The mediating effect of depressive symptoms on the association between childhood friendship and physical function in middle-aged and older adults: Evidence from the China Health and Retirement Longitudinal Study (CHARLS)

Purpose This study examines the extent to which depressive symptoms mediate the link between childhood friendship (CF) and physical function among middle-aged and older adults in China. Methods The data originated from the China Health and Retirement Longitudinal Study. The Sobel test and multivariable logistic regression were performed to examine the mediating role of depressive symptoms (measured by the 10-item Center for Epidemiologic Studies Depression Scale) in the association between CF (measured by a standardized retrospective questionnaire) and physical function, which was measured by basic activities of daily living (BADL) disability, instrumental activities of daily living (IADL) disability, and grip strength. Results A total of 12,170 participants aged 45 or older were included in this cross-sectional study. After controlling for covariates, worse CF was associated with an increased prevalence of BADL disability (OR = 1.18; 95% CI = 1.05–1.32), IADL disability (OR = 1.25; 95% CI = 1.12–1.40), and low grip strength (OR = 1.21; 95% CI = 1.09–1.34). The proportion of the mediating effect of depressive symptoms was 49% in CF and BADL, 41% in CF and IADL, and 12% in CF and grip strength. Depressive symptoms and worse CF have a joint effect on BADL disability (OR = 3.30; 95% CI = 2.82–3.85), IADL disability (OR = 3.52; 95% CI = 3.03–4.09), and low grip strength (OR = 1.65; 95% CI = 1.43–1.92).


Introduction
Physical function is the ability to perform both basic and instrumental activities of daily living (ADL), and the ability of older adults to reside in the community depends to a large extent on their level of physical function [1].Physical function in older adults could be re ected by the basic activities of daily living (BADL), instrumental activities of daily living (IADL), and grip strength [2].The BADL scale represents the most basic activities involved in daily independent functioning, such as bathing, dressing, and eating [3].IADL described activities necessary for adaptation, including community activities such as shopping, cooking, transportation, and housekeeping [3].Additionally, grip strength is an objective measure of upper body strength and has been demonstrated to decline with age [4].Impaired physical function can lead to increased mortality and hospitalization rates and is signi cantly associated with falls, dependence, and cognitive decline [4][5][6].In China, more than 40 million older adults are currently suffering from impaired physical function, and this number is expected to reach 65 million by 2030 [7].Therefore, by investigating factors associated with the development of long-term physical function among older adults, interventions focusing on the improvement of physical function in adults could be developed.
Childhood experiences have a far-reaching in uence across the entire life span and adverse childhood experiences play an important role in the development of physical function in later life [8].Previous studies have demonstrated that poor economic status, being abused, and being neglected during childhood were associated with decreased physical function later in life [9].However, there is a dearth of empirical studies assessing whether and how childhood friendship (CF) affects impaired physical function in middle-aged and older adults.
Depression is a common psychiatric disorder among middle-aged and older adults, affecting an estimated 322 million people worldwide [10].According to data from baseline data of the China Health and Retirement Longitudinal Study, 30% of men and 45% of women aged 45 and above in China suffered from depression [11].This indicates that depression has become a prominent problem and a serious disease burden for society in China [12].
Meanwhile, accumulating studies have indicated that worse CF contributes to depression, in both young and old adults.
Rebecca A Schwartz-Mette et al., in a study of young adults, showed that friendship problems can elicit depressive symptoms linked to hopelessness and helplessness [13].Another study demonstrated that those who had lower CF scores were signi cantly associated with a higher risk of depressive symptoms in 13,354 Chinese individuals aged 45 + years [14].
Furthermore, depression has been inversely related to the level of physical function in older people [15,16].A 3-year followup study in the United States comprising 3,421 participants found that depression was an independent risk factor for impaired physical function [17].Similar results were reported in a Japanese American population-based survey of older adults [15].However, the extent to which and how depressive symptoms affect the association between CF and physical function in middle aged and older adults is unclear and thus is investigated in this study.Therefore, by using a nationally representative sample that provided both contemporaneous and retrospective data in China, this study was performed with the aim of investigating the associations between CF and physical function among Chinese middle-aged and older adults, and sought to further reveal the mediating role of depressive symptoms in the association between CF and physical function and the joint effect of depressive symptoms and CF in this association.

Participants
Data were derived from the China Health and Retirement Longitudinal Studies (CHARLS), a publicly available dataset and a national longitudinal survey of Chinese community-dwelling adults [18].Using a sample method of multistage strati ed probability-proportionate-to-size, CHARLS offered a wide range of information on adults over age 45 recruited from approximately 10,000 households in 28 provinces, and their household demographics, socioeconomic status, family structure, and health circumstances.The information collection process used face-to-face computer-assisted interviews, with two to three years of follow-up for each respondent.
For this analysis, the life history data in 2014 and the wave of the survey in 2015 were used.The life history data were collected in a special module designed with the aim of studying the life course of the respondents by using the life history survey questionnaire.The demographic, depressive symptom and physical function information of the respondents was derived from the 2015 wave, and information about their childhood was derived from the 2014 life history module.The number of eligible respondents was 12,170 (Fig. 1).

Assessments of childhood friendship
Friendship is a close relationship between two peers that involves mutual liking and affection [19].According to the questionnaire released by CHARLS, three items from the life course questionnaire were used to measure friendships before the age of 17.The three questions were as follows: (i) When you were a child, did you feel lonely because you did not have friends? 1 = Often, sometimes, 0 = rarely or never (ii) When you were a child, did you have a group of good friends to play with?(1 = Often/sometimes, 0 = rarely/never) (iii) When you were a child, did you have a good friend?(0 = yes,1 = no).The total CF score was the sum of the scores for each question, ranging from 0 to 3 [20].We categorized the participants into the following two groups based on their CF scores: good (0 points) and poor (1-3 points).

Assessments Of Physical Function BADL/IADL disability
The CHARLS asked respondents if they required assistance with any of the six BADLs (dressing, bathing, eating, getting into and out of bed, toileting and controlling urination and defecation) or the ve IADLs (preparing a hot meal, shopping for groceries, doing housework, taking medicines and managing money).All respondents were requested to choose from four options: do not have any di culty; have di culty but can still do it; have di culty and need help; and cannot do it.In our analysis, respondents who needed help (last two options) with any item were considered to have a BADL/IADL disability [7].

Grip strength
Grip strength (kilogram) was estimated by using a dynamometer (WCS-100, Nantong, China).Participants squeezed the handles as hard as possible in a standing position with their arms hanging naturally at their sides.Two measurements were taken with each hand.The average value of four measurements was used in the analysis.A grip strength of less than 18 kg for women and less than 26 kg for men was de ned as low grip strength [21].Moreover, a grip strength value of 0 or above 100 kg was considered a missing value and was not included in the analysis [22].

Assessments of depressive symptoms
Depressive symptoms were examined by the 10-item abbreviated version of the Centre for Epidemiological Studies Depression Scale (CES-D), which has been widely used [23].The 10-items refer to the depressive feelings and behaviour of the respondents during the week before the individuals took the depression assessment.Each of the scale items evaluated a depressive symptom with an assigned value ranging from 0 to 3 (0 = less than 1 day; 1 = 1 to 2 days; 2 = 3 to 4 days; 3 = 5 to 7 days).Thus, the total score of the scale ranged from 0 to 30, with a higher score indicating a higher level of depression.
A cut-off score ≥ 10 was used to identify the respondents who had signi cant depressive symptoms [23].

Assessments Of Covariates
The analysis included both childhood and adulthood characteristics, with the childhood characteristic measured by the question "Before you were 17 years old, what was the economic situation of your family in relation to the average family in your community/village at that time?" [9].The demographic variables included age (years), gender (male, female), marital status (married/partnered, single or others), and educational attainment (junior high school and more, junior high school and less).Health related variables included smoking (yes, no), alcohol consumption (nonregular, regular), BMI (normal, nonnormal) [24], depressive symptoms (yes, no), and self-reported health (fair/good/very good, bad/very bad).

Statistical analysis
Frequencies and percentages were calculated for descriptive analysis.The chi-square test was used to describe the differences in the prevalence of declined physical function among different attributes in the whole sample.Logistic regression was used to analyse the relationship between CF and physical function and the joint effect of depressive symptoms and CF on this relationship (Model 1 as an unadjusted model) and an adjusted model was used for covariates (Model 2 as an adjusted model).Then, we repeated the analysis between CF and physical function strati ed by participants' gender.Odds ratios (ORs) and 95% con dence intervals (CIs) were calculated and reported from the logistic models.
In addition, Baron and Kenny′s path analysis [25] and Sobel test [26] were adopted to verify whether there were mediating effects of depressive symptoms between CF and physical function.Baron and Kenny's path analysis used regression to test the following four paths (Appendix Fig. 1): (1) Path c: X (independent variable) on Y (dependent variable); (2) Path a: X on M (mediating variable); and (3) Path b: X and M as predictors of Y; if these three pathways were found to be signi cant, the mediating effect was said to be established.( 4) Path c′: X and M on Y.A signi cant Path c′ indicated partial mediation, whereas nonsigni cant Path c′ indicated full mediation.The signi cance of the mediating effect was veri ed by the Sobel test of mediation effect (Z).All analyses were performed using Stata software version 15.A two-sided P < 0.05 was considered statistically signi cant.

Study sample characteristics
The demographic characteristics of the respondents are described in Table 1 Note: CF = childhood friendship, BADL = basic activities of daily living, IADL = instrumental activities of daily living, BMI = body mass index, SRH = self-report health status, CFS = childhood nancial situation.

Association between CF and physical function
Table 2A reports the relationship between CF and physical function.The results adjusted for covariates showed that CF was recognized as signi cantly associated with physical function in adulthood.Compared with respondents who did not experience poor CF, those who did were more likely to experience BADL disability (OR = 1.18, 95% CI 1.05-1.32),IADL disability (OR = 1.25, 95% CI 1.12-1.40),and low grip strength (OR = 1.21, 95% CI 1.09-1.34).However, there was no signi cant difference in such an association between men and women.

The Mediating Effect Of Depressive Symptoms
Table 3 reports the mediating role of CF on physical function in later life through depressive symptoms.The Sobel test showed that CF had a direct effect on physical function as well as an indirect effect on physical function through depressive symptoms.Figure 2 presents the coe cients of the associations.The direct effects of CF on BADL disability, IADL disability, and low grip strength were 0.027 (P < 0.05), 0.042 (P < 0.05), and − 0.732 (P < 0.05), respectively.Moreover, the indirect effects of CF on BADL disability, IADL disability, and low grip strength were 0.026 (P < 0.05), 0.029 (P < 0.05), and − 0.095 (P < 0.05), respectively.The total effects of CF on BADL disability, IADL disability, and low grip strength were 0.053 (P < 0.05), 0.072 (P < 0.05), and − 0.827 (P < 0.05), respectively.Finally, the mediated effect ratios for depressive symptoms were 49% in CF and BADL, 41% in CF and IADL, and 12% in CF and grip strength.

Discussion
In this national cross-sectional study in China, we found a positive association between CF and physical function in middleaged and older individuals.This nding showed that those who had better CF were more likely to have a lower rate of impairments in physical function.It was particularly interesting that we demonstrated the mediating role of depressive symptoms on the relationship between CF and physical function in elderly individuals.Moreover, the joint effects of CF and depressive symptoms on physical function in later life were observed.
Better CF was associated with better physical function later in life.This nding was consistent with a previous study showing that early life experience can in uence the health of individuals in later life [27].This could be for several reasons.First, repeated stress in early life can cause long-term health problems, including an increasing prevalence of chronic diseases in later life, which may act as stressors of impaired physical function [28].Second, individuals with worse CF have an increased risk of being exposed to a low socioeconomic status in adulthood, which may also be associated with impaired physical function [29].Third, poor CF may increase the likelihood of engaging in risky behaviour in adolescence and adulthood such as smoking; and drinking, which are also possibly associated with impaired physical function [27].
Theoretically, evidence also supports that adverse environments, such as experiencing fewer positive friendship experiences may lead to changes in gene expression and cause functional and structural changes in the brain, autonomy and immune system [30].This nding further indicated that community-based intervention programs promoting children's friendship opportunities (e.g., interaction with friends, teamwork, and club activities) may improve the long-term physical function of individuals.
Our ndings have further con rmed the mediating effect of depressive symptoms on the association between CF and the levels of physical function in middle-aged and older adults.On the one hand, a previous large study demonstrated that middle-aged and older adults in China with adverse CF were at higher risk of depressive symptoms as they aged [14].
Children who grew up in environments lacking friendships tended to be lonely and more vulnerable to psychological and emotional trauma [14].On the other hand, depressive symptoms were demonstrated to be signi cantly associated with a higher prevalence of impaired physical function among middle-aged and older adults in China [31].The reason is that people with depressive symptoms were more likely to close themselves off, which may lead to impaired role performance and interfere with self-care and mobility, as well as loss of opportunities to be socially and physically active leading to decreased physical function [32,33].Therefore, developing appropriate interventions for prevention and controlling the development of depressive symptoms in adults (e.g., through child companionship, social interaction, and cultivation of interests) may help reduce the decline in physical function in later life.
It is important to note that the increased risk of impaired physical function depends on the presence of adverse CF in patients with depressive symptoms, as the impaired physical function risk among positive CF participants with depressive symptoms did not increase in our ndings.The potential mechanism behind the joint effect of adverse CF and depressive symptoms is unknown.This may be explained by several reasons.First, as both adverse CF and depressive symptoms are associated with the volume of the hippocampus, one might speculate that the observed excess impaired physical function risk may be related to reduced hippocampal volume [30,34].Second, adverse CF is associated with less physical activity, which may add to the decline in physical activity levels induced by depressive symptoms, resulting in an excess risk of impaired physical function [35].Third, middle-aged and older adults with depressive symptoms may experience secondary psychological trauma in the process of recalling adverse CF experiences, which may aggravate depressive symptoms and result in an excess risk of impaired physical function [34,36].Therefore, middle-aged and older adults with coexisting adverse CF and depressive symptoms were at higher risk for impaired physical function.
Based on the results of the joint effect of depressive symptoms and CF on physical function, we further found that depressive symptoms were more strongly associated with physical function than CF in middle-aged and older adults.
Compared with CF, both depressive symptoms and physical function were measured in 2015 and were less susceptible to memory bias, so the results obtained are more accurate and not weakened.Moreover, previous studies have reported a variety of mechanisms linking depression and physical function [37,38].Via behavioural mechanisms, depressed individuals are less likely to pay adequate attention to personal health and follow treatment regimens and are more likely to engage in risky behaviours such as smoking, drinking and physical inactivity [39].Other contributing mechanisms could be multiple pathophysiological pathways, such as in ammation and oxidative stress, that have been found to be increased in both persons with depression and in those with poorer physical function [37].
Although our hypothesis of causality between CF and physical function is supported by life-course theory, we cannot rule out the possibility of reverse causality given our cross-sectional study design.If the relationship is indeed bidirectional, such that CF affects depressive symptoms leading to impaired physical function, and conversely, physical function affects depressive symptoms leading to more severe friendship problems, then small effects in the short term may snowball into larger practical effects over time.

Policy Implications
Our results have several health policy implications for China.First, instead of focusing on the health of speci c age groups, health care policies in China should consider their long-term impacts over the life cycle.For example, the fact that childhood health matters to achieve successful ageing, including avoidance of disability, maintenance of high physical function, and sustained engagement in social and productive activities, implies that interventions to improve elderly health should take into consideration the events in the life course, beginning in childhood.The CDC's comprehensive approach to preventing adverse childhood experiences uses multiple strategies from the best available evidence [40].The fth of these strategies is interventions to strengthen primary care through screening, referral and support to identify and address childhood exposure to bad friendships, thereby mitigating immediate and long-term harm and advancing trauma-informed care for children, adolescents, and adults with a history of adverse exposure to CF [41,42].

Limitations And Strengths
Our study has several limitations.First, this study used the Baron and Kenny's classic mediation model as the primary analysis.The well-recognized analytical limitations include that this approach cannot handle exposure-mediator interactions (interaction between CF and depressive symptoms in this study).This is inevitable and may alter our ndings.Therefore, further research is needed.Second, the association between CF and physical function may vary across developmental stages, from early childhood to adolescence.For example, preadolescents and adolescents may experience different friendship de cits, which may have different consequences for physical function in later life.Third, it may happen that participants with depressive symptoms report their childhood experiences differently than those without depressive symptoms.For example, depressed individuals may tend to remember negative events, which could have an impact on our results.Fourth, we were not able to control for people's behavioural problems in childhood, genetic factors or friendship problems in adulthood that might confound the relationship between worse CF and physical function in the regression analysis.Fifth, CF in our study was measured in middle-aged and older adults, and hence was subject to recall bias.
Previous research has reported the reliability of retrospective measures, which cannot be simply replaced by prospective measures [43].Nevertheless, further studies are needed to evaluate whether prospective and retrospective CF have different associations with impaired physical function.Given the lack of longitudinal studies that span from childhood to the elderly in China, our current ndings would be valuable when assessing the in uence of childhood experience on later life.Finally, although we combined three items to measure the status of CF, no other details of childhood friendship questions were measured in CHARLS.We did not consider the number, quality and long-term nature of friends, all of which have been found to be associated with worse friendship outcomes [13].Therefore, it is necessary to further investigate the effects of various CF patterns on physical function later in life.
Despite these limitations, this study has some strengths.For instance, to the best of our knowledge, this study is the rst to analyse the association between CF and physical function in middle-aged and aged adults.We validated depression as a risk factor for physical function in middle-aged and older adults, and further extended the ndings to childhood.Moreover, this study was a large-scale nationwide sample, data were derived from diverse regions of China, and consistent and standardized survey methods were used, so our data are representative of the middle-aged and older population in China.

Conclusions
In this study, adverse CF was demonstrated to be signi cantly associated with impaired physical function among Chinese middle-aged and older adults.Moreover, depressive symptoms mediated this association between CF and physical function in later life.Our ndings have provided evidence for the future development of early screening programs and targeted interventions to alleviate the impact of adverse CF and depressive symptoms on quality of life in later life.

of study participants Figure 2 3 Mediation
Figure 2 . Of the 12,170 respondents, 51.6% (6,279/12,170) were female.A total of 51.8% (6,243/12,170) of the participants were over the age of 60, and only 11.19% (1,252/12,170) of the respondents had an education level above junior high school.Furthermore, 55.92% (6,805/12,170) of the participants reported a poor friendship experience.A total of 18.39% (2,238/12,170) had a BADL disability, 20.35% (2,477/12,170) had an IADL disability and 19.66% (2,393/12,170) had low grip strength.Participants with impaired physical function were more likely to be female, over age 60, less educated, and had poor CF.Signi cant differences were found in impaired physical function in age, sex, education level, marital status, smoking, alcohol consumption, depressive symptoms, self-reported health status, child nancial status, and CF (all P < 0.05).
Note: CF = childhood friendship, BADL = basic activities of daily living, IADL = instrumental activities of daily living, BMI = body mass index, SRH = self-report health status, CFS = childhood nancial situation.

Table 2
The mediating effect of depressive symptoms between CF and BADL, IADL, and grip strength Joint Effect Between Cf And Depressive Symptoms On Physical Function Table2Breports the association of the joint effects of CF and depressive symptoms on physical function.The results of joint effects analysis adjusted for covariates showed that participants with both poor CF and depressive symptoms had a 3.30-fold higher risk of BADL disability (OR = 3.30, 95% CI 2.82-3.85),a 3.52-fold higher risk of IADL disability (OR = 3.52, 95% CI 3.03-4.09),and a 1.65-fold higher risk of low grip strength (OR = 1.65, 95% CI 1.43-1.92)than those with neither condition.
Note: CF = childhood friendship, BADL = basic activities of daily living, IADL = instrumental activities of daily living, ***p < 0.001.Models were adjusted for age, sex, education level, marital status, self-rated health, BMI and children′s nancial status.The Model 2 = adjusted age, sex, education level, marital status, self-rated health, BMI, depressive symptoms, and children′s nancial status.