Abstract
Background
Combined vision and hearing impairment, referred to as dual sensory impairment (DSI), has been associated with depressive symptoms, but it remains unclear whether depressive symptoms worsen mainly after DSI is first observed or whether worsening is already evident beforehand.
Methods
We used data from the China Health and Retirement Longitudinal Study (CHARLS) from 2011 to 2020. A total of 5,210 middle-aged and older adults who were free of DSI and significant depressive symptoms at baseline were included. Depressive symptoms were assessed using the 10-item Center for Epidemiologic Studies Depression Scale (CESD-10). Piecewise linear mixed-effects models were used to examine depressive symptom trajectories before and after DSI was first observed. Multiple imputation was used for missing baseline covariates. Models were adjusted for sociodemographic factors, health behaviors, chronic conditions, and body mass index.
Results
During follow-up, 3,371 participants were first observed with DSI, whereas 1,839 remained free of DSI. In fully adjusted models, participants who were first observed with DSI during follow-up had higher baseline CESD-10 scores (β = 0.777, 95% confidence interval [CI]: 0.526 to 1.028) and showed a faster increase in depressive symptoms before DSI was first observed than those who remained free of DSI (β = 0.166, 95% CI: 0.077 to 0.255). After DSI was first observed, this group continued to show a more unfavorable depressive symptom trajectory (β = 0.149, 95% CI: 0.090 to 0.209). However, the slope difference did not further increase thereafter (β = -0.017, 95% CI: -0.148 to 0.115).
Conclusions
Participants who were first observed with DSI during follow-up showed faster worsening of depressive symptoms before DSI was first observed, rather than only afterward. These findings suggest that the association between DSI and depressive symptoms may partly reflect shared aging-related risk processes that begin earlier.
Clinical trial number
Not applicable.
Supplementary Information
The online version contains supplementary material available at https://doi.org/10.1186/s40359-026-05481-7.
Keywords: Dual sensory impairment, Vision impairment, Hearing impairment, Depressive symptoms, Trajectory, CHARLS
Introduction
Depressive symptoms are among the most common and burdensome mental health problems in middle-aged and older adults. Late-life depression not only impairs quality of life, but is also closely related to functional limitations, poor chronic disease management, cognitive decline, greater health care use, and increased mortality [1–3]. In later life, depressive symptoms rarely occur as an isolated psychological problem. They are often intertwined with sensory decline, physical impairment, cognitive deterioration, chronic disease burden, and reduced social participation [4, 5]. Understanding how depressive symptoms evolve over time in the broader process of aging-related health decline may therefore help identify people at higher risk of poor mental health and clarify whether depressive symptoms are primarily a consequence of a specific health event or part of a more systemic deterioration that begins earlier.
Sensory decline is a common but often underestimated component of aging-related health deterioration. Vision and hearing impairments may limit access to environmental information, weaken communication, reduce social engagement, and increase loneliness and functional dependence, all of which can have sustained effects on mental health [6–8]. A large body of evidence has shown that either vision impairment (VI) or hearing impairment (HI) alone is associated with a higher risk of depressive symptoms [9–12]. Dual sensory impairment (DSI), defined as the co-occurrence of VI and HI, may impose a greater psychological and functional burden because individuals face concurrent losses in visual input and auditory communication [13, 14]. DSI may therefore represent not only an accumulated state of sensory impairment, but also a marker of broader vulnerability in later life.
Previous studies have consistently supported an association between DSI and depressive symptoms. Evidence from aging cohorts such as the China Health and Retirement Longitudinal Study (CHARLS), the National Health and Aging Trends Study (NHATS), and the Tromsø Study suggests that people with DSI tend to have higher levels of depressive symptoms and a greater risk of incident or persistent depression during follow-up [10–13]. These studies have provided important evidence linking sensory function with mental health and highlight the clinical value of jointly considering vision, hearing, and depressive symptoms in older adults. However, most existing studies have primarily asked whether sensory impairment predicts subsequent depression risk, with much less attention to how depressive symptoms change before and after sensory impairment becomes apparent.
This distinction is not merely semantic; it is central to how the association between DSI and depression should be interpreted. Much of the existing literature implicitly follows a temporal framework in which DSI occurs first and then increases depression risk through sensory deprivation, communication difficulties, social isolation, or functional dependence [6–8, 13]. If this framework is correct, depressive symptoms would be expected to accelerate mainly after DSI emerges. Conversely, if depressive symptoms are already worsening faster before DSI is first observed, the association may not reflect a simple exposure-outcome relationship. Instead, it may partly reflect shared vulnerability or common aging-related processes that begin earlier, such as chronic disease burden, vascular or neurodegenerative changes, inflammation, functional decline, adverse health behaviors, and reduced social participation [5, 15, 16]. Identifying the timing of depressive symptom worsening relative to when DSI is first observed is therefore critical for distinguishing a consequence of sensory impairment from a marker of shared deterioration.
Recent work on DSI and cognitive function has begun to use pre- and post-onset trajectory designs to compare the rate of cognitive change before and after DSI is first observed [17]. This approach does more than test whether DSI is associated with an adverse health outcome; it also helps determine whether worsening of that outcome occurs after DSI is first observed or is already present beforehand. For cognitive decline, this temporal pattern can help separate the potential consequences of sensory deprivation from shared pathological processes. Depressive symptoms can be reconsidered within the same framework. They may be a psychological response to sensory decline, but they may also be an early manifestation of increasing overall health vulnerability. To date, however, longitudinal evidence on depressive symptom trajectories before and after DSI was first observed remains limited, leaving room for existing studies to overemphasize DSI as an antecedent risk factor for depression.
Against this background, we used data from CHARLS from 2011 to 2020 and included middle-aged and older adults who were free of DSI and significant depressive symptoms at baseline. We treated DSI as a first observed event during follow-up and used piecewise linear mixed-effects models to examine depressive symptom trajectories before and after that event. Unlike previous studies that mainly estimated the effect of DSI on subsequent depression risk, this study focused on whether depressive symptom worsening emerges after DSI is first observed or is already present beforehand. Through this design, we aimed to evaluate whether DSI is better understood as the starting point of accelerated depressive symptom worsening or as a clinical marker that co-occurs with depressive symptom worsening within a broader process of health deterioration.
Methods
Data source and study population
This study used data from five waves of CHARLS conducted in 2011, 2013, 2015, 2018, and 2020. CHARLS is a nationally representative longitudinal survey that collects information on demographic characteristics, family structure, health status, health care use, economic conditions, and social participation among middle-aged and older adults in China. The baseline survey was conducted in 2011 and enrolled residents aged 45 years or older from multiple provinces, with repeated follow-up surveys conducted in subsequent years.
We conducted a longitudinal cohort study to examine depressive symptom trajectories before and after DSI was first observed among middle-aged and older adults who were free of DSI and significant depressive symptoms at baseline. The participant selection process is shown in Fig. 1. Starting from the 2011 CHARLS baseline sample, we sequentially excluded participants with missing baseline sensory function or 10-item Center for Epidemiologic Studies Depression Scale (CESD-10) information, missing age or age younger than 45 years, missing baseline DSI status, prevalent DSI at baseline, missing CESD-10 scores both at baseline and across all follow-up waves, and no follow-up sensory function assessment. We further excluded participants with missing baseline CESD-10 scores or baseline CESD-10 scores ≥ 10. The final analytic sample included 5,210 participants, of whom 3,371 were first observed with DSI during follow-up and 1,839 remained free of DSI throughout follow-up.
Fig. 1.
Flowchart of participant selection
Definition of DSI
Sensory function was defined using self-reported vision and hearing questions in CHARLS. Vision was assessed using questions on both distance and near vision, and hearing was assessed using a self-reported hearing question. Response options for these items were excellent, very good, good, fair, and poor. Consistent with previous studies, VI was defined as reporting fair or poor distance or near vision, and HI was defined as reporting fair or poor hearing. DSI was defined as the presence of both VI and HI in the same survey wave [17].
Incident DSI was defined using the longitudinal sequence of sensory assessments. All included participants were free of DSI at the 2011 baseline. We then assessed whether participants first met the definition of DSI in 2013, 2015, or 2018. Participants first observed with DSI in 2013 were assigned a first-observed DSI time of 2 years after baseline; those first observed with DSI in 2015 after remaining free of DSI in 2013 were assigned a first-observed DSI time of 4 years; and those first observed with DSI in 2018 after remaining free of DSI in both 2013 and 2015 were assigned a first-observed DSI time of 7 years. Participants who were never observed to have DSI during follow-up were classified as the never DSI group. Because this study focused on depressive symptom trajectories before and after DSI was first observed, participants first observed with DSI in 2020 could not contribute depressive symptom data after that time point. Therefore, the 2020 wave was not used to define incident DSI.
Assessment of depressive symptoms
Depressive symptoms were assessed using the CESD-10. The scale includes 10 items, with total scores ranging from 0 to 30; higher scores indicate more severe depressive symptoms. Consistent with commonly used criteria in CHARLS-based studies, a CESD-10 score of 10 or higher was used to define significant depressive symptoms [18]. In the main analysis, participants with baseline CESD-10 scores ≥ 10 were excluded to examine depressive symptom trajectories before and after DSI was first observed among those without significant depressive symptoms at baseline. The longitudinal outcome was the continuous CESD-10 score measured in 2011, 2013, 2015, 2018, and 2020.
Covariates
Baseline covariates were selected a priori based on previous studies and their plausible associations with both sensory impairment and depressive symptoms. Age, sex, education, and marital status represented sociodemographic factors; smoking and alcohol drinking represented health behaviors; and hypertension, diabetes, heart disease, chronic lung disease, and body mass index (BMI) represented baseline health status and general health vulnerability. These variables were included to account for measured baseline differences between the groups. The assumed role and rationale for adjustment of each covariate are summarized in Supplementary Table 1. Sex was classified as male or female. Education was categorized as high school or above versus below high school, and marital status as married or unmarried. Smoking and alcohol drinking were defined using baseline self-reported information. Chronic conditions included hypertension, diabetes, heart disease, and chronic lung disease, all based on self-reported physician diagnoses at baseline. BMI was calculated from baseline measured height and weight.
Statistical analysis
Baseline characteristics of the final analytic sample were summarized and compared according to whether participants were first observed with DSI during follow-up. To assess potential selection, we also compared baseline sociodemographic, behavioral, and health characteristics between included and excluded participants aged 45 years or older. Continuous variables were presented as means and standard deviations, and categorical variables as counts and percentages. Between-group differences were assessed using Wilcoxon rank-sum tests for continuous variables and chi-square tests or Fisher’s exact tests for categorical variables.
To evaluate depressive symptom trajectories before and after first observed DSI, we used piecewise linear mixed-effects models. The continuous CESD-10 score was modeled as the outcome, with a random intercept for each participant to account for within-person correlation across repeated measurements. Time was calculated from the 2011 baseline, with the 2011, 2013, 2015, 2018, and 2020 waves corresponding to 0, 2, 4, 7, and 9 years after baseline, respectively. For participants first observed with DSI during follow-up, the survey wave in which they first met the DSI definition was set as event time 0. CESD-10 measurements obtained before or at this time point were treated as pre-onset/onset observations, whereas measurements obtained after this time point were treated as post-onset observations. Event time was calculated as the number of years from baseline at each survey wave minus the number of years from baseline when DSI was first observed. For participants who were never observed to have DSI during follow-up, the mean first-observed DSI time among participants with incident DSI was used as the reference time in the main analysis, allowing depressive symptom trajectories before and after this reference time to be estimated on a comparable time scale. In the model, pre-onset time was defined as the smaller value between years since baseline at each wave and the first-observed DSI time, whereas post-onset time was set to 0 before and at the time DSI was first observed and represented the number of years elapsed thereafter. The model included DSI group, pre-onset time, post-onset time, and interaction terms between DSI group and pre- or post-onset time. No separate immediate level-change term was included; therefore, the trajectory was assumed to be continuous at the time DSI was first observed. This piecewise model allowed the rate of change in depressive symptoms to be estimated separately before and after DSI was first observed.
Four parameters were of primary interest. The first was the baseline difference in CESD-10 scores between the incident DSI and never DSI groups. The second was the difference in CESD-10 slope before DSI was first observed between these two groups. The third was the slope difference after DSI was first observed. The fourth was the change in slope difference after versus before DSI was first observed, representing whether the DSI-related excess slope increased further after DSI was first observed. In this analysis, pre- and post-onset labels were anchored to the first observed DSI wave rather than the true clinical onset of DSI. This parameter was used to assess whether depressive symptom worsening was mainly expressed as additional post-onset acceleration. The main analysis reported the fully adjusted model. Covariates included age, sex, education, marital status, smoking, alcohol drinking, hypertension, diabetes, heart disease, chronic lung disease, and BMI. Missing baseline covariates were handled using multiple imputation with 10 imputed datasets and 10 iterations. Multiple imputation was used only for missing baseline covariates, not for longitudinal CESD-10 outcomes; the mixed-effects models used all available repeated CESD-10 measurements. Models were fitted separately in each imputed dataset, and estimates and standard errors were combined using Rubin’s rules. For visual comparison, we also plotted trajectories from an unadjusted piecewise linear mixed-effects model with the same time structure and random-intercept specification but without baseline covariates. To further document data availability and imputation, CESD-10 observation counts by wave and timing, the number of valid CESD-10 observations per participant, and baseline covariate missingness and distributions before and after multiple imputation are provided in Supplementary Tables 2–4.
We conducted twelve sensitivity analyses to assess the robustness of the findings. First, we repeated the main model among participants with complete covariate data to evaluate the influence of multiple imputation. Second, we repeated the analysis in a broader cohort without excluding participants with baseline CESD-10 scores ≥ 10, to examine the influence of excluding baseline depressive symptoms. Third, to address uncertainty related to the timing of first observed DSI, we repeated the main analysis after stratifying participants with incident DSI by the wave in which DSI was first observed (2013, 2015, or 2018). Fourth, to evaluate whether the results depended on the reference-time assignment for the never DSI group, we repeated the analysis using alternative reference times of 4 and 7 years for participants who were never observed with DSI. Fifth, to account for heterogeneity in individual depressive symptom trajectories, we fitted an alternative model including participant-specific random intercepts and random slopes for calendar time. Sixth, to assess whether the findings were sensitive to the bounded and right-skewed distribution of CESD-10 scores, we repeated the analysis using log-transformed CESD-10 scores [log(CESD-10 + 1)] as the outcome. Seventh, we repeated the analysis using a stricter DSI definition requiring poor distance or near vision and poor hearing. Eighth, we required DSI to be present at two consecutive waves; participants first observed with DSI in 2018 or without an available confirmation assessment were excluded from this analysis. Ninth, we repeated the main analysis after excluding CESD-10 observations from the 2020 wave. Tenth, to examine whether the findings differed according to baseline sensory status, we stratified participants into those with no sensory impairment and those with single sensory impairment at baseline and repeated the main analysis within each stratum, comparing participants with incident DSI with never-DSI participants who had the same baseline sensory status. Eleventh, we repeated the analysis among participants with complete CESD-10 data across all five waves. Twelfth, we fitted an inverse-probability-of-observation weighted model, with stabilized weights truncated at the 1st and 99th percentiles.
In addition, we conducted a supplementary comparator analysis to examine whether the trajectory pattern related to DSI differed from that related to single sensory impairment. This analysis was restricted to participants without VI or HI at baseline. According to the first observed sensory impairment type during follow-up, participants were classified as incident HI-only, incident VI-only, incident DSI, or never sensory impairment. Each incident sensory impairment group was compared separately with the never sensory impairment group using the same piecewise linear mixed-effects model and covariate adjustment strategy as in the main analysis. All analyses were performed using R software. Two-sided P values < 0.05 were considered statistically significant.
Results
Baseline characteristics
The final analytic sample included 5,210 middle-aged and older adults who were free of DSI and significant depressive symptoms at baseline. Among them, 3,371 were first observed with DSI during follow-up, whereas 1,839 were never observed with DSI. Among participants with incident DSI, 1,692 were first observed with DSI in 2013, 1,094 in 2015, and 585 in 2018.
Table 1 shows the baseline characteristics of the study population. The mean age was 57.29 years, 53.2% of participants were men, and the mean baseline CESD-10 score was 4.07. Compared with participants who remained free of DSI throughout follow-up, those who were first observed with DSI during follow-up were older, more likely to be women, had lower educational attainment, had slightly lower BMI, and had higher baseline CESD-10 scores. Specifically, the mean baseline CESD-10 score was 4.26 in the incident DSI group and 3.73 in the never DSI group. Differences in marital status, smoking, alcohol drinking, hypertension, diabetes, heart disease, and chronic lung disease did not reach statistical significance.
Table 1.
Baseline characteristics of participants by incident DSI status
| Characteristic | Overall n = 5,210 |
Incident DSI n = 3,371 |
Never DSI n = 1,839 |
p-value |
|---|---|---|---|---|
| Age (years) | 57.29 ± 8.96 | 57.70 ± 8.93 | 56.55 ± 8.98 | < 0.001 |
| Sex | 0.044 | |||
| Female | 2,434 (46.8%) | 1,610 (47.8%) | 824 (44.9%) | |
| Male | 2,771 (53.2%) | 1,759 (52.2%) | 1,012 (55.1%) | |
| Education | < 0.001 | |||
| Below high school | 4,238 (81.4%) | 2,889 (85.8%) | 1,349 (73.4%) | |
| High school or above | 970 (18.6%) | 480 (14.2%) | 490 (26.6%) | |
| Marital status | 0.218 | |||
| Married | 4,732 (90.8%) | 3,074 (91.2%) | 1,658 (90.2%) | |
| Unmarried | 478 (9.2%) | 297 (8.8%) | 181 (9.8%) | |
| Current smoking | 0.113 | |||
| Yes | 2,136 (41.0%) | 1,355 (40.2%) | 781 (42.5%) | |
| No | 3,073 (59.0%) | 2,015 (59.8%) | 1,058 (57.5%) | |
| Current drinking | 0.120 | |||
| Yes | 1,938 (37.2%) | 1,228 (36.4%) | 710 (38.6%) | |
| No | 3,272 (62.8%) | 2,143 (63.6%) | 1,129 (61.4%) | |
| Hypertension | 0.358 | |||
| Yes | 1,078 (20.7%) | 710 (21.1%) | 368 (20.0%) | |
| No | 4,119 (79.3%) | 2,651 (78.9%) | 1,468 (80.0%) | |
| Diabetes | 0.552 | |||
| Yes | 237 (4.6%) | 149 (4.5%) | 88 (4.8%) | |
| No | 4,932 (95.4%) | 3,194 (95.5%) | 1,738 (95.2%) | |
| Heart disease | 0.693 | |||
| Yes | 375 (7.2%) | 246 (7.3%) | 129 (7.0%) | |
| No | 4,815 (92.8%) | 3,110 (92.7%) | 1,705 (93.0%) | |
| Chronic lung disease | 0.869 | |||
| Yes | 330 (6.4%) | 212 (6.3%) | 118 (6.4%) | |
| No | 4,863 (93.6%) | 3,146 (93.7%) | 1,717 (93.6%) | |
| BMI (kg/m²) | 23.84 ± 3.78 | 23.75 ± 3.79 | 24.00 ± 3.77 | 0.023 |
| Baseline CESD-10 score | 4.07 ± 2.77 | 4.26 ± 2.77 | 3.73 ± 2.74 | < 0.001 |
Values are presented as mean ± SD or n (%). Incident DSI was defined as DSI first observed during follow-up among participants without DSI at baseline. DSI, dual sensory impairment; CESD-10, 10-item Center for Epidemiologic Studies Depression Scale. P values were calculated using Wilcoxon rank-sum tests for continuous variables and chi-square or Fisher exact tests for categorical variables. Percentages were calculated among participants with non-missing data; totals may vary because of missing covariate data
Included participants were younger, better educated, more likely to be married, and had fewer chronic conditions than excluded participants (Supplementary Table 5).
Depressive symptom trajectories around the first observation of DSI
Figure 2 shows the unadjusted (A) and fully adjusted (B) predicted CESD-10 trajectories. The two models showed similar temporal patterns: participants in the incident DSI group already had higher CESD-10 scores and faster increases before DSI was first observed, and their scores remained higher thereafter.
Fig. 2.
Unadjusted and fully adjusted CESD-10 trajectories before and after DSI was first observed. A shows predicted trajectories from the unadjusted piecewise linear mixed-effects model, and (B) shows trajectories from the fully adjusted model after multiple imputation of baseline covariates. The red line represents participants with incident DSI (n = 3,371), and the blue line represents participants with never DSI (n = 1,839). The vertical dotted line indicates the time when DSI was first observed. Both models used the same piecewise time structure and included a participant-specific random intercept. The fully adjusted model additionally included age, sex, education, marital status, current smoking, current drinking, hypertension, diabetes, heart disease, chronic lung disease, and BMI. Shaded areas indicate 95% confidence intervals. DSI, dual sensory impairment; CESD-10, 10-item Center for Epidemiologic Studies Depression Scale; BMI, body mass index
Table 2 presents the results from the fully adjusted linear mixed-effects models after multiple imputation. Compared with the never DSI group, participants in the incident DSI group had higher baseline CESD-10 scores (β = 0.777, 95% CI: 0.526 to 1.028, P < 0.001). Before DSI was first observed, their CESD-10 scores increased significantly faster than those of the never DSI group (β = 0.166, 95% CI: 0.077 to 0.255, P < 0.001). After DSI was first observed, the incident DSI group continued to show a faster increase in CESD-10 scores than the never DSI group (β = 0.149, 95% CI: 0.090 to 0.209, P < 0.001).
Table 2.
Fully adjusted associations of incident DSI with depressive symptom trajectories
| Parameter | β | 95% CI | p-value |
|---|---|---|---|
| Baseline difference: DSI vs. never DSI | 0.777 | 0.526, 1.028 | < 0.001 |
| Pre-onset slope difference: DSI vs. never DSI | 0.166 | 0.077, 0.255 | < 0.001 |
| Post-onset slope difference: DSI vs. never DSI | 0.149 | 0.090, 0.209 | < 0.001 |
| Change in slope difference after versus before first observed DSI | -0.017 | -0.148, 0.115 | 0.803 |
Values are pooled regression coefficients (β) and 95% confidence intervals from fully adjusted linear mixed-effects models across 10 imputed datasets. Models included a random intercept for each participant and adjusted for age, sex, education, marital status, current smoking, current drinking, hypertension, diabetes, heart disease, chronic lung disease, and BMI. The change in slope difference after versus before first observed DSI represents whether the DSI-related slope difference increased after first observed DSI compared with the pre-onset period. DSI, dual sensory impairment; CESD-10, 10-item Center for Epidemiologic Studies Depression Scale; CI, confidence interval
Further analysis showed that the between-group slope difference did not further increase after DSI was first observed compared with the pre-onset period (β = -0.017, 95% CI: -0.148 to 0.115, P = 0.803). This finding suggests that the less favorable depressive symptom trajectory in the incident DSI group was already present before DSI was first observed and persisted afterward, with no evidence of additional acceleration after DSI was first observed.
Sensitivity and supplementary analyses
The complete-case analysis showed that participants in the incident DSI group had higher baseline CESD-10 scores, significant between-group slope differences both before and after DSI was first observed, and no further increase in the slope difference after DSI was first observed compared with the pre-onset period (Supplementary Table 6). In the broader cohort that included participants with baseline CESD-10 scores ≥ 10, the incident DSI group still had higher baseline CESD-10 scores and a faster pre-onset increase in depressive symptoms, although the between-group slope difference after DSI was first observed was not statistically significant (Supplementary Table 7). This finding further suggests that the less favorable depressive symptom trajectory in the incident DSI group did not mainly emerge after DSI was first observed. In analyses stratified by the wave in which DSI was first observed, the faster pre-onset increase in CESD-10 scores remained evident among participants first observed with DSI in 2013, 2015, and 2018, whereas post-onset slope differences were not statistically significant in any stratum (Supplementary Table 8). Thus, the main temporal pattern was not driven by a single wave in which DSI was first observed. Analyses using alternative reference times of 4 and 7 years for the never DSI group also yielded results consistent with the main analysis (Supplementary Tables 9 and 10).
Results were also similar when a random-slope model was used to account for heterogeneity in individual depressive symptom trajectories (Supplementary Table 11). Analyses using poor-only and two-consecutive-wave DSI definitions yielded significant pre-onset slope differences (Supplementary Tables 13 and 14). In these analyses, as in the random-slope and log-transformed outcome models, the change in slope difference after versus before DSI was first observed was significantly negative, suggesting attenuation rather than additional acceleration (Supplementary Tables 11–14). Results were similar after excluding CESD-10 observations from the 2020 wave (Supplementary Table 15). Among participants with incident DSI, 1,084 had no sensory impairment at baseline, whereas 2,287 had baseline single sensory impairment, including 1,858 with VI-only and 429 with HI-only (Supplementary Table 16). Stratified analyses by baseline sensory status supported the main interpretation, showing faster increases in CESD-10 scores before DSI was first observed and no evidence of additional acceleration thereafter in either stratum (Supplementary Table 17). Wave-specific CESD-10 availability and missingness by DSI status are shown in Supplementary Table 18. Complete five-wave and inverse-probability-of-observation weighted analyses supported the main interpretation (Supplementary Table 19). Sex-interaction analysis showed no statistically significant interactions for the main DSI-related trajectory parameters (Supplementary Table 20).
Finally, in a supplementary comparator analysis restricted to participants without any sensory impairment at baseline, we compared those who were first observed with HI-only, VI-only, or DSI during follow-up with those who remained free of sensory impairment. Participants who were first observed with DSI showed a significantly faster pre-onset increase in CESD-10 scores (β = 0.461, 95% CI: 0.168 to 0.754, P = 0.002), whereas similar pre-onset slope differences were not observed among those who were first observed with HI-only (β = 0.098, 95% CI: -0.185 to 0.381, P = 0.497) or VI-only (β = 0.139, 95% CI: -0.111 to 0.389, P = 0.275) (Supplementary Table 21). The corresponding predicted trajectory plot showed a more pronounced separation of CESD-10 trajectories before incident DSI than before incident single sensory impairment (Supplementary Fig. 1). Together, these findings indicate that a faster increase in depressive symptoms before DSI was first observed was more evident in the incident DSI group.
Discussion
Using longitudinal data from CHARLS from 2011 to 2020, this study examined depressive symptom trajectories before and after DSI was first observed among middle-aged and older adults in China. We found that participants in the incident DSI group already showed a faster increase in depressive symptoms before DSI was first observed. After DSI was first observed, this unfavorable trajectory persisted. However, the slope difference after DSI was first observed did not further increase compared with the pre-onset period. Overall, these findings suggest that depressive symptom trajectories among participants in the incident DSI group had already diverged before DSI was first observed. The magnitude of this trajectory difference was modest. For example, the pre-onset slope difference of 0.166 points per year corresponds to approximately 0.66 additional CESD-10 points over a 4-year interval on average. Thus, this finding should be interpreted as a mean trajectory difference and an early risk signal at the population level, rather than as an individual transition across the threshold for significant depressive symptoms.
This temporal pattern differs from the one-way interpretation often implied in previous studies. Prior work on sensory impairment and depression has largely emphasized DSI as a risk factor for depressive symptoms [10, 19, 20]. This interpretation is biologically and psychosocially plausible: concurrent VI and HI may reduce access to environmental information and disrupt social communication, thereby increasing loneliness, functional dependence, and emotional distress [20, 21]. However, because these studies often treated sensory impairment as a baseline exposure or a follow-up risk state, they tended to present a framework in which DSI precedes depression. Interestingly, contrary to this conventional expectation, our trajectory analysis showed that participants in the incident DSI group already had a faster increase in CESD-10 scores before DSI was first observed. In other words, the unfavorable depressive symptom trajectory related to DSI did not suddenly emerge after DSI became observable, but may have begun earlier. The association between DSI and depression observed in previous studies may therefore partly reflect pre-existing mental health deterioration or systemic health decline before DSI was first observed, rather than only the direct psychological consequences of DSI [22, 23].
The trajectory analysis also showed no additional increase in slope after DSI was first observed. That is, although the unfavorable depressive symptom trajectory among participants in the incident DSI group persisted thereafter, it did not show further acceleration after DSI became observable. The random-slope and log-transformed outcome models further suggested attenuation rather than additional acceleration after DSI was first observed. This attenuation may have several explanations, including possible ceiling-related constraints on further increases among participants with relatively elevated CESD-10 scores, psychological adaptation, or increased family, clinical, or social support after sensory difficulties become apparent. Worsening depressive symptoms may be an early manifestation of co-evolving sensory decline, functional vulnerability, and reduced social participation [20, 21, 23]. DSI may therefore be less a discrete external exposure than a clinical expression of accumulated aging-related vulnerability [24, 25]. At the biological level, this shared vulnerability may involve chronic low-grade inflammation, vascular aging and microvascular brain injury, and sleep-related neurovascular and glymphatic regulation. These processes may affect peripheral sensory organs, central sensory processing, emotional regulation, and physical function, thereby creating conditions in which sensory decline and depressive symptoms worsen together [26–29]. From this perspective, depressive symptoms may not simply lag behind DSI, but may appear as an early signal of increasing vulnerability before sensory decline reaches the threshold for DSI. These pathways were not directly measured in CHARLS. Future cohort studies should combine repeated objective sensory assessments with longitudinal measures of depressive symptoms, frailty, cognition, and social participation, as well as biological markers, vascular assessments, neuroimaging, and sleep-related indicators, while allowing potential sex-related and life-course heterogeneity to be evaluated. Such multidimensional designs could help distinguish shared determinants from mediating pathways and support earlier risk stratification. A conceptual overview of these potential pathways and priorities for future longitudinal research is provided in Supplementary Table 22.
The comparator analysis using single sensory impairment provided further support for this interpretation. Among participants without any sensory impairment at baseline, pre-onset acceleration in depressive symptoms was mainly observed among those who were first observed with DSI, but not among those who were first observed with HI-only or VI-only. The corresponding trajectory plot also showed a more pronounced separation of depressive symptom trajectories before incident DSI. This suggests that the unfavorable pre-onset depressive trajectory before DSI was not a nonspecific pattern preceding all forms of sensory impairment, but may more likely reflect the greater accumulated sensory burden and overall vulnerability represented by DSI. Previous studies have also suggested that, compared with single sensory impairment, DSI may be associated with greater functional decline, cognitive decline, and other adverse aging-related outcomes [30–33]. Therefore, the clinical importance of DSI may lie not only in its association with depressive symptom risk, but also in the possibility that its emergence indicates earlier mental health risk and broader health vulnerability. In eye-care or hearing-related clinical settings, gradual worsening of depressive symptoms in middle-aged and older adults may serve as an early risk signal for coexisting or emerging sensory vulnerability. More broadly, loneliness, pain-related symptoms such as headache and low back pain, increasing or excessive analgesic use, and, among women, menstrual irregularity, premenstrual symptoms, and hot flashes may represent potential correlates or early warning features of depressive symptoms [34–39], although their temporal and causal roles require confirmation in longitudinal studies. When visual or hearing impairment is identified, especially among those with single sensory impairment who may later progress to DSI, mental health assessment and early intervention should not wait until dual impairment is clearly present. Sensory function assessment and depressive symptom screening should be integrated earlier into geriatric health assessment to help identify coexisting sensory decline and psychological vulnerability.
This study has several strengths. First, we used multiple waves of nationally representative longitudinal data from CHARLS, allowing us to examine changes in sensory function and depressive symptoms over an extended period among middle-aged and older adults in China. Second, by focusing on participants who were free of DSI and significant depressive symptoms at baseline, we reduced the influence of pre-existing DSI and significant depressive symptoms when evaluating temporal patterns. Third, we used piecewise linear mixed-effects models and treated the first observed DSI wave as the event time, allowing depressive symptom trajectories before and after DSI was first observed to be compared. This pre- and post-onset trajectory design has been applied to the study of DSI and cognitive decline, and provided a methodological basis for re-examining the temporal pattern between DSI and depressive symptoms. Fourth, multiple sensitivity analyses and the comparator analysis using single sensory impairment supported the robustness and specificity of the main findings.
Several limitations should also be noted. First, VI and HI were defined using self-reported information, which may introduce information bias or misclassification. Depressive symptoms may lower the threshold for reporting sensory difficulties and influence individuals’ perceptions of health and sensory function; if participants with worsening depressive symptoms were more likely to report VI or HI, such differential reporting could exaggerate the difference in depressive symptom trajectories before DSI was first observed. Second, because CHARLS was conducted at discrete follow-up waves, we could only identify the survey wave in which DSI was first observed, rather than the exact timing at which participants transitioned from non-DSI to meeting the DSI definition. We therefore used the term “first observed” rather than true onset. Although the onset-wave stratified sensitivity analysis supported the main temporal pattern, uncertainty in the timing of this transition should still be considered when interpreting the pre- and post-onset slopes. If the true transition to DSI occurred between waves before it was first observed, some observations classified as pre-onset may have occurred after the actual transition, potentially inflating pre-onset differences and attenuating post-onset changes. Moreover, the pre-onset worsening of depressive symptoms may partly reflect subclinical or progressive sensory decline that had not yet reached the study definition of DSI. Therefore, our findings cannot fully distinguish a shared co-evolutionary process from an earlier sensory-first process that was not captured by the study definition. The findings should therefore not be interpreted as indicating that sensory function was completely normal before DSI was first observed. Rather, mental health and sensory function may already have been worsening together before DSI became clearly identifiable. Third, CESD-10 assesses depressive symptoms rather than clinically diagnosed depressive disorder, so the findings should not be interpreted directly as risks of clinical depression. Fourth, the relatively healthier analytic sample may limit the generalizability of the findings to less healthy populations. Although the sensitivity analyses yielded consistent results, informative missingness related to health deterioration or mortality could not be fully addressed. Finally, we adjusted only for baseline covariates and did not account for time-varying health, functional, cognitive, social, sleep-related, pain-related, or medication-use factors. These factors may reflect shared vulnerability underlying both sensory decline and worsening depressive symptoms. Accounting for their longitudinal trajectories might attenuate the observed pre-onset slope difference or help clarify whether they act as shared determinants or mediating processes. Therefore, the findings should not be interpreted as a simple exposure–outcome effect.
In conclusion, among middle-aged and older Chinese adults who were free of DSI and significant depressive symptoms at baseline, those who were first observed with DSI during follow-up already showed faster worsening of depressive symptoms before DSI was first observed. This unfavorable trajectory persisted after DSI was first observed, but there was no evidence of additional acceleration thereafter. This pattern appeared more pronounced for DSI than for single sensory impairment. These findings suggest that the first observation of DSI may not mark the starting point of depressive symptom worsening, but may instead represent an observable marker of earlier mental health risk and broader vulnerability.
Supplementary Information
Below is the link to the electronic supplementary material.
Acknowledgements
AcknowledgmentsThe authors would like to acknowledge the participants and investigators of the China Health and Retirement Longitudinal Study (CHARLS).
Author contributions
Author ContributionsDaolei Wu contributed to statistical analyses and validation, interpretation of the findings, and substantive manuscript revision. Limin Fu participated in data collection and the initial drafting and editing of the manuscript. Jionghao Guo, Zhaoli Lei, and Jiajing Wu participated in data analysis and manuscript revision. Yang Li participated in data collection and manuscript preparation, provided methodological guidance, supervised the study, and reviewed and finalized the manuscript. All authors read and approved the final manuscript.
Funding
This study received no external funding.
Data availability
Data Availability StatementThe data supporting the findings of this study are available on the CHARLS website (https://charls.pku.edu.cn/). This study was conducted in accordance with the Declaration of Helsinki. CHARLS was approved by the Institutional Review Board of Peking University (approval number: IRB00001052-11015 for the household survey and IRB00001052-11014 for blood samples).
Declarations
Consent for publication
Not applicable.
Supplementary Information
The separate supplementary file contains twenty-two supplementary tables (Supplementary Tables 1–22) and one supplementary figure (Supplementary Fig. 1).
Ethics Statement
This study was conducted in accordance with the Declaration of Helsinki. CHARLS was approved by the Institutional Review Board of Peking University (approval number: IRB00001052-11015 for the household survey and IRB00001052-11014 for blood samples).
Consent to participate
Written informed consent was obtained from all the participants.
Conflict of interest
The authors declare no conflicts of interest.
Competing interests
The authors declare no competing interests.
Footnotes
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
References
- 1.Rong J, Wang X, Cheng P, Li D, Zhao D. Global, regional and national burden of depressive disorders and attributable risk factors, from 1990 to 2021: results from the 2021 Global Burden of Disease study. Br J Psychiatry. 2025;227:688–97. 10.1192/bjp.2024.266. [DOI] [PubMed] [Google Scholar]
- 2.Wang Y, Qin C, Chen H, Liang W, Liu M, Liu J. Global, regional, and national burden of major depressive disorders in adults aged 60 years and older from 1990 to 2021, with projections of prevalence to 2050: Analyses from the Global Burden of Disease Study 2021. J Affect Disord. 2025;374:486–94. 10.1016/j.jad.2025.01.086. [DOI] [PubMed] [Google Scholar]
- 3.Szymkowicz SM, Gerlach AR, Homiack D, Taylor WD. Biological factors influencing depression in later life: role of aging processes and treatment implications. Transl Psychiatry. 2023;13:160. 10.1038/s41398-023-02464-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Ahrenfeldt LJ, Möller S, Nielsen DL, Kjær NK, Søndergaard J, Lykkegaard J. Sensory impairments and depressive symptoms in Europe: a cross-national cohort study. Aging Ment Health. 2024;28:1591–9. 10.1080/13607863.2024.2345790. [DOI] [PubMed] [Google Scholar]
- 5.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]
- 6.Wei J, Li Y, Gui X. Association of hearing loss and risk of depression: a systematic review and meta-analysis. Front Neurol. 2024;15:1446262. 10.3389/fneur.2024.1446262. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Lawrence BJ, Jayakody DMP, Bennett RJ, Eikelboom RH, Gasson N, Friedland PL. Hearing Loss and Depression in Older Adults: A Systematic Review and Meta-analysis. Gerontologist. 2020;60:e137–54. 10.1093/geront/gnz009. [DOI] [PubMed] [Google Scholar]
- 8.Virgili G, Parravano M, Petri D, Maurutto E, Menchini F, Lanzetta P, et al. The Association between Vision Impairment and Depression: A Systematic Review of Population-Based Studies. J Clin Med. 2022;11:2412. 10.3390/jcm11092412. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Ji X, Zhang Y. The impact of vision impairment on depressive symptoms among older adults: The mediating role of dispositional optimism. PLoS ONE. 2025;20:e0328053. 10.1371/journal.pone.0328053. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Xie T, Liu D, Guo J, Zhang B. The longitudinal effect of sensory loss on depression among Chinese older adults. J Affect Disord. 2021;283:216–22. 10.1016/j.jad.2021.01.081. [DOI] [PubMed] [Google Scholar]
- 11.Killeen OJ, Xiang X, Powell D, Reed NS, Deal JA, Swenor BK, et al. Longitudinal Associations of Self-Reported Visual, Hearing, and Dual Sensory Difficulties With Symptoms of Depression Among Older Adults in the United States. Front Neurosci. 2022;16:786244. 10.3389/fnins.2022.786244. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Simning A, Fox ML, Barnett SL, Sorensen S, Conwell Y. Depressive and Anxiety Symptoms in Older Adults With Auditory, Vision, and Dual Sensory Impairment. J Aging Health. 2019;31:1353–75. 10.1177/0898264318781123. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Cosh S, von Hanno T, Helmer C, Bertelsen G, Delcourt C, Schirmer H, et al. The association amongst visual, hearing, and dual sensory loss with depression and anxiety over 6 years: The Tromsø Study. Int J Geriatr Psychiatry. 2018;33:598–605. 10.1002/gps.4827. [DOI] [PubMed] [Google Scholar]
- 14.Wang J, Duan L, Zeng R, Xie F, Wu Z, Luo Y, et al. Change in depressive symptom scores to assess the risk of new-onset dual sensory impairment in middle-aged and older adults: a nationwide cohort study. Front Public Health. 2025;13:1520552. 10.3389/fpubh.2025.1520552. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Zhou S, Liang Z, Li Q, Song S, Wang Z, Xu M, et al. Association of cumulative average sensory impairments with cognitive function and depressive symptoms: Two prospective cohort studies. J Affect Disord. 2025;369:16–24. 10.1016/j.jad.2024.09.140. [DOI] [PubMed] [Google Scholar]
- 16.D’Onofrio AM, Ferrajoli GF, Balzoni LM, Massetti M, Zanzarri A, Marano G, et al. Late-Onset Depression in an Aging World: A Multidimensional Perspective on Risks, Mechanisms, and Treatment. Geriatr (Basel). 2026;11:13. 10.3390/geriatrics11010013. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Zou M, Gong H, Chen B, Chen S, Teng J, Zhou J, et al. Trajectories of cognitive decline before and after new-onset dual sensory impairment: findings from two longitudinal studies. Alzheimers Dement. 2026;22:e71333. 10.1002/alz.71333. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Zeng Y, Ke Z, Cai S, Zhuang X, Pan L, Liao L. Association of depression with sarcopenia and the impact of sleep: the first longitudinal evidence from the CHARLS. BMC Public Health. 2026;26:1232. 10.1186/s12889-026-26852-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Wang EB, Garcia Morales EE, Gross AL, Lin FR, Reed NS, Deal JA. Residential Differences and Depression Among Older Adults With Dual Sensory Loss. JAMA Otolaryngol Head Neck Surg. 2025;151:202–10. 10.1001/jamaoto.2024.4488. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Huang C, He X, Zhang X. A study on the mechanism of how sensory impairment affects depression in the elderly: the mediating roles of daily activity capability and social participation. Front Psychol. 2024;15:1410422. 10.3389/fpsyg.2024.1410422. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Huang AR, Cudjoe TKM, Rebok GW, Swenor BK, Deal JA. Hearing and vision impairment and social isolation over 8 years in community-dwelling older adults. BMC Public Health. 2024;24:779. 10.1186/s12889-024-17730-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Khalaila R, Grebe L, Allen IE. A Prospective Association Between Sensory Impairment and Cognitive Performance Among Older Community-Dwelling Adults: The Role of Depressive Symptoms. J Appl Gerontol. 2024;43:1997–2006. 10.1177/07334648241254362. [DOI] [PubMed] [Google Scholar]
- 23.Xin B, He Y, Wu Y, Zhang D, Liu J, Yang S, et al. Associations between sensory impairment and functional limitations among older Chinese adults: mediating roles of social isolation and cognition. BMC Geriatr. 2024;24:797. 10.1186/s12877-024-05383-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Janower A, St John P. Dual Sensory Impairment and Functional Status in a Prospective Cohort Study. Can Geriatr J. 2023;26:364–71. 10.5770/cgj.26.669. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Jang YS, Yun I, Park YS, Park E-C, Jang S-I. Association between changes in sensory impairment and functional disability in older adults: A longitudinal study. Arch Gerontol Geriatr. 2023;114:105108. 10.1016/j.archger.2023.105108. [DOI] [PubMed] [Google Scholar]
- 26.Furman D, Campisi J, Verdin E, Carrera-Bastos P, Targ S, Franceschi C, et al. Chronic inflammation in the etiology of disease across the life span. Nat Med. 2019;25:1822–32. 10.1038/s41591-019-0675-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Climie RE, Alastruey J, Mayer CC, Schwarz A, Laucyte-Cibulskiene A, Voicehovska J, et al. Vascular ageing: moving from bench towards bedside. Eur J Prev Cardiol. 2023;30:1101–17. 10.1093/eurjpc/zwad028. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Rensma SP, van Sloten TT, Launer LJ, Stehouwer CDA. Cerebral small vessel disease and risk of incident stroke, dementia and depression, and all-cause mortality: A systematic review and meta-analysis. Neurosci Biobehav Rev. 2018;90:164–73. 10.1016/j.neubiorev.2018.04.003. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Hauglund NL, Andersen M, Tokarska K, Radovanovic T, Kjaerby C, Sørensen FL, et al. Norepinephrine-mediated slow vasomotion drives glymphatic clearance during sleep. Cell. 2025;188:606–e62217. 10.1016/j.cell.2024.11.027. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Armstrong NM, Vieira Ligo Teixeira C, Gendron C, Brenowitz WD, Lin FR, Swenor B, et al. Associations of dual sensory impairment with incident mobility and ADL difficulty. J Am Geriatr Soc. 2022;70:1997–2007. 10.1111/jgs.17764. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Assi L, Ehrlich JR, Zhou Y, Huang A, Kasper J, Lin FR, et al. Self-reported dual sensory impairment, dementia, and functional limitations in Medicare beneficiaries. J Am Geriatr Soc. 2021;69:3557–67. 10.1111/jgs.17448. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Tomida K, Lee S, Bae S, Makino K, Chiba I, Harada K, et al. Association of dual sensory impairment with changes in life space: A longitudinal study with two-year follow-up. Maturitas. 2022;165:78–84. 10.1016/j.maturitas.2022.07.017. [DOI] [PubMed] [Google Scholar]
- 33.Maharani A, Dawes P, Nazroo J, Tampubolon G, Pendleton N. Sense-Cog WP1 group. Visual and hearing impairments are associated with cognitive decline in older people. Age Ageing. 2018;47:575–81. 10.1093/ageing/afy061. [DOI] [PubMed] [Google Scholar]
- 34.Lee SL, Pearce E, Ajnakina O, Johnson S, Lewis G, Mann F, et al. The association between loneliness and depressive symptoms among adults aged 50 years and older: a 12-year population-based cohort study. Lancet Psychiatry. 2021;8:48–57. 10.1016/S2215-0366(20)30383-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Chou K-L. Reciprocal relationship between pain and depression in older adults: evidence from the English Longitudinal Study of Ageing. J Affect Disord. 2007;102:115–23. 10.1016/j.jad.2006.12.013. [DOI] [PubMed] [Google Scholar]
- 36.Stubbs B, Chen T-Y, Vancampfort D, Wu M-S, Werneck AO, Firth J, et al. Pain diagnoses and pain medication prescriptions before, during and after major depression diagnosis: a nationwide cohort of 521,244 people with new onset depression. J Affect Disord. 2026;397:120854. 10.1016/j.jad.2025.120854. [DOI] [PubMed] [Google Scholar]
- 37.Schmidt PJ, Luff JA, Haq NA, Vanderhoof VH, Koziol DE, Calis KA, et al. Depression in women with spontaneous 46, XX primary ovarian insufficiency. J Clin Endocrinol Metab. 2011;96:E278–87. 10.1210/jc.2010-0613. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Hou L, Chen L, Zhang W. The longitudinal predictive effect of self-reported frequency of premenstrual syndrome on depression: findings from the Australian Longitudinal Study on Women’s Health. Front Public Health. 2023;11:1126190. 10.3389/fpubh.2023.1126190. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Strauss JR. The reciprocal relationship between menopausal symptoms and depressive symptoms: a 9-year longitudinal study of American women in midlife. Maturitas. 2011;70:302–6. 10.1016/j.maturitas.2011.08.002. [DOI] [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
Data Availability StatementThe data supporting the findings of this study are available on the CHARLS website (https://charls.pku.edu.cn/). This study was conducted in accordance with the Declaration of Helsinki. CHARLS was approved by the Institutional Review Board of Peking University (approval number: IRB00001052-11015 for the household survey and IRB00001052-11014 for blood samples).