Association between social capital and trajectories of cognitive function among Chinese older adults

Abstract

Background

The association between social capital and trajectories of cognitive function (CF) is still unclear among older adults in mainland China. The present study aims to examine the association using a longitudinal cohort from the Ningxia Healthy Aging Cohort.

Methods

Four waves of longitudinal data were abstracted to identify the CF trajectories using the conditional latent class growth model (LCGM). CF was measured using the Chinese version of the Mini-Mental State Examination (MMSE) from 2012 to 2023. Social capital, including social cohesion, social interaction, civic participation, and spiritual capital, was assessed using a series of standard instruments at baseline. Multinomial logistic regression was used to test the association between social capital and each trajectory class.

Results

Three distinct CF trajectories were identified: Stable (72.5%), slow decline (25.8%), and deterioration (1.7%). High spiritual capital was associated with a decreased risk of developing a deterioration trajectory (OR = 0.36, 95% CI = 0.14–0.93), while high civic participation (OR = 0.61, 95% CI = 0.46–0.81) was associated with a decreased risk of developing a slow decline trajectory. Low CF at baseline was the first important risk factor, followed by age, civic participation, sex, monthly income, spiritual capital, and obesity.

Conclusions

The CF trajectories are heterogeneous among older Chinese adults. Civic participation was associated with a lower risk of a slow decline trajectory, and spiritual capital was slightly associated with a lower risk of a deterioration trajectory. These findings support the development of social-integrated interventions for vulnerable groups.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12889-025-21570-5.

Introduction

Age-related health problems have challenged the health system under the rapid acceleration of global population aging [1]. Cognitive impairment has become a common neurological problem among older adults. The overall prevalence of mild cognitive impairment (MCI) ranges from 13.24 to 18.03% globally [2], and it is similar to that reported in China (from 13.5 to 17.4%) [3]. Over 60% of Alzheimer’s disease (AD) patients live in underdeveloped countries, and the estimated dementia prevalence for adults aged 60 and above in India is 7.4% [4], whereas this prevalence is 6.04% in mainland China [5]. To date, there is no curative treatment available for advanced cognitive impairment syndrome such as AD, so early and accurate identification of high-risk individuals is crucial for timely intervention.

The cognitive function (CF) in older adults changes dynamically across evaluation moments and the situation is becoming increasingly complex over time [6]. Longitudinal repeated cognitive assessments are more sensitive and accurate than cross-sectional measures to capture the natural process of temporal cognitive change. Previous studies have identified the heterogeneity in CF trajectories [7]. Studies from various countries, including the USA [8], Spain [9], and China [10] have suggested that there are three to five CF trajectories among older adults. Individuals with adverse CF trajectories (such as moderate or rapid decline) are associated with an increased risk of several health outcomes, including chronic conditions, disability in activities and instrumental of daily living, and motor cognitive risk syndrome [11, 12]. When compared with older adults with a stable trajectory of CF, those with a rapidly decreased trajectory of CF have a significantly higher risk of all-cause dementia, AD, and all-cause mortality [13].

Various factors, including education, sex [14], income [15], physical health [16], and genetic variants [17], are associated with CF trajectories. Individuals with a higher baseline MMSE score and a lower prevalence of the ApoE ε4 allele genotype are related to developing a stable CF trajectory [18]. Social capital (SC) was defined as a collection of various social resources of the public, as it “represents the characteristics of social organization, such as social networks, reciprocity, norms, and trust, that can improve the efficiency of society” [19]. According to the original social capital theory, three dimensions of SC were identified, including social cohesion, civic participation, and social interaction [20, 21]. Social cohesion refers to the level of solidarity and connection that exists among neighbors [22]. Civic participation indicates the extent to which citizens participate in social organizations and activities [23]. Social interaction is a social activity in which individuals interact with each other in their daily lives [24]. With the development of the SC theory, spiritual capital was conceptualized as a unique dimension of SC in recent years [25], it refers to social resources available to individuals and groups through their social connections with a spiritual community [26]. Previous studies have reported that the correlation between SC and cognition may vary with differences in the dimensions of SC [27]. Chinese older adults with high social cohesion were likely to tend to a high-stable trajectory and with a lower risk of persistently low CF trajectory, and those with low social cohesion were likely to have a low-decline trajectory [28]. Recently, several studies have highlighted the possible protective effect of spiritual capital on CF among older adults [29, 30]. One study revealed that spiritual practice was significantly associated with slower trajectories of cognitive decline over time among those aged 60 and above [31]. However, spiritual capital (measured by religious attendance and importance) has been found to have mixed effects on women’s cognition trajectories [32].

Despite the increasing research on the relationship between SC and CF trajectories among older adults, most previous studies have been conducted in Western countries [33, 34]. One study from mainland China revealed that spiritual capital might stimulate and activate the CF among the oldest-old people by enhancing their optimism, happiness, and engagement [35]. However, traditional approaches typically classify CF into cognitive decline, MCI, and AD based on the cumulative changes in disease, neglecting the dynamic changes in CF and the impact of education level on CF assessment. The above topic is still limited, especially in the special context of Chinese culture.

The relationship between SC (especially the spiritual capital) and CF trajectories among older adults in mainland China is still unclear. To address this gap, the current study aims to examine the associations between SC and CF trajectories among community-dwelling older adults using a prospective cohort from the Ningxia Hui Autonomous Region, where over one-third of the population claims to be Muslim and is believed to engage in relatively frequent spiritual practices [36]. Based on the existing evidence, we hypothesize that the CF trajectories would be heterogeneous among older adults, and SC would be associated with a lower risk of developing unfavorable CF trajectories.

Measures

Participants and process

Data for the present study were abstracted from the Ningxia Healthy Aging Cohort (NHAC) established between 2012 and 2016. Participants enrolled using a judgment sampling protocol. First, five districts or counties with a high percentage of minorities were selected. Second, ten communities with a high percentage of older adults in the total population were selected. A total of 2,907 individuals were screened for eligibility (Fig. 1), and 215 cases were excluded due to the missing value in the socio-demographical characteristics or physical health conditions; results 2,692 participants included at baseline were followed up on cognitive function every two years from 2017 to 2023. Thus, the final sample size included 1,969 participants who finished at least two waves of measurement of CF and were enrolled in the final analysis. The inclusion criteria for the participants’ enrollment were (1) living in the local address for more than 6 months; (2) aged 55 years and over; (3) written informed consent provided; and (4) completed with CF examination. The exclusion criteria were: (1) a history of severe mental illness and (2) unable to complete the survey due to severe hearing, vision, speech impairments, or other serious physical illnesses. As shown in Table S3, we compared the difference between those included participants and excluded participants in the socio-demographical characteristics at baseline. It was found that the proportion of those aged 65 and above and females with lower educational attainment and monthly income was slightly lower among the included group, while the proportion of those living alone and married was slightly higher among the excluded group.

Fig. 1.

The flowchart of the participant’s enrollment

Measurements

Cognitive function

The Mini-Mental State Examination (MMSE) scale was used to assess CF (including orientation, memory, attention and calculation, language, and executive function) [37] at baseline and during the follow-up period. This instrument is the most widely used instrument for assessing CF and screening for cognitive impairment and has high sensitivity and specificity globally [38]. Education was significantly associated with MMSE scores, and the cutoff points of MMSE varied following an individual’s educational attainment among Chinese older adults [39]. The low CF at baseline was defined according to Peterson’s criteria [40] with modification as Cui et al. suggested [39]: MMSE ≤ 17 for those with no formal school education; MMSE ≤ 20 for those with primary school education (1–6 years of education); and MMSE ≤ 24 for those with junior high school education or above (≥ 6 years of education). To compare MMSE across different educational levels and measurement years, we employed group-centering (Xgroup-centered​=X − µ​) to generate c-MMSE scores for the education stratum (Chinese classification: without formal education, 1–6 years of education, ≥ 6 years of education). Then, based on individual MMSE scores, each test score is centralized by subtracting the mean within the groups.

Social capital

SC was measured at baseline using a series of standard instruments. The social cohesion subscale consists of four items on a 5-point Likert scale: (1) People around here are willing to help their neighbors; (2) People in my neighborhood generally get al.ong with each other; (3) People in my neighborhood can be trusted; (4) People in my neighborhood share the same values. The social interaction was measured using a 5-item Likert scale: (1) You and other people in the community (village) help each other; (2) When a neighbor is not at home or going out, you can help him look after the house or property; (3) people in the community (village) talk about each other’s matters; (4) participate in group activities together with people in the community (village); (5) communicate with each other on the street. The civic participation subscale inquires whether the individuals participate in six types of organizations or social activities: (1) Hobby activity groups (e.g., senior university, senior activity center); (2) volunteer groups; (3) government departments (e.g., neighborhood committees, street offices); (4) party organizations (e.g., party branches); (5) work-related organizations (e.g., trade unions, retirement management committees); (6) religious organizations [21, 41]. Both instruments are widely used to assess the social capital of Chinese adults [42]. Spiritual capital was measured using a five-item scale that consists of the frequency of attendance of organizational religious activities (religious services, prayer groups, scripture study groups, etc.), non-organizational religious activities (private prayer, scripture study, listening to religious radio, and so forth), and the intrinsic religiosity (which reflects the degree of personal religious commitment or motivation). The Chinese version of the scale had acceptable psychometric properties [43]. The Cronbach’s alpha in the present sample was 0.846 for social cohesion, 0.828 for social interaction, and 0.918 for spiritual capital.

Social cohesion and social interaction were divided into high and low levels based on the median scores of social cohesion and social interaction. Taking part in any kind of activity was defined as a high level of civic participation. A total score greater than 24 was defined as high spiritual capital [44].

Covariates

The socio-demographical characteristics at baseline including age (≥ 55), sex, ethnicity, household registration, living status, educational attainment (junior school and above, primary, illiterate), monthly income; health-related behaviors (smoking and drinking); physical health conditions (obesity (BMI score ≥ 28 kg/m²), age-adjusted Charlson comorbidity index (aCCI), and cognitive function at baseline. The degree of illness comorbidity was divided into three categories: no/mild comorbidity (aCCI score 0–1), moderate comorbidity (aCCI score 2–3), and severe comorbidity (aCCI score ≥ 4) [45], the scoring calculation method for aCCI can be found in Table S4.

Statistical analysis

First, we used the conditional latent class growth model (LCGM) [46] to identify the latent trajectories of CF. A final model was selected based on [47]: (1) the Akaike information criterion (AIC) and the Bayesian information criterion (BIC), where the lowest AIC and BIC values indicated better model fitness. While the sample size influences the AIC and BIC, the aBIC accounts for the sample size by adjusting for a large sample size. (2) Entropy is used to measure the accuracy of classification ranging from 0 to 1, with higher values of entropy indicating better classification. (3) P for the Vuong- Lo–Mendell likelihood ratio test (VLRT) and P for the bootstrap likelihood ratio test (BLRT) that were used to identify the meaningful trajectories.

Second, the Chi-square test was performed to determine the proportion of covariates in different CF trajectory groups. A multiple logistic regression model was used to examine the associations between different CF trajectories and covariates (reference: stable group). Odds ratios (ORs) and 95% confidence interval (95% CI) were calculated.

Third, the algorithm of generalized linear modeling (GLM) using the H2OFlow of H2O.ai (http://docs.h2o.ai/h2o/latest-stable/h2o-docs/flow.html) was employed to explore the importance of influencing predictors in different trajectory groups. Standardized coefficients were calculated to compare the relative contributions of different predictors to the model. In this, the data were randomly split into training and validation datasets (ratio = 0.7), and 10-fold cross-validation was set [48].

The LCGM was estimated in Mplus (Version 8.4), and other statistical analyses were performed using SPSS version 25.0 (Version 25 for Windows). An alpha level of < 0.05 was considered statistically significant for a two-tailed test.

Results

Trajectories of CF

We used the conditional LCGM with linear, nonlinear, and time parameter-free estimation models to fit the optimal latent CF trajectories after controlling for sex. As shown in Table S1, the entropy of the 3-class recorded the highest values in all the models, and the AIC, BIC, and aBIC of the 3-class model were acceptable, and the BLRT indicated statistical significance. Compared to the free-estimate and nonlinear models, the AIC and BIC of the linear model are relatively low, and the VLRT and BLRT are still significant. Thus, the LCGM with a linear 3-class model was selected as the final latent trajectory model (Fig. 2). The first class constantly maintained a higher and stable CF level, named the “Stable group”, with a proportion of 72.5%; The second class showed a trend of decline (Slope= -0.09, P = 0.002), was named the “Slow decline group”, with a proportion of 25.8%. The third class decreased dramatically over time and was named the “deterioration group”, with a proportion of 1.7%.

Fig. 2.

Trajectories of CF. (VLRT = Vuong–Lo–Mendell-Rubin Likelihood Ratio Test; BLRT = Parametric Bootstrapped Likelihood Ratio Test

Demographic characteristics

The demographic characteristics of the participants showed in Table 1. The final sample consisted of 1086 (55.15%) males and 883 (44.85%) females, with an average age of 64.1 ± 5.5 years and a total follow-up person-year of 13113.94. The majority of the participants were ethnic minorities (82.68%) and lived with their families (87.10%), over half were illiterate (53.53%). Over two-thirds of the participants had a low level of civic participation (73.59%) and social cohesion (65.01%), more than half of the participants had a low level of social interaction (53.48%), and with a high level of spiritual capital (57.33%). A small proportion of the participants reported smoking (11.68%) or drinking (6.25%). 15.24% reported severe aCCI and 15.8% met the low CF criteria.

Table 1.

Characteristics of the total sample and by CF trajectory groups [n (%)]

Variable	Trajectory group				χ²	P
	Tatal (n = 1969)	Deteriorate (n = 30)	Slow decline (n = 470)	Stable (n = 1469)
Age					49.104	< 0.001
≥65	883 (44.85)	26 (86.67)	257 (54.68)	600 (40.84)
< 65	1086 (55.15)	4 (13.33)	213 (45.32)	869 (59.16)
Sex					79.776	< 0.001
Female	1086 (55.15)	13 (43.33)	343 (72.98)	730 (49.69)
Male	883 (44.85)	17 (56.67)	127 (27.02)	739 (50.31)
Ethnicity					1.720	0.423
Minorities	1628(82.68)	23 (76.67)	396 (84.26)	1209 (82.30)
Han	341(17.32)	7 (23.33)	74 (15.74)	260 (17.70)
Household registration					7.804	0.020
Urban	718(36.47)	11 (36.67)	146 (31.06)	561 (38.19)
Rural	1251(63.53)	19 (63.33)	324 (68.94)	908 (61.81)
Educational attainment					29.660	< 0.001
junior and above	443(22.50)	8 (26.67)	66 (14.04)	369 (25.12)
Primary	472(23.97)	7 (23.33)	109 (23.19)	356 (24.23)
Illiterate	1054 (53.53)	15 (50.00)	295 (62.77)	744 (50.65)
Living alone					1.470	0.479
Yes	254(12.90)	2 (6.67)	57 (12.13)	195 (13.27)
No	1715(87.10)	28 (93.33)	413 (87.87)	1274 (86.73)
Marital status					2.295	0.317
Married	1569(79.69)	24 (80.00)	363 (77.23)	1182 (80.46)
Not married	400(20.31)	6 (20.00)	107 (22.77)	287 (19.54)
Monthly income a					23.786	< 0.001
≥2000	705(35.80)	9 (30.00)	125 (26.60)	571 (38.87)
< 2000	1264(64.20)	21 (70.00)	345 (73.40)	898 (61.13)
Obesity					6.573	0.037
Yes	342(17.37)	5 (16.67)	100 (21.28)	237 (16.13)
No	1627(82.63)	25 (83.33)	370 (78.72)	1232 (83.87)
Smoking					15.475	< 0.001
Yes	230(11.68)	4 (13.33)	31 (6.60)	195 (13.27)
No	1739(88.32)	26 (86.67)	439 (93.40)	1274 (86.73)
Drinking					6.164	0.046
Yes	123(6.25)	2(6.67)	18(3.83)	103(7.01)
No	1846(93.75)	28(93.33)	425(96.17)	1366(92.99)
aCCI					28.530*	< 0.001
None/mild	300(15.24)	1 (3.33)	52 (11.06)	247 (16.81)
Moderate	1452(73.74)	17 (56.67)	364 (77.45)	1071 (72.91)
Severe	217(11.02)	12 (40.00)	54 (11.49)	151 (10.28)
Low CF					623.449	< 0.001
Yes	301(15.8)	10(33.3)	249(51.3)	52(3.5)
No	1668(84.7)	20(66.7)	241(48.7)	1417(96.5)
Spiritual capital					24.243	< 0.001
High	1131(57.44)	9 (30.00)	236 (50.21)	886 (60.31)
Low	838(42.56)	21 (70.00)	234 (49.79)	583 (39.69)
Civic participation					22.102	< 0.001
High	1449(73.59)	19 (63.33)	309 (65.74)	1121 (76.31)
Low	520(26.41)	11 (36.67)	161 (34.26)	348 (23.69)
Social interaction					0.215	0.898
High	916(46.52)	14 (46.67)	223 (47.45)	679 (46.22)
Low	1053(53.48)	16 (53.33)	247 (52.55)	790 (53.78)
Cohesion					0.963	0.618
High	689(34.99)	11 (36.67)	173 (36.81)	505 (34.38)
Low	1280(65.01)	19 (63.33)	297 (63.19)	964 (65.62)

Note Chi-square test was used to examine the relationship between different CF trajectories and socio‐demographic characteristics, health‐related behaviors, and health status; ^aChinese Yuan; aCCI = age‐adjusted Charlson comorbidity index; CF = Cognitive function; *:Due to insufficient sample size, Fisher’s exact test was adopted

The Chi-square tests showed significant differences between different trajectory groups in terms of age, sex, household registration, educational attainment, monthly income, obesity, smoking status, alcohol consumption, aCCI, low CF at baseline, spiritual capital, and civic participation.

Multinomial logistic regression

As shown in Table 2, when the covariates were controlled, compared with the stable group, those aged 65 years and above were likely to have a higher risk of deterioration group (OR = 7.65, 95% CI = 2.17–27.01, p = 0.002) and slow decline group (OR = 1.99, 95% CI = 1.49–2.65, p < 0.001). Being female (OR = 2.11, 95% CI = 1.59–3.10, p < 0.001), with obesity (OR = 1.41, 95% CI = 1.02–1.95, p = 0.040), and moderate aCCI (OR = 1.57, 95% CI = 24.93–142.20, p = 0.040) had an increased risk of being slow decline group. Those with low CF at baseline had the highest risk of deterioration group (OR = 54.54, 95% CI = 1.49–2.65, p < 0.001) and slow decline group (OR = 27.12, 95% CI = 18.97–38.77, p < 0.001). Meanwhile, those with high levels of spiritual capital are less likely to be in the deterioration group (OR = 0.36, 95% CI = 0.14–0.93, p = 0.036). High civic participation (OR = 0.61, 95% CI = 0.46–0.81, p = 0.001) was associated with a lower risk of a slow decline trajectory.

Table 2.

Multinomial logistic regression analysis on the trajectory of cognitive function among older adults (N = 1969)

Variable	Deterioratea		Slow declinea
	OR(95%CI)	P	OR(95%CI)	P
Age (≥65)	7.65(2.17–27.01)	0.002	1.99(1.49–2.65)	< 0.001
Sex (female)	0.43(0.16–1.15)	0.092	2.22(1.59–3.10)	< 0.001
Ethnicity (Han)	1.27(0.33–4.91)	0.726	0.67(0.42–1.07)	0.094
Household registration (urban)	0.95(0.31–2.94)	0.932	0.96(0.66–1.38)	0.807
education (primary)	1.15(0.41–3.21)	0.790	1.14(0.82–1.58)	0.425
education (junior and above)	0.87(0.28–2.75)	0.817	0.67(0.43–1.04)	0.072
Live alone (yes)	0.58(0.11–3.23)	0.536	1.12(0.73–1.71)	0.604
Marital status (married)	0.98(0.32–3.01)	0.973	1.00(0.70–1.41)	0.984
Monthly income (≥2000)b	0.64(0.22–1.85)	0.406	0.77(0.55–1.07)	0.122
Obesity (yes)	1.20(0.41–3.52)	0.737	1.41(1.02–1.95)	0.040
Smoking (yes)	0.62(0.14–2.68)	0.519	0.89(0.52–1.52)	0.674
Drinking (yes)	0.45(0.06–3.34)	0.435	0.89(0.44–1.81)	0.742
aCCI (moderate)	0.89(0.09–9.10)	0.924	1.57(1.02–2.42)	0.040
aCCI (severe)	3.19(0.28–35.85)	0.348	1.49(0.83–2.66)	0.185
Low CF (yes)	59.54(24.93–142.20)	< 0.001	27.12(18.97–38.77)	< 0.001
Spiritual capital (high)	0.36(0.14–0.93)	0.036	0.84(0.63–1.13)	0.251
Civic participation (high)	0.46(0.20–1.09)	0.077	0.61(0.46–0.81)	0.001
Social interaction (high)	1.04(0.46–2.38)	0.923	1.09(0.84–1.43)	0.509
Cohesion (high)	1.64(0.69–3.88)	0.264	1.18(0.90–1.55)	0.240

Note^a Reference Group: Stable group; ^b Chinese Yuan; CI = Confidence Interval; aCCI = age-adjusted Charlson Comorbidity Index; CF = Cognitive function

Importance of predictors of CF trajectories

As shown in Fig. 3, low CF at baseline was the first important factor (standardized coefficient = 0.253), followed by age, civic participation, sex, monthly income, spirituality, and obesity. The participants were female, older, and with obesity, which are negatively associated with stable CF trajectories. However, those with higher monthly incomes, spirituality, and civic participation are positively associated with stable CF trajectories.

Fig. 3.

The variable importance in trajectories of CF. The factors are ranked based on their importance in the trajectories of CF. The blue bar represents the factor having a positive coefficient on the CF trajectories, while the red bar represents the factor having a positive coefficient on the CF trajectories

Sensitivity analysis

Considering the number of follow-up times, we conducted a sensitivity analysis for participants who completed at least three waves of measurement (N = 964), and a similar trajectory model persisted (Table S1, Figure S1). As shown in Table S5, multinomial regression analysis demonstrated that spiritual capital was associated with the slow decline trajectory (OR = 0.68, 95% CI: 0.49–0.94) and the deterioration trajectory (OR = 0.19, 95% CI: 0.05–0.70). As shown in Figure S2A, after excluding participants under age 65, the shapes of trajectories were largely maintained. In addition, after excluding participants with low CF at baseline. The proportion of CF trajectories changed: rapid decline (12.5%), high-stable (18.1%), and stable ( 69.4%) (Figure S2B). As shown in Table S6, the multinomial regression analysis demonstrated that high civic participation was negatively associated with rapid decline trajectory (OR = 0.42, 95% CI: 0.26–0.67) and stable trajectory (OR = 0.69, 95% CI: 0.48–0.99). High cohesion was associated with a lower risk of a rapid decline trajectory (OR = 0.63, 95% CI: 0.41–0.98).

Discussion

This study is the first to examine the link between social capital and CF trajectories using longitudinal data collected repeatedly in minority communities in mainland China.We identified three distinct CF trajectories, and two dimensions of social capital (spiritual capital, civic participation) are positively associated with a favorable CF trajectory (stable), indicating that interventions targeting improved social capital in the community may be beneficial for maintaining better CF among older adults. Our findings also emphasize the importance of caring for and supporting vulnerable populations, such as women, low CF, and those with comorbidities in rural areas. This study provides primary evidence to support the development of social participation and spiritual practices in alleviating cognitive impairment in Chinese older adults, especially in ethnic minority areas [35].

Our findings indicate that CF is a heterogeneous process composed of multiple potential developmental trajectories among older adults, the finding consistent with previous studies [49]. We provide evidence that the potential CF trajectories of most community-dwelling individuals could be roughly divided into three to four different classes in different studies and populations [50]. In this study, a large proportion of older adults had a stable trajectory, and nearly one-third had a slow decline trajectory over the year of follow-up. We identified a relatively uncommon and accelerated rate of decline (deterioration trajectory) groups that we are supposed to serve as potential prototypes for probing AD and dementia in the future [51]. The minimum proportion of this trajectory is less than 2% among those aged 55 and over, and 3.1% among those aged over 65 years old, this is similar to the study in the USA that found 2.8% of participants experienced a rapid cognitive decline trajectory among those aged 70 years and above [8]. Hu et al. found that about 3.5% of participants showed a rapid cognitive decline trajectory among those aged 80 years and above in mainland China [10]. However, in whole trajectories, we did not find a group whose CF improved over time. The possible reasons might be that an individual’s cognitive performance, such as memory, recall, execution, and learning, will gradually decline with age throughout the whole life cycle [52]. The specificity of brain nerve cells and brain regions will decrease with age, leading to a decline in CF among older adults [53].

Consistent with the previous studies, civic participation, and spiritual capital may be potential protective resources for CF maintenance [54, 55]. Individuals with frequent community participation may suffer less from psychological stress through several pathways, including providing social support, accessing information and other resources, enhancing self-confidence, and bolstering self-worth [56]. Additionally, spiritual practices such as Taichi, Baduanjing, and Qigong have beneficial effects on CF, particularly in memory enhancement [57]. However, unlike other studies [58, 59], this study has not found the effect of social cohesion and social interaction on the CF trajectory. One possible explanation is the lack of consensus on the measurement standards of social cohesion in existing research. Another reason for these results may be differences in the study design (cross-sectional and longitudinal), demographic characteristics, cultural background, and confounding factors.

Our findings also identified several risk factors for adverse CF trajectories, including advanced age, being female, obesity, comorbidities, and low CF at baseline. In the context of traditional Chinese culture, females have fewer opportunities for higher socioeconomic status than males do in early life in remote provinces in rural areas (where there is a preference for male offspring over female ones) [38, 60]. Older adults with comorbidities have a higher risk of cognitive impairment [61], and obesity was associated with reduced cognitive scores [62]. Balasubramanian et al. suggested that obesity promotes the development of cognitive impairment and increases the incidence of AD [63]. In addition, individuals with obesity in early life were associated with lower executive function, lower MMSE scores, and impaired memory, ability, and executive function in later life [64]. Previous studies have concluded that cognitive decline is commonly considered a high-risk disease for AD in clinical practice [65] compared with healthy age-matched individuals, and nearly 10% of MCI adults will progress to AD and dementia based on the meta-analysis of 41 robust inception cohort studies [66].

Limitations

Although this study used a population-based prospective design in a large community population of older adults, there were several study limitations. First, a 2-year interval follow-up and the loss to follow-up may affect the accuracy of the findings, for example, those who develop cognitive problems are more likely to become lost to follow-up or refuse to participate, which may result in underestimating the association. Despite our efforts to address the concerns and provide more comprehensive analyses, the fact that a significant portion of our major findings relies on a relatively small percentage (less than 2%) of cases still casts doubt on the external validity of our results. Second, our participants were enrolled from one province of China using a nonrandom sampling method, so the generability of the findings was limited. Third, we did not consider the dynamic changes in health-related behaviors, disease factors over time, and the possible bidirectional causality of which individuals with cognitive impairment may also impair their ability to engage in social participation, interaction, and church services. Finally, cognitive scores may improve with repeated administration of the tests (i.e., practice effects), even the two-year interval time for each measurement.

Conclusions

The CF trajectories are heterogeneous among older adults, and civic participation, and spiritual capital are slightly associated with a lower risk of developing unfavorable trajectories (slow decline and deterioration) among older adults in mainland China. These findings provide primary clues to support the development of social-integrated interventions for vulnerable groups, such as older women, with obesity, those with more comorbidities, and those with low CF at baseline.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Abbreviations

CF

cognitive function

LCGM

latent class growth model

SC

social capital

MMSE

Mini-Mental State Examination

NHAC

Ningxia Healthy Aging Cohort

aCCI

age-adjusted Charlson comorbidity index

OR

Odds ratio

95% CI

95% confidence interval

AIC

Akaike information criterion

BIC

Bayesian information criterion

VLRT

Vuong–Lo–Mendell likelihood ratio test

BLRT

bootstrap likelihood ratio test

Author contributions

Mengyun Xiong and Zhaojing Wen: data collection, visualization, writing–original draft and formal analysis. Zhizhong Wang: Conceptualization and funding acquisition. Zhizhong Wang and Yan Tong: data curation and project administration. Xue Hu, Yan Tong, and Wanrui Ma: data collection. Zhizhong Wang and Wanrui Ma: writing–review & editing. Zhizhong Wang, Yan Tong, and Wanrui Ma: methodology. All authors have read and agreed to the published version of the manuscript.

Funding

This work was supported by the National Natural Science Foundation of China under Grant number 81860599; and Dongguan Outstanding Young Medical Talent Recruitment and Training Program, the Dongguan Science and Technology of Social Development Program (Grant No. 20231800940182).

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

This study was performed in line with the principles of the Declaration of Helsinki. The study was approved by the Institutional Review Board of Ningxia Medical University (approval numbers 2015151 and 2018115). All participants provided written informed consent before enrollment.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.