Analysis of antiretroviral therapy and its influencing factors among HIV/AIDS patients in Yili Prefecture, Xinjiang using a multi-state Markov model. Chin J Public Health. 2023;39(12):1513–9. [Google Scholar]
- 28.Jack CR Jr, Therneau TM, Wiste HJ, et al. Transition rates between amyloid and neurodegeneration biomarker states and to dementia: a population-based, longitudinal cohort study. Lancet Neurol. 2016;15(1):56–64. 10.1016/S1474-4422(15)00323-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Jackson DJ, Evans MD, Gangnon RE, et al. Evidence for a causal relationship between allergic sensitization and rhinovirus wheezing in early life. Am J Respir Crit Care Med. 2012;185(3):281–5. 10.1164/rccm.201104-0660OC. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Knol MJ, VanderWeele TJ. Recommendations for presenting analyses of effect modification and interaction. Int J Epidemiol. 2012;41(2):514–20. 10.1093/ije/dyr218. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Hou D, Liu S, Sun Y, Liu C, Shang X, Pei L, et al. Estimated glucose disposal rate associated with risk of frailty and likelihood of reversion. J Cachexia Sarcopenia Muscle. 2025;16(2):e13814. 10.1002/jcsm.13814. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Yuan M, Xu C, Fang Y. The transitions and predictors of cognitive frailty with multi-state Markov model: a cohort study. BMC Geriatr. 2022;22(1):550. 10.1186/s12877-022-03220-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Wu Y, Su B, Zhong P, Zhao Y, Chen C, Zheng X. Association between chronic disease status and transitions in depressive symptoms among middle-aged and older Chinese population: insights from a Markov model-based cohort study. J Affect Disord. 2024;363:445–55. 10.1016/j.jad.2024.07.116. [DOI] [PubMed] [Google Scholar]
- 34.Schulz D. Depression development: from lifestyle changes to motivational deficits. Behav Brain Res. 2020;395:112845. 10.1016/j.bbr.2020.112845. [DOI] [PubMed] [Google Scholar]
- 35.Ebrahimi OV, Borsboom D, Hoekstra RHA, et al. Towards precision in the diagnostic profiling of patients: leveraging symptom dynamics as a clinical characterisation dimension in the assessment of major depressive disorder. Br J Psychiatry. 2024;224(5):157–63. 10.1192/bjp.2024.19. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Scher C, Nepomnyaschy L, Amano T. Comparison of cognitive and physical decline as predictors of depression among older adults. J Appl Gerontol. 2023;42(3):387–98. 10.1177/07334648221139255. [DOI] [PubMed] [Google Scholar]
- 37.Klijs B, Nusselder WJ, Looman CW, Mackenbach JP. Contribution of chronic disease to the burden of disability. PLoS ONE. 2011;6(9):e25325. 10.1371/journal.pone.0025325. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Chen YY, Wen Y, Shu XY. Influencing factors of disability status among older adults with multimorbidity. Chin Gen Pract. 2025;28(32):4024-30,4054. [Google Scholar]
- 39.Finan PH, Goodin BR, Smith MT. The association of sleep and pain: an update and a path forward. J Pain. 2013;14(12):1539–52. 10.1016/j.jpain.2013.08.007. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.Cruz-Jentoft AJ, Bahat G, Bauer J, et al. Sarcopenia: revised European consensus on definition and diagnosis. Age Ageing. 2019;48(4):601. 10.1093/ageing/afz046. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41.Giebel C, Silva-Ribeiro W, Watson J, et al. A systematic review on the evidence of misdiagnosis in dementia and its impact on accessing dementia care. Int J Geriatr Psychiatry. 2024;39(10):e6158. 10.1002/gps.6158. [DOI] [PubMed] [Google Scholar]
- 42.Teixeira AL, Gregg A, Gentry MT, et al. Cognitive deficits in late-life depression: from symptoms and assessment to therapeutics. Focus (Am Psychiatr Publ). 2025;23(2):183–94. 10.1176/appi.focus.20240046. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43.Miller AH, Raison CL. The role of inflammation in depression: from evolutionary imperative to modern treatment target. Nat Rev Immunol. 2016;16(1):22–34. 10.1038/nri.2015.5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44.Dantzer R, O’Connor JC, Freund GG, Johnson RW, Kelley KW. From inflammation to sickness and depression: when the immune system subjugates the brain. Nat Rev Neurosci. 2008;9(1):46–56. 10.1038/nrn2297. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45.Duman RS, Aghajanian GK, Sanacora G, Krystal JH. Synaptic plasticity and depression: new insights from stress and rapid-acting antidepressants. Nat Med. 2016;22(3):238–49. 10.1038/nm.4050. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 46.Cui X, Ding Q, Yu S, Zhang S, Li X. The deficit in cognitive reappraisal capacity in individuals with anxiety or depressive disorders: meta-analyses of behavioral and neuroimaging studies. Clin Psychol Rev. 2024;114:102480. 10.1016/j.cpr.2024.102480. [DOI] [PubMed] [Google Scholar]
- 47.Smoski MJ, Keng SL, Schiller CE, Minkel J, Dichter GS. Neural mechanisms of cognitive reappraisal in remitted major depressive disorder. J Affect Disord. 2013;151(1):171–7. 10.1016/j.jad.2013.05.073. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 48.Hodes GE, Kana V, Menard C, Merad M, Russo SJ. Neuroimmune mechanisms of depression. Nat Neurosci. 2015;18(10):1386–93. 10.1038/nn.4113. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 49.Hage Z, Madeira MM, Koliatsis D, Tsirka SE. Convergence of endothelial dysfunction, inflammation and glucocorticoid resistance in depression-related cardiovascular diseases. BMC Immunol. 2024;25(1):61. 10.1186/s12865-024-00653-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 50.Turner ST, Carmel S, O’Rourke N, Raveis VH, Tovel H, Cohn-Schwartz E. Social support and symptoms of depression in late life: bidirectional associations over time. Int J Environ Res Public Health. 2022;19(23):16065. 10.3390/ijerph192316065. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 51.Ai F, Li E, Dong A, Zhang H. Association between disability and cognitive function in older Chinese people: a moderated mediation of social relationships and depressive symptoms. Front Public Health. 2024;12:1354877. 10.3389/fpubh.2024.1354877. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 52.Choi NG, Zhou Y, Marti CN. Possible and probable dementia and depressive/anxiety symptoms: mediation effects of social engagement and loneliness. Res Aging. 2025;48(5–6):299–311. 10.1177/01640275251380531. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 53.Ditzen B, Heinrichs M. Psychobiology of social support: the social dimension of stress buffering. Restor Neurol Neurosci. 2014;32(1):149–62. 10.3233/RNN-139008. [DOI] [PubMed] [Google Scholar]
- 54.Halahakoon DC, Lewis G, Roiser JP. Cognitive impairment and depression-cause, consequence, or coincidence? JAMA Psychiat. 2019;76(3):239–40. [DOI] [PubMed] [Google Scholar]
- 55.Li LW, Conwell Y. Effects of changes in depressive symptoms and cognitive functioning on physical disability in home care elders. J Gerontol A Biol Sci Med Sci. 2009;64(2):230–6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 56.Kong LL, Xie W, Dong ZY, et al. Cognitive frailty and its association with disability among Chinese community-dwelling older adults: a cross-sectional study. BMC Geriatr. 2024;24(1):189. 10.1186/s12877-024-04773-0. [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
Data Availability Statement
All of our research data are available from the CHARLS(http://charls.pku.edu.cn) datasets.