The China Health and Retirement Longitudinal Study (CHARLS) is administered by the National Institute of Development and Research in China and was initiated in 2011. Data collection occurs biennially, with a total of five waves completed by 2020. The inaugural baseline survey was conducted from June 2011 to March 2012, encompassing 17,705 participants. Utilizing a Probability Proportional to Size (PPS) sampling method, participants were randomly selected from 28 provinces, 150 counties/districts, and 450 villages/residential committees. Data collection was performed through Computer-Assisted Personal Interviewing (CAPI) conducted in participants’ homes. The study received ethical approval from the Biomedical Ethics Review Committee of Peking University (IRB00001052–11015), and all participants provided informed consent.

CHARLS employs a longitudinal design with stratified sampling and PPS methods to ensure a nationally representative sample of Chinese families. It targets residents aged 45 and above, thereby offering comprehensive insights into the demographic and health status of China’s urban and rural elderly populations. To investigate the association between perceived caregiver stress and anxiety and the incidence of depression in offspring, the sample was restricted to individuals with available family baseline information in 2014, matched with corresponding data from 2015 based on participant IDs. Participants with missing family information in 2014, absent depression data in 2015, or incomplete covariate information in 2015 were excluded. Ultimately, 6,450 participants were included in the subsequent analyses. The specific criteria for inclusion and exclusion are detailed in Figure 1 .

Depressive symptoms were evaluated using the Center for Epidemiologic Studies Depression Scale (CES-D-10) within the CHARLS database. The CES-D-10 is a validated instrument for measuring depressive symptoms among elderly populations in China. It comprises 10 items, each with four response options: “Rarely or none of the time” (<1 day), “Some or a little of the time” (1-2 days), “Occasionally or a moderate amount of time” (3-4 days),” Most or all of the time” (5-7 days). Each response is scored from 0 to 3, resulting in a total score range of 0 to 30, with higher scores indicating greater severity of depressive symptoms. A cutoff score of ≥10 is employed to identify individuals exhibiting significant depressive symptoms.

In 2014, CHARLS incorporated a life course questionnaire that included a section on caregiver mental health. Two specific items were selected to assess caregiver anxiety and stress:A. “When you were a child, did your female caregiver often feel tense or anxious?”A little of the time (0); Some of the time (1); Good part of the time (2); Most of the time (3). B. “When you were a child, did your male caregiver often feel tense or anxious?” A little of the time (0); Some of the time (1); Good part of the time (2); Most of the time (3). Responses of 2, 3, or 4 were indicative of the individual having experienced caregiver stress and anxiety during childhood. These responses were scored accordingly, with 0 points assigned to responses indicating minimal or no stress/anxiety, and higher scores reflecting greater levels of caregiver stress and anxiety. The total scores for male and female caregivers were calculated separately and subsequently combined to form an overall caregiver anxiety score. Additionally, analyses were conducted to separately evaluate the impact of female and male caregiver anxiety and stress on the prevalence of depression in offspring during adulthood.

This study primarily investigates the relationship between perceived caregiver anxiety/stress and offspring depression, recognizing that depression is influenced by a multitude of factors. To account for potential confounders, the following covariates were included in the analysis:Demographic Variables: Gender, age, residence (urban/rural), education level, marital status, and race. Health-Related Factors: Body Mass Index (BMI), obesity, hypertension, diabetes.Lifestyle Factors: Smoking and alcohol consumption.Socioeconomic Indicators: Family economic status during childhood.These covariates were selected based on their established associations with the incidence of depression, ensuring a comprehensive adjustment for potential confounding variables in the analysis of the primary relationship under investigation (14, 15).

Continuous variables are presented as means ± standard deviations (SD) or medians with interquartile ranges (IQR), while categorical variables are expressed as percentages. To rigorously examine the association between individual depression and perceived caregiver stress and anxiety, we employed multivariable logistic regression analyses. Three distinct logistic models were utilized to calculate the odds ratios (OR) and 95% confidence intervals (CI) for the relationship between caregiver stress and anxiety and depression in offspring: Model 1: Crude (unadjusted) association. Model 2: Adjusted for age, sex, education, location, marital status, BMI, obesity, and nationality.Model 3: Further adjusted for smoking, alcohol consumption, hypertension, diabetes mellitus (DM), and childhood financial status, in addition to the covariates included in Model 2.Subgroup analyses were conducted to explore the relationship between caregiver stress and anxiety and offspring depression across various demographic and health-related strata. All statistical analyses were performed using R version 4.3.2.

Table 1 delineates the baseline characteristics of participants who met the inclusion criteria during the 2015 follow-up survey. The non-depressed group had an average age of 60.87 ± 9.19 years, whereas the depressed group averaged 61.22 ± 8.95 years (p = 0.13), indicating no significant difference in age distribution between the two cohorts. In terms of Body Mass Index (BMI), the depressed group exhibited a slightly lower mean BMI of 23.98 ± 4.13 compared to the non-depressed group’s mean BMI of 24.24 ± 3.70, a difference that reached statistical significance (p = 0.01). Caregiver anxiety scores were substantially higher in the depressed group (1.29 ± 1.84) than in the non-depressed group (0.87 ± 1.56), with the difference being highly statistically significant (p < 0.0001). Gender distribution revealed that females constituted a significantly larger proportion of the depressed group (73.91%) compared to the non-depressed group (61.99%, p < 0.0001). Similarly, the depressed group had a higher percentage of unmarried individuals (16.71%) compared to the non-depressed group (11.18%, p < 0.0001). Educational attainment analysis showed that a greater proportion of the depressed group had primary or lower education levels (78.18%) compared to the non-depressed group (64.03%, p < 0.0001). Rural residency was more prevalent in the depressed group (70.06%) than in the non-depressed group (59.35%, p < 0.0001). Regarding smoking status, the depressed group had a lower proportion of current smokers (17.80%) compared to the non-depressed group (20.85%, p < 0.0001). Additionally, underweight status was more common in the depressed group (6.74%) compared to the non-depressed group (4.48%, p < 0.001).Caregiver characteristics indicated significant differences between the depressed and non-depressed groups for both male and female caregivers. Specifically, higher caregiver anxiety and stress scores were associated with the depressed group. Furthermore, childhood financial status analysis revealed that a higher percentage of the depressed group reported poorer economic conditions, with 28.14% considering themselves “far worse off than others” compared to 19.18% in the non-depressed group (p < 0.0001). No significant differences were observed between the two groups concerning ethnicity, alcohol consumption, hypertension, or diabetes mellitus. Detailed data are presented in Table 1 .

We found a significant association between perceived caregiver stress and anxiety and late-life depression. Although the association decreased as covariates were added, it remained statistically significant. In the crude risk model, perceived caregiver stress and anxiety scores were significantly positively correlated with child depression, with a 95% CI of 1.61 (1.45, 1.79), p < 0.0001. In Model 1, the 95% CI was 1.59 (1.43, 1.78), p < 0.0001. In Model 2, the 95% CI was 1.49 (1.33, 1.66), p < 0.0001 (detailed data can be found in Table 2 ).

We further performed regression analysis on perceived caregiver stress and anxiety and child depression, separating male and female caregivers. Among male caregivers, the association with child depression was significant in the crude and partially adjusted models. However, after full adjustment for covariates, the association was no longer significant. In contrast, for female caregivers, the association remained significant across all three models.

For male caregivers, we used “A little of the time” as the reference group. In the crude model, the 95% CI for “Some of the time” was 1.4 (1.20, 1.62), p < 0.0001, for “Good part of the time” was 1.71 (1.43, 2.05), p < 0.0001, and for “Most of the time” was 1.9 (1.54, 2.34), p < 0.0001, with p for trend < 0.0001. In Model 1, the 95% CI for “Some of the time” was 1.4 (1.20, 1.62), p < 0.0001, for “Good part of the time” was 1.61 (1.33, 1.94), p < 0.0001, and for “Most of the time” was 1.82 (1.47, 2.26), p < 0.0001, with p for trend < 0.0001. In Model 2, the 95% CI for “Some of the time” was 1.01 (0.83, 1.23), p = 0.9, for “Good part of the time” was 1.14 (0.89, 1.47), p = 0.31, and for “Most of the time” was 1.22 (0.90, 1.67), p = 0.2, with p for trend = 0.16 (more information can be found in Table 3 ).

For female caregivers, the risk of child depression showed a significant positive correlation with perceived caregiver stress and anxiety, particularly in groups more frequently monitored by female caregivers, where depression risk was higher. Even after controlling for various confounding factors, the association remained significant, although with a relatively lower risk ratio. This suggests that female caregiving may play an important role in depression risk. Similarly to male caregivers, we used “A little of the time” as the reference group. In the crude model, the 95% CI for “Some of the time” was 1.51 (1.31, 1.73), p < 0.0001, for “Good part of the time” was 1.75 (1.47, 2.08), p < 0.0001, and for “Most of the time” was 1.92 (1.59, 2.33), p < 0.0001, with p for trend < 0.0001. In Model 1, the 95% CI for “Some of the time” was 1.56 (1.35, 1.80), p < 0.0001, for “Good part of the time” was 1.68 (1.41, 2.01), p < 0.0001, and for “Most of the time” was 1.86 (1.53, 2.26), p < 0.0001, with p for trend < 0.0001. In Model 2, the 95% CI for “Some of the time” was 1.49 (1.24, 1.79), p = 0.9, for “Good part of the time” was 1.39 (1.09, 1.77), p = 0.01, and for “Most of the time” was 1.47 (1.11, 1.96), p = 0.01, with p for trend < 0.0001 (more information can be found in Table 4 ).

We conducted a subgroup analysis on the association between perceived caregiver anxiety and child depression to explore potential interactions based on different demographic and health characteristics. The analysis results are as follows: in the gender subgroup, the p for interaction = 0.3, with the hazard ratio (HR) being 1.580 (95% CI: 1.392, 1.793), P < 0.0001 in females, and HR = 1.782 (95% CI: 1.474, 2.155), P < 0.0001 in males. For marital status, p for interaction = 0.205, with HR = 1.899 (95% CI: 1.438, 2.515), P < 0.0001 in unmarried individuals and HR = 1.563 (95% CI: 1.396, 1.751), P < 0.0001 in married individuals. Regarding education level, p for interaction = 0.924, with HR = 1.593 (95% CI: 1.291, 1.964), P < 0.0001 in individuals with high school education, HR = 1.234 (95% CI: 0.301, 4.483), P = 0.754 in those with college education or higher, and HR = 1.607 (95% CI: 1.421, 1.818), P < 0.0001 in those with elementary education or less. For geographical location, p for interaction = 0.083, with HR = 1.384 (95% CI: 1.151, 1.664), P < 0.001 in urban areas and HR = 1.688 (95% CI: 1.485, 1.920), P < 0.0001 in rural areas. Regarding smoking status, p for interaction = 0.191, with HR = 1.604 (95% CI: 1.416, 1.817), P < 0.0001 in never-smokers, HR = 1.491 (95% CI: 1.175, 1.891), P < 0.001 in current smokers, and HR = 2.156 (95% CI: 1.545, 3.014), P < 0.0001 in former smokers. For alcohol consumption, p for interaction = 0.444, with HR = 1.653 (95% CI: 1.466, 1.864), P < 0.0001 in non-drinkers and HR = 1.503 (95% CI: 1.216, 1.858), P < 0.001 in drinkers. In terms of ethnicity, p for interaction = 0.025, with HR = 1.670 (95% CI: 1.498, 1.862), P < 0.0001 in non-minorities and HR = 1.062 (95% CI: 0.723, 1.554), P = 0.758 in minorities. Regarding hypertension, p for interaction = 0.949, with HR = 1.612 (95% CI: 1.426, 1.822), P < 0.0001 in those without hypertension and HR = 1.624 (95% CI: 1.328, 1.986), P < 0.0001 in those with hypertension. For diabetes status, p for interaction = 0.554, with HR = 1.638 (95% CI: 1.462, 1.836), P < 0.0001 in those without diabetes and HR = 1.503 (95% CI: 1.157, 1.952), P = 0.002 in those with diabetes. For childhood economic status, p for interaction = 0.252, with HR = 1.591 (95% CI: 0.422, 5.573), P = 0.473 in those who considered their financial status “much better than others,” HR = 1.021 (95% CI: 0.690, 1.501), P = 0.917 in those who considered themselves “slightly better than others,” HR = 1.639 (95% CI: 1.404, 1.913), P < 0.0001 in those who considered themselves “the same as others,” HR = 1.406 (95% CI: 1.093, 1.810), P = 0.008 in those who considered themselves “slightly worse than others,” and HR = 1.481 (95% CI: 1.204, 1.823), P < 0.001 in those who considered themselves “much worse than others.” Detailed data are shown in Figure 2 and Supplementary Table 1 .

We analyzed the association between perceived caregiver anxiety and depression risk in children for different subgroups, focusing on male and female caregivers, and stratified by gender, marital status, education level, geographical location, smoking, alcohol consumption, ethnicity, hypertension, diabetes, and financial status. Overall, the association between perceived caregiver stress and late-life depression risk was significant in most subgroups, particularly regarding gender, marital status, education level, geographical location, smoking, and alcohol consumption. While not all interactions were statistically significant, the general trend indicated that the longer the father’s anxiety persisted, the higher the risk of depression in the child. The subgroup analysis for male caregivers showed the following results: regarding gender, p for interaction = 0.67. For females, as perceived male caregiver stress and anxiety increased, the risk of depression in children also increased, with HRs of 1.663 (95% CI: 1.333, 2.075), P < 0.0001 for the “Good part of the time” group and 2.035 (95% CI: 1.566, 2.650), P < 0.0001 for the “Most of the time” group. For males, the HRs were 1.879 (95% CI: 1.352, 2.595), P < 0.001 and 1.799 (95% CI: 1.240, 2.589), P = 0.002 for these groups, respectively. For marital status, p for interaction = 0.596, with HR = 2.690 (95% CI: 1.534, 4.868), P < 0.001 in the “Most of the time” group for unmarried individuals and HR = 1.777 (95% CI: 1.413, 2.234), P < 0.0001 for married individuals. Regarding education level, p for interaction = 0.324, with HR = 1.701 (95% CI: 1.069, 2.663), P = 0.022 for high school graduates and HR = 1.839 (95% CI: 1.443, 2.348), P < 0.0001 for those with elementary education or less. For geographical location, p for interaction = 0.538, with HR = 1.610 (95% CI: 1.086, 2.365), P = 0.016 in urban residents and HR = 1.958 (95% CI: 1.520, 2.527), P < 0.0001 in rural residents. For smoking status, p for interaction = 0.654, with HR = 2.079 (95% CI: 1.608, 2.691), P < 0.0001 in never-smokers and HR = 2.025 (95% CI: 0.991, 4.057), P = 0.048 in former smokers. For alcohol consumption, p for interaction = 0.993, with HR = 1.871 (95% CI: 1.476, 2.373), P < 0.0001 in non-drinkers and HR = 1.994 (95% CI: 1.262, 3.152), P = 0.003 in drinkers. Regarding ethnicity, p for interaction = 0.324, with HR = 1.938 (95% CI: 1.560, 2.408), P < 0.0001 in non-minorities and HR = 1.357 (95% CI: 0.540, 3.299), P = 0.502 in minorities. For hypertension, p for interaction = 0.922, with HR = 1.931 (95% CI: 1.515, 2.460), P < 0.0001 in those without hypertension and HR = 1.818 (95% CI: 1.185, 2.786), P = 0.006 in those with hypertension. For diabetes, p for interaction = 0.228, with HR = 2.027 (95% CI: 1.612, 2.548), P < 0.0001 in those without diabetes. Regarding childhood financial status, p for interaction = 0.466, with HR = 1.966 (95% CI: 1.378, 2.797), P < 0.001 in those who considered their financial status the same as others and HR = 1.329 (95% CI: 0.948, 1.864), P = 0.099 in those who considered themselves much worse than others. More data can be found in Figure 3 and Supplementary Table 2 .

The subgroup analysis of perceived female caregiver stress and anxiety and the association with late-life depression showed similar results to the analysis for male caregivers. Regarding gender, p for interaction = 0.609, with the depression risk in late-life increasing significantly as female caregiver stress and anxiety increased. In the “Good part of the time” and “Most of the time” groups, HRs were 1.767 (95% CI: 1.431, 2.181), P < 0.0001 and 1.828 (95% CI: 1.447, 2.310), P < 0.0001 for females. For males, the highest depression risk was observed in the “Most of the time” group, with HR = 2.225 (95% CI: 1.581, 3.116), P < 0.0001. For marital status, p for interaction = 0.91, with HR = 2.175 (95% CI: 1.312, 3.663), P = 0.003 in the “Most of the time” group for unmarried individuals and HR = 1.874 (95% CI: 1.522, 2.305), P < 0.0001 for married individuals. For education level, p for interaction = 0.393, with HR = 1.371 (95% CI: 0.884, 2.083), P = 0.148 for high school graduates and HR = 1.998 (95% CI: 1.602, 2.495), P < 0.0001 for those with elementary education or less. Regarding geographical location, p for interaction = 0.246, with HR = 1.547 (95% CI: 1.093, 2.173), P = 0.013 in urban residents and HR = 2.080 (95% CI: 1.646, 2.632), P < 0.0001 in rural residents. For smoking status, p for interaction = 0.489, with HR = 1.930 (95% CI: 1.530, 2.436), P < 0.0001 in never-smokers and HR = 2.437 (95% CI: 1.321, 4.464), P = 0.004 in former smokers. Regarding alcohol consumption, p for interaction = 0.661, with HR = 1.848 (95% CI: 1.487, 2.295), P < 0.0001 in non-drinkers and HR = 2.205 (95% CI: 1.469, 3.317), P < 0.001 in drinkers. For ethnicity, p for interaction = 0.178, with HR = 1.949 (95% CI: 1.601, 2.373), P < 0.0001 in non-minorities and HR = 1.550 (95% CI: 0.660, 3.578), P = 0.304 in minorities. For hypertension, p for interaction = 0.768, with HR = 2.029 (95% CI: 1.626, 2.532), P < 0.0001 in those without hypertension and HR = 1.645 (95% CI: 1.120, 2.409), P = 0.011 in those with hypertension. For diabetes, p for interaction = 0.244, with HR = 2.067 (95% CI: 1.679, 2.545), P < 0.0001 in those without diabetes. Regarding childhood financial status, p for interaction = 0.639, with HR = 2.040 (95% CI: 1.485, 2.797), P < 0.0001 in those who considered their financial status the same as others and HR = 1.467 (95% CI: 1.073, 2.008), P = 0.016 in those who considered themselves much worse than others. More data can be found in Figure 4 and Supplementary Table 3 .