Developmental trajectories and heterogeneity of social engagement among Chinese older adults: a growth mixture model

Abstract

Background

Social engagement is closely related to well-being among older adults. However, studies on the changing trajectory and influencing factors (especially time-varying factors) of social engagement are limited. This study aimed to examine the social engagement trajectory of older Chinese adults and explore its time-fixed and time-varying factors, thus providing evidence for the development of strategies to promote a rational implementation for healthy aging.

Methods

This study included 2,195 participants from a subset of four surveys from the Chinese Longitudinal Healthy Longevity Survey conducted from 2008 to 2018 (with the latest survey completed in 2018), with follow-ups conducted approximately every three years. Growth mixture modeling was used to explore the social engagement trajectory of older adults and the effects of time-varying variables. In addition, multinomial logistic regression was employed to analyze the association between time-fixed variables and latent classes.

Results

Three distinct trajectories of social engagement among older adults in China were identified: slow declining (n = 204; 9.3%), which meant social engagement score decreased continuously, but social engagement level improved; slow rising (n = 1,039; 47.3%), marked by an increased score of social engagement, but with an depressed engagement level; and middle stabilizing (n = 952; 43.4%), which meant social engagement score and engagement level remained quite stable. A time-fixed analysis indicated that age, marital status, educational level, and annual family income had a significant impact on social engagement (P < 0.05). In contrast, the time-varying analysis showed that a decline in functional ability, insufficient exercise (means no exercise at present), deteriorating self-reported health and quality of life, negative mood, monotonous diet, and reduced community services were closely related to the reduction in social engagement levels (P < 0.05).

Conclusion

Three trends were observed at the social engagement level. Older adults with initially high levels of social engagement exhibited a continuous upward trend, whereas those with initially low levels experienced a decline in their social engagement, and those with initially intermediate levels remained quite stable. Considering the primary heterogeneous factors, it is imperative for governments to enhance basic services and prioritize the well-being of older adults. Additionally, families should diligently monitor the emotional well-being of older adults and make appropriate arrangements for meals.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12877-024-05448-6.

Introduction

Population aging is a global phenomenon driven by decreasing mortality and fertility [1]. Based on statistical data, population aging is increasing. Individuals aged 65 years and above were reported to make up 9.3% of the global population in 2020, and this is predicted to increase to 16.0% (1.5 billion) worldwide by 2050 [2]. However, as a developing country with a large number of older adults, China is facing a more severe crisis. Since its classification as an aging society in 2000 [3], China’s aging rate has accelerated rapidly. The proportion of the population aged 65 years and older is predicted to rise from 13.50% (2020) to 35% (2050) [4, 5], which has aroused great public concern.

Maintaining a regular lifestyle and good health has become an important issue for older adults owing to economic growth and longevity expectations, and both organizations and governments have long been attempting to address this issue. In 2015, the World Health Organization proposed the concept of healthy aging, which refers to the process of developing and maintaining functional ability that enable well-being in older age. The World Report on Aging and Health emphasized that to create age-friendly environments and support healthy aging, governments should take measures such as providing opportunities for social engagement [6]. And there are benefits for those who are also receiving aged care services - e.g. opportunities for civic engagement [7], as well as in aged care [8] where these individuals are even more frail. Since then, social engagement, regarded as a key component of healthy aging, has drawn great attention and has gradually become an important part of policy formulation. In 2021, the State Council in China issued a guideline to promote the development of national undertakings for the aged and improve the elderly care service system during the 14th Five-Year Plan period (2021–2025), and advocated for the positive implementation of social engagement.

Social engagement refers to an individual’s participation in activities that involve interacting with others in the community and shared spaces, and can vary based on available time, resources, societal context, and personal preferences [9]. Based on previous studies, social engagement impacts human health, both physically and psychologically. A previous study has indicated that improved social group participation serves as a protective factor against cognitive decline in older females [10]. Ding et al. [11] found a bidirectional link between social engagement and depressive symptoms among adults aged 50 years and above, which indicates that both factors have a negative impact on each other. A study involving 9,645 middle-aged and older adults in China reported that social engagement had a positive effect on activities of daily living (ADLs) [12]. In addition, a prospective European cohort study of 8,623 participants provided evidence of a potential connection between social engagement and lower all-cause mortality, especially in physically active individuals [13]. Furthermore, Douglas et al. [14] revealed that the mediation between social engagement and health is social support and an individual’s sense of community social cohesion. Consequently, social engagement is closely associated with the health status and living conditions of older adults, attracting ample research interest. This confirms its indispensable role in successful aging and in many conceptual models of human functioning [15].

Some studies that have focused on potential factors related to social engagement, but have not concentrated on its health influence, have identified age, sex, educational level, financial condition, and neighborhood cohesion to be factors related to social engagement [16–18]. Trajectory modeling techniques are employed to examine the changing trajectory of social engagement [19, 20]. This is an innovative method to identify latent subgroups and better describe intra- and inter-individual variability in health outcome patterns over time [21], and has been used in the medical field. Using latent growth mixture modeling and latent class growth analysis, Starzer et al. [22] analyzed positive and negative symptoms trajectories in the following 20 years after the first psychotic episode in patients with schizophrenia spectrum disorder. Huang et al. [23] employed latent mixture modeling to investigate the trajectory of the cardiovascular health score and predict the risk of myocardial infarction in patients with hypertension. Additionally, Chen et al. [24] adopted latent growth mixture modeling to identify the diet quality trajectory and studied its relationship with cognitive performance.

The widespread use of trajectory modeling techniques inspired us. Consequently, we utilized growth mixture modeling (GMM) to explore the heterogeneous trajectory of social engagement among older Chinese adults. Moreover, given that, to our knowledge, current studies have been limited to the factors influencing social engagement, especially time-varying variables, our study employed an impact analysis of time-fixed and time-varying variables to fill the research gap and provide evidence for appropriate healthy aging practices.

Methods

Data and population sample

The Chinese Longitudinal Healthy Longevity Survey (CLHLS) is a national prospective cohort study conducted by Peking University and Duke University covering 23 provinces in China [25]. The baseline survey began in 1998, with follow-ups conducted approximately every three years. By 2018, the study had conducted eight waves, visiting 113,000 households. The study used a standard questionnaire to obtain demographic characteristics and data on lifestyle, medical history, mental health, and physical health of older adults aged 65 years or over [26] to better understand health-related factors related to longevity. Detailed information on the CLHLS has been provided elsewhere [27].

In this study, we adopted data from the 2008, 2011–2012, 2014, and 2018 waves to identify social engagement trajectories. Figure 1 illustrates the selection process. A total of 16,954 participants were enrolled in the 2008 wave. We included 2,195 participants for analysis, after excluding a total of 14,514 individuals who were lost to follow-up in the subsequent waves or passed away before the 2018 wave, 163 survival individuals with missing social engagement data and 82 individuals under 65 years old.

Fig. 1.

Sample selection in this study. A total of 2195 participants were included

Variables

Dependent variable

We assessed the social engagement level as the dependent variable using seven items in the CLHLS questionnaire: doing housework, doing garden work, reading newspapers and books, raising domestic animals or pets, playing cards or mahjong, watching television or listening to the radio, and taking part in organized activities. The answers were given on a five-point scale to classify the frequency (1: “almost every day,” 2: “not every day, but at least once a week,” 3: “not weekly, but at least once a month,” 4: “not monthly, but sometimes,” and 5: “never”). The total score ranged from 7 to 35, with higher scores indicating lower participation.

Time-varying independent variables

Eight time-varying variables were included: physical exercise (1 = yes; 2 = no), self-reported health (1 = very good, 2 = good, 3 = average, 4 = bad, and 5 = very bad), self-reported quality of life (1 = very good, 2 = good, 3 = average, 4 = bad, and 5 = very bad), functional ability, positive emotions, negative emotions, dietary diversity, and services available in the community.

Functional ability was measured by the basic ADLs (BADLs) and the instrumental ADLs (IADLs). There were six aspects to estimate BADLs (bathing, dressing, toileting, continence, indoor transferring, and feeding) and eight items to evaluate IADLs (visiting neighbors, shopping, cooking, washing clothes, walking 1 km, carrying 5 kg of weight, crouching and standing three times, and taking public transportation). The responses were divided into a three-point scale (1: “no assistance,” 2: “need some help,” and 3: “near independence”). Higher scores indicated participants with poor functional abilities.

Based on a previous study, we appraised older adults’ mental health using five questions. Two questions were about positive emotions: (a) “Do you look at the bright side of things?” and (b) “Do you feel as happy as when you were younger?” The other three items referred to negative emotions: (a) “Do you feel anxious or fearful?”, (b) “Do you feel isolated and lonely?”, and (c) “Do you feel useless as you age?” Responses were scored on a five-point scale: (1: “always,” 2: “often,” 3: “sometimes,” 4: “seldom,” and 5: “never”) [5]. We calculated the positive and negative emotions respectively. Higher scores indicated less positive or negative emotions.

Dietary diversity was reflected in the intake frequency of 13 food items: vegetables, fresh fruit, meat, fish, eggs, bean products, salt-preserved vegetables, sugar, tea, garlic, dairy products, nuts, mushrooms (algae). We defined “intake almost every day or weekly” or “quite often” as eating this kind of food, which was scored as one point. Participants with higher scores had greater food diversity [28].

According to the questionnaire, the condition of services available in the community were measured by nine items. These included personal care, home visit, psychological consulting, daily shopping, social and recreation, legal aid, healthcare education, neighborhood-relation, and other social services. The total score ranged from 1 to 9, and the number of available services varied more in the community when the score was higher.

Time-fixed independent variables

Time-fixed independent variables contained nine categorical variables, including age (1 = 65–74 years; 2 = 75 years or above), sex (1 = male; 2 = female), marital status (1 = married and living with a spouse; 2 = other), residence (1 = city or town; 2 = rural), yearly household income (1 = 10,000 yuan or less; 2 = more than 10,000 yuan), education (0 = no schooling; 1 = some schooling), financial support (0 = insufficient; 1 = sufficient), current alcohol consumption (0 = no; 1 = yes), and current smoking (0 = no; 1 = yes).

Statistical analysis

We utilized Mplus 7.4 for all data analyses to identify the latent trajectories and probe the impact of time-varying variables. In addition, SPSS 26 was employed for descriptive analysis and exploration of possible time-fixed variables. Categorical variables were presented as numbers (percentages) for descriptive analysis and continuous variables were described using mean ± standard deviation [29].

To identify the heterogeneous trajectories of social engagement levels among Chinese older adults, we utilized GMM, which is a parametric model suitable for longitudinal data. Unlike classical models, the GMM notices latent classes with similar characteristics in the entire population and permits individual distinctions among the same classes, depicting average growth curves. We used three criteria to evaluate model fit and determine the best classification: (a) the Akaike Information Criteria (AIC), Bayesian Information Criteria (BIC), and sample-size adjusted BIC (aBIC): the lower the value, the more appropriate the model; (b) Lo-Mendell-Rubin Likelihood Ratio Test (LMR) and Bootstrap Likelihood Ratio Test (BLRT): the model with a higher P-value showed greater k classes fit; and (c) entropy: ranged from 0 to 1. When the value was close to 1, the model was sufficiently separated into several classes. Additionally, we ensured that the proportion of each class in the final model was greater than 5% [21, 30–32].

After identifying the social engagement trajectory, we used time-fixed and time-varying variables to analyze the potential influence. For the time-fixed analysis, we applied data from the 2008 wave to test the effect on trajectory classification through multinomial logistic regression. For the time-varying analysis, we inputted variables into the final GMM model separately to explore their impact on the trajectory.

Results

Sample characteristics

A total of 2,195 participants were included in this study, with a mean age of 75.44 ± 8.03 years (range 65–108). The other basic characteristics are shown in Table 1 and Supplementary Table 1. Approximately half of the population was female (53.8%), with some schooling (51.1%), married and living with a spouse (57.5%), and with a yearly household income of no more than 10,000 yuan (58.8%). Most lived in rural areas (66.8%), 77.3% had sufficient financial support, 78.1% did not currently consume alcohol, and 77.6% did not smoke at present. In addition, characteristics of time-varying variables showed that during 2008 and 2018, the scores for functional ability (from 15.37 ± 3.00 to 20.15 ± 7.55) and services available in the community (from 0.70 ± 1.42 to 1.85 ± 2.22) presented a continuous upward trend, while scores for negative emotions (from 11.66 ± 2.30 to 11.40 ± 2.25) and self-reported quality of life (from 2.38 ± 0.80 to 2.09 ± 0.79) underwent a general decrease. There was a decline after the rising trend in dietary diversity (from 5.77 ± 2.59 to 5.74 ± 2.62) and self-reported health (from 2.44 ± 0.91 to 2.60 ± 0.93). On the contrary, an increase followed by a drop was found in the physical exercise (from 1.65 ± 0.48 to 1.68 ± 0.47) and positive emotions scores (from 5.21 ± 1.41 to 4.79 ± 1.63). The level of social engagement scores showed an uptrend as a whole, the average of which was 24.73 ± 4.64 in 2008, 24.68 ± 5.07 in 2011, 25.22 ± 5.17 in 2014, and 27.80 ± 5.26 in 2018.

Table 1.

Basic characteristics of time-fixed and time-varying variables

Variable	N(%)/Mean ± SD
Age, years
65–74	1186(54.0)
≥75	1009(46.0)
Sex, (%)
Male	1014(46.2)
Female	1181(53.8)
Residence, (%)
City/town	728(33.2)
Rural	1467(66.8)
Marital status, (%)
Married and living with a spouse	1262(57.5)
Other marital status	933(42.5)
Education, (%)
No schooling	1074(48.9)
Some schooling	1121(51.1)
Financial support, (%)
Insufficient	498(22.7)
Sufficient	1697(77.3)
Yearly household income, (%)
≤10,000	1290(58.8)
>10,000	905(41.2)
Current alcohol consumption, (%)
No	1715(78.1)
Yes	480(21.9)
Current smoking, (%)
No	1704(77.6)
Yes	491(22.4)
Functional ability
Year of 2008	15.37 ± 3.00
Year of 2011	16.02 ± 3.86
Year of 2014	16.86 ± 4.74
Year of 2018	20.15 ± 7.55
Physical exercise
Year of 2008	1.65 ± 0.48
Year of 2011	1.56 ± 0.50
Year of 2014	1.62 ± 0.49
Year of 2018	1.68 ± 0.47
Self-reported health
Year of 2008	2.44 ± 0.91
Year of 2011	2.55 ± 0.89
Year of 2014	2.61 ± 0.90
Year of 2018	2.60 ± 0.93
Self-reported quality of life
Year of 2008	2.38 ± 0.80
Year of 2011	2.28 ± 0.83
Year of 2014	2.17 ± 0.79
Year of 2018	2.09 ± 0.79
Positive emotions
Year of 2008	5.21 ± 1.41
Year of 2011	4.48 ± 1.65
Year of 2014	4.56 ± 1.61
Year of 2018	4.79 ± 1.63
Negative emotions
Year of 2008	11.66 ± 2.30
Year of 2011	11.56 ± 2.21
Year of 2014	11.47 ± 2.21
Year of 2018	11.40 ± 2.25
Dietary diversity
Year of 2008	5.77 ± 2.59
Year of 2011	5.89 ± 2.64
Year of 2014	5.91 ± 2.52
Year of 2018	5.74 ± 2.62
Services available in the community
Year of 2008	0.70 ± 1.42
Year of 2011	1.28 ± 1.72
Year of 2014	1.65 ± 1.94
Year of 2018	1.85 ± 2.22

Latent trajectory and classification of social engagement

Table 2 shows the results of the linear and quadratic unconditional models for five classes, respectively. Finally, we considered the three-class quadratic model to be the most appropriate because of its high entropy; low AIC, BIC, and aBIC; and statistical significance in the LMR and BLRT. The percentage of each group in the model was > 5%. Supplementary Table 2 provides more details regarding this model.

Table 2.

Fix index of GMM model for social engagement

Model	k	AIC	BIC	aBIC	Entropy	LMR	BLRT	Class probability
Linear
1	9	51903.334	51954.579	51925.985	-	-	-	-
2	12	51832.323	51900.651	51862.525	0.562	0.0000	0.0000	0.718,0.282
3	15	51811.434	51896.843	51849.186	0.713	0.0156	0.0000	0.719,0.275,0.005
4	18	51791.031	51893.522	51836.333	0.697	0.0150	0.0000	0.513,0.402,0.081,0.005
5	21	51778.027	51897.599	51830.879	0.715	0.0475	0.0000	0.382 ,0.364,0.215,0.034,0.004
Quadratic
1	13	51623.607	51697.628	51656.325	-	-	-	-
2	17	51519.619	51616.416	51562.405	0.595	0.0000	0.0000	0.680,0.320
3	21	51448.640	51568.212	51501.492	0.731	0.0009	0.0000	0.473,0.434,0.093
4	25	51417.252	51559.601	51480.172	0.781	0.0000	0.0000	0.469,0.434,0.092,0.005
5	29	51417.044	51582.168	51490.031	0.724	0.2301	0.0632	0.427,0.425,0.106,0.036,0.006

AIC: Akaike Information Criteria, BIC: Bayesian Information Criteria, aBIC: sample-size adjusted BIC, LMR: Lo-Mendell-Rubin Likelihood Ratio Test, BLRT: Bootstrap Likelihood Ratio Test. Best fitting model (by AIC, BIC, aBIC, Entropy, LMR and BLRT) is bold

There were three latent trajectories of social engagement, as shown in Fig. 2: slow declining (Class 1, 9.3%), with a constant decreased social engagement score; slow rising (Class 2, 47.3%), with slowly ascending scores from 2008 to 2018; and middle stabilizing (Class 3, 43.4%), the social engagement score of which remained quite stable, with slight growth in general. A higher score indicated a lower engagement level. As a result, participants in Class 1 experienced an upward trend over ten years, while participants in Class 2 and Class 3 experienced a downward trend.

Fig. 2.

Latent trajectory of social engagement among Chinese older adults

The influence of time-fixed variables

The classifications of social engagement trajectory and slow rising (Class 2) were used as dependent variables and references, respectively, for the multinomial logistic regression analysis. The results are summarized in Table 3. Older adults aged 65–74 years (odds ratio [OR] = 5.251, 95% confidence interval [CI] = 3.587–7.687 and OR = 3.283, 95% CI = 2.689–4.008, respectively) and married and living with a spouse (OR = 1.732, 95% CI = 1.187–2.527 and OR = 1.283, 95% CI = 1.044–1.575, respectively) had a higher risk of having slow declining (Class 1) and middle stabilizing (Class 3) trajectories, with more social engagement. Meanwhile, those without schooling (OR = 0.462, 95% CI = 0.318–0.672 and OR = 0.608, 95% CI = 0.492–0.751, respectively) or with a yearly household income of no more than 10,000 yuan (OR = 0.662, 95% CI = 0.470–0.932 and OR = 0.763, 95% CI = 0.621–0.936, respectively) had lower risk of slow declining (Class 1) and middle stabilizing (Class 3) trajectories, with less engagement in social activities.

Table 3.

Multinomial logistic regression of time-fixed variables related to social engagement

Variables	Slow declining (Class 1)				middle stabilizing (Class 3)
	OR	95% CI	P value		OR	95% CI	P value
Age, years
65–74	5.251	3.587–7.687	<0.01		3.283	2.689–4.008	<0.01
≥75	1.000				1.000
Sex, (%)
Male	1.147	0.772–1.704	0.497		0.926	0.730–1.176	0.530
Female	1.000				1.000
Residence, (%)
City/town	1.674	1.190–2.354	<0.01		1.196	0.970–1.474	0.094
Rural	1.000				1.000
Marital status, (%)
Married and living with a spouse	1.732	1.187–2.527	<0.01		1.283	1.044–1.575	<0.05
Other marital status	1.000				1.000
Education, (%)
No schooling	0.462	0.318–0.672	<0.01		0.608	0.492–0.751	<0.01
Some schooling	1.000				1.000
Financial support, (%)
Insufficient	0.774	0.503–1.191	0.244		1.139	0.907–1.431	0.262
Sufficient	1.000				1.000
Yearly household income, (%)
≤10,000	0.662	0.470–0.932	<0.05		0.763	0.621–0.936	<0.01
>10,000	1.000				1.000
Current alcohol consumption, (%)
No	0.653	0.442–0.963	<0.05		0.862	0.671–1.109	0.248
Yes	1.000				1.000
Current smoking, (%)
No	0.921	0.614–1.382	0.692		0.831	0.642–1.076	0.161
Yes	1.000				1.000

Reference class: slow rising (Class 2); OR: Odds ratio; 95% CI: 95% Confidence Interval

Compared with the slow rising (Class 2) trajectory, participants who lived in city or town (OR = 1.674, 95% CI = 1.190–2.354) were more likely to have a slow declining (Class 1) trajectory, characterized by higher social engagement level; while those who did not consume alcohol currently (OR = 0.653, 95% CI = 0.442–0.963) were less likely to have a slow declining (Class 1) trajectory, characterized by a lower social engagement level. Our study showed no significant association between sex, financial support, current smoking status, and social engagement.

The influence of time-varying variables

Figure 3 presents the results of the time-varying analyses. The ascending score of functional ability (β = 0.340 ~ 0.510), physical exercise (β = 1.344 ~ 2.291), self-reported health (β = 0.476 ~ 0.732), self-reported quality of life (β = 0.403 ~ 0.644), and positive emotions (β = 0.280 ~ 0.381), related to an increased social engagement score but a lower engagement level. In contrast, the declining scores of negative emotions (β=-0.252~-0.179), dietary diversity (β=-0.409~-0.291), and services available in the community (β=-0.294~-0.094), associated with a higher social engagement score but less engagement. In summary, the decrease in social engagement was mainly caused by a decline in functional ability, insufficient physical exercise, deteriorating self-reported health and quality of life, negative mood, a monotonous diet, and reduced community services.

Fig. 3.

The influence of time-varying variables on social engagement trajectory

Black extending lines represent the 95% credible interval of each estimate. (A) The influence of functional ability; (B) The influence of physical exercise; (C) The influence of self-reported health; (D) The influence of self-reported quality of life; (E) The influence of positive emotions; (F) The influence of negative emotions; (G) The influence of dietary diversity; (H) The influence of services available in the community

Discussion

The present study utilized the GMM to identify older adult-related changes in social engagement in China. There were three different trajectories: slow declining (n = 204; 9.3%), slow rising (n = 1,039; 47.3%), and middle stabilizing (n = 952; 43.4%). Generally, the social engagement of older adults in China presented three opposing developmental tendencies. With a higher initial level, the social engagement scores in the slow rising (Class 2) trajectory increased further in the subsequent waves. In contrast, the score for the slow declining (Class 1) trajectory was quite low at first and then dropped further over the next ten years. And score for middle stabilizing (Class 3) trajectory remained quite stable, with only a slight rise. However, higher scores were associated with lower levels of social engagement. As a result, older adults who had a better engagement level subsequently initially tended to engage in social activities more, while those with a lower initial level preferred to be involved less later.

To explore the factors influencing social engagement, we adopted a time-fixed and time-varying analysis. The results of our study indicated that age was an essential variable. Younger adults tended to have a higher level of social engagement than older adults. Possible explanations for why older participants participated less include transport difficulties, the need for companionship, and their withdrawal from social life [33]. In contrast, younger older adults tended to participate more because of their more favorable participation in ADLs and social connections. Meanwhile, retirement provided them with sufficient time to participate in activities they liked.

Marital status was another factor associated with social engagement. Our study suggests that older adults who are married and living with a spouse tend to engage more than those with other marital statuses. Similarly, a Chinese longitudinal study indicated that married older adults tend to be divided into groups with higher social engagement [34]. A cross-sectional study in Brazil regarded living with partners as a favorable factor [35]. Based on increasing pleasure for engaging in social activities with their spouse and more time to participate through partners’ shared daily routines, married older adults are more likely to engage more [36]. However, Utz et al. [37] reported opposite results, showing that in order to relieve negative effects, widowed individuals experience a constant growth in informal activities after spousal loss. This discrepancy was probably caused by the different extents of individuals and activities in various studies. Apart from widowed older adults, individuals with other marital statuses in our study included single older adults and those with separated marital statuses, which might have affected the outcome.

Education significantly affects the lives of older adults. Highly educated individuals are more likely to want to live a healthy lifestyle, to care about their health status, and to participate in sports and other activities; thus, the activity frequency of these individuals is much higher than that in those with lower education [38, 39]. Furthermore, as the education level increases, self-confidence and communication skills improves [40]. This helps older adults become more involved in social activities and communicate positively with others, ultimately improving their level of social engagement. In addition, highly educated older adults have a lower risk of experiencing social network contraction [41]. This probably offers them more opportunities to participate in social activities and gain sufficient support from their families and friends.

We observed a positive correlation between yearly household income and social engagement. Older adults with higher incomes had more opportunity to engage in leisure activities, such as gardening and raising pets. However, due to the “side effects” of poverty, older adults with lower incomes tended to participate less, although some leisure activities cost little and were acceptable to most retired Chinese females, such as public square dancing (guang chang wu) [42, 43]. Kennedy et al. [44] found that income was significantly associated with health, and that the percentage of individuals reporting fair or poor health was much higher in the lowest-income category than in the highest-income category. Lower-income individuals suffer the strongest impact of income inequality. Limited by poor health, these individuals have fewer chances and a lower desire to engage in social activities.

Previous studies have primarily focused on how social engagement influences functional abilities. Gao et al. [45] reported that engaging in organized social activities is beneficial for maintaining physical activity and cognitive ability, thus protecting older adults against functional disabilities. In the Japan Gerontological Evaluation Study of Aging longitudinal study, Saito et al. [46] indicated a possible connection between a low frequency of outings and limited social relationships, which might cause a recession in functional abilities. However, our study found that functional ability was positively related to social engagement. Poor physical function acts as a barrier against active participation, whereas favorable functional ability makes it possible. The results of our study showed that physical exercise was positively related to social engagement. According to Chinese rural research, engagement in sports or social clubs is the highest among all types of engagement. This may be due to accessibility [47]. Meanwhile, an umbrella review suggested that body-mind exercise promotes older adults’ physical function and balance [48], which might also improve their social engagement level.

Our study found that dietary diversity had a positive impact on social engagement. Participants with higher dietary diversity scores have higher intakes of antioxidants, proteins, and various micronutrients, thus improving their memory status and overall health [49]. Various foods reduce inflammation and oxidative stress and promote healthier aging [50]. Aihemaitijiang et al. [28] indicated that maintaining high dietary diversity in the long term is instrumental for physical function. Duan et al. [51] reported that improved food diversity is conducive to a lower incidence of frailty. As a result, older adults with higher dietary diversity are likely to be healthy and have more energy to participate in activities. An international study also found a link between food consumption and social engagement. It revealed that in China, Russia, and India, lower engagement in social activities was related to limited fruit and vegetable consumption [52]. However, the specific underlying mechanism remains unclear and requires further investigation.

In this study, self-reported health status was found to affect social engagement. However, a cross-sectional survey conducted in South Korea indicated that self-reported health was influenced by social engagement, and this impact was more notable in adults aged 65 years and above [53]. Therefore, a bidirectional connection may exist between self-reported health and social engagement. Our study also indicated that self-reported quality of life is associated with social engagement. A similar outcome was shown among Chinese older adults in the greater Chicago area [54]. One explanation that a study on the relationship between multimorbidity and social participation found was that, through increased resilience or multimorbidity coping, better life satisfaction facilitated constant social engagement [55]. Moreover, emotions are closely related to social engagement. Based on the findings of our study, positive emotion improved older adults’ social engagement level, while in another study, “high activity” older adults had less negative emotions and more positive emotions compared to “low activity” older adults; however, there was no significant distinction between “low activity” and “moderate activity” [34]. It seemed that for some individuals, these two factors influenced and promoted each other.

Our findings confirmed that the variety of services available in a community was closely related to social engagement. A systematic review showed that proximity to resources, recreational facilities, social support, and public transportation can positively affect social engagement [56]. Therefore, to promote engagement, a community environment is crucial. Services in our study also included psychological consultation and legal aid, which are essential for older adults to receive timely therapy, thus gaining enough energy to participate in social activities. Furthermore, an empirical analysis showed that community home older-adult care services changed the social participation arrangement of chronically ill older adults, with a reduction in economic participation and an increase in voluntary and political participation [57]. The results showed that community services are an important factor affecting older adults’ social engagement patterns and frequencies. And it is important to provide aged care service provision, for it can help contribute to better well-being [58].

Based on the results of our study, the social engagement level of older adults in China was unsatisfactory because of a decline in this in some of the participants. In such cases, governments should pay more attention to older adults with advanced age, poor marital status, low educational levels, and insufficient financial status. Several subsidy policies should be formulated to ensure the basic lives of low-income individuals and reduce their financial stress. Additionally, governments must strengthen the construction of universities for older adults. With both educational and social functions, a university for older adults is likely to promote their social support and satisfy their need for spiritual support [59]. Basic facilities and services are vital as they provide older adults with more opportunities to participate in social activities and maintain a regular lifestyle. Therefore, it is important to promote infrastructure construction and service diversity. Families should closely monitor the emotions of older adults, make arrangements for their diets, and lead them to improved social engagement, thereby improving their physical and mental health later in life.

However, this study had some limitations. First, we excluded older adults who were dead or lost to follow-up in the four waves, which might have caused selection bias because these individuals were more likely to have poor health conditions and unsatisfactory social engagement levels. Second, we only found the relationship between social engagement and influencing factors, the exact causative relation needs further exploration. Third, our study mainly focused on the influential variables associated with social engagement, without exploring how social engagement affects others. As a result, we could not ascertain the mechanism among them clearly and entirely, but only generated potential mechanisms based on current studies. Finally, mean age of participants chose in this study was 75.44 years, suggesting that the older adults were relatively young at beginning, which might cause limitations on our findings.

Conclusion

Our study employed GMM analysis to identify three distinct patterns of social engagement: slow declining, slow rising, and middle stabilizing. Additionally, the classification was closely associated with four time-fixed variables (age, marital status, education, and yearly household income) and eight time-varying variables (physical exercise, self-reported health, self-reported quality of life, functional ability, positive emotions, negative emotions, dietary diversity, and services available to the community). Furthermore, to enhance social involvement, governments should enact subsidy policies and enforce infrastructure construction, whereas families should prioritize the dietary requirements of older adults. Overall, our study highlights the importance of promoting social engagement and healthy aging.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Abbreviations

ADLs

Activities of daily living

BADLs

Basic activities of daily living

IADLs

Instrumental activities of daily living

GMM

Growth mixture modeling

AIC

Akaike Information Criteria

BIC

Bayesian Information Criteria

aBIC

sample-size adjusted BIC

LMR

Lo-Mendell-Rubin Likelihood Ratio Test

BLRT

Bootstrap Likelihood Ratio Test

CLHLS

Chinese Longitudinal Healthy Longevity Survey

OR

Odds ratio

CI

Confidence interval

Author contributions

L.W.: Conceptualization, Funding acquisition, Methodology, Project administration, Writing – review & editing. C.Y.: Conceptualization, Methodology, Writing – review & editing. H.Z.: study design, data analysis, results interpretation, manuscript drafting. C.Z.: study design, supervision of data analysis, results interpretation, manuscript revision. N.W.: study design, data collection. All authors read and approved the final manuscript.

Funding

This research was supported by the National Key R&D Program of China (2020YFC2002901), National Natural Science Foundation of China (81960611, 81560550, 81960620, 82460670) and Jiangxi Provincial Natural Science Foundation (20202ACBL206016, 20192BAA208005, 20232ACB216006).

Data availability

The datasets analyzed during the current study are available online from Peking University Open Research Data, https://opendata.pku.edu.cn/dataverse/CHADS.

Supplementary Information

Ethics approval and consent to participate

The study used data from Chinese Longitudinal Healthy Longevity Survey (CLHLS), which got the approval of Biomedical Ethics Committee of Peking University (IRB00001052-13074). And the informed consent forms were provided to all the participants and their legal representatives.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.