Community involvement, trust, and health-related outcomes among older adults in India: a population-based, multilevel, cross-sectional study

Abstract

Objectives

this study examined whether individual and contextual measures of structural and cognitive social capital were associated with six health-related outcomes across older adults in India.

Methods

data were collected from a representative sample of adults aged 60 and above across India in 2011–12 (n = 9,174). Personal community involvement and having someone to trust represented individual measures of structural and cognitive social capital. These measures were then aggregated to represent contextual measures of social capital, that is, the mean village level of community involvement and the village proportion having someone to trust. To examine associations between all four social capital indicators and six outcomes including self-rated health, psychological well-being, subjective well-being, memory, activities of daily living (ADL), and instrumental activities of daily living, we fit pooled, sex-stratified, and place-stratified multilevel regression models and adjusted for demographic and socio-economic factors.

Results

personal community involvement was positively associated with all outcomes among the full sample. Adjusted odds ratios ranged from 1.05 (95% CI 1.02; 1.08) for good self-rated health to 1.42 (95% CI 1.33; 1.53) for high-ADL function. Personally having someone to trust was associated with four outcomes. Village-level social capital measures were less frequently associated with outcomes than personal social capital measures. Association strength between six health-related outcomes and individual and contextual measures of structural and cognitive social capital varied, however, among older people in India by sex, place and outcome.

Discussion

interventions to promote healthy ageing by increasing community involvement and trust may need to be tailored to population subgroups.

Introduction

Social capital may significantly impact individual health and well-being through social influence, provisioning of social support, access to material resources and social engagement [1–3]. These pathways are strongly associated with physiological stress, psychological well-being, health-related behaviours and exposure to infectious diseases [1, 4]. Structural social capital relates to externally observable aspects of social organisation [5–7] and can be ascertained via indicators of community involvement. In contrast, cognitive social capital is conceptualised as shared norms, values, attitudes and beliefs [8, 9] and can be measured as feelings of interpersonal trust. In addition, social capital can be measured at the personal level as well as at a higher population level (e.g. the contextual village level).

There is a limited understanding of the extent to which social capital is associated with health and well-being across low- and middle-income countries according to recent reviews [10–12]. Although a few existing studies in developing countries have found that various measures of social capital are positively associated with health and well-being, such associations are not consistently observed [12–18]. The dearth of knowledge in this area is even greater when focusing on older adult health.

India is soon to be the most populous country with a substantial ageing population and an ever-increasing economic burden of diseases among older adults. Research is needed on the extent to which personal and contextual measures of structural and cognitive social capital are associated with health and well-being outcomes for older men and women in rural and urban locations across India. Thus, this study uses a multilevel framework to investigate the associations between social capital (measured at the individual level and village level, separately) and health-related outcomes, including well-being, functionality, psychological distress and cognitive ability, among a nationally representative sample of older adults in India from 2011 to 2012.

We first estimate the associations between outcomes and personal community involvement and personally having someone to trust while adjusting for other explanatory factors. We hypothesize that both individual-level measures of social capital are positively associated with all outcomes. We then assess whether the average village community involvement and the village proportion of having someone to trust are independently associated with outcomes while adjusting for the individual social capital measures and other factors. We hypothesize that both contextual social capital measures are positively associated with all outcomes.

Method

Sample

Data are from the study ‘Building a Knowledge Base on Population Ageing in India’ conducted in Himachal Pradesh, Kerala, Maharashtra, Odisha, Punjab, Tamil Nadu and West Bengal [19]. These seven states represent different regions and have a higher share of older adults compared with the national average [19]. The study design included a random sample of 1,280 households per state equally split between rural and urban areas. Each state sample was divided into 40 villages in rural areas and 40 wards in urban areas, though rural areas in Himachal Pradesh included 48 villages to account for a large number of small villages. Sixteen households with at least one adult aged 60+ years were randomly selected per village or ward. All persons aged 60+ years in the sampled households were eligible. In total, the sample included 10,604 older adults from 8,792 households spread across 568 villages or wards. The individual response rate was 93%. After excluding respondents with missing information on variables in this study (average missingness per variable was 0.31%), the final sample consisted of 9,174 older adults, nested within 565 villages or wards and 145 districts in India. Table 1 provides a brief description of the socio-demographic distribution of older adults included in this study. Further details about the final sample distribution by gender and location across several background characteristics are available in Supplementary Tables 1 and 2, available at Age and Ageing online.

Table 1.

Distribution of older adults in India in 2011/12 who were a part of the Building a Knowledge Base on Population Ageing in India study (N = 9,174)

	N	%
Age
60–64 years	3,271	35.7
65–69 years	2,538	27.7
70–74 years	1,586	17.3
75–79 years	851	9.3
80 years and above	928	10.1
Sex
Male	4,351	47.4
Female	4,823	52.6
Currently married
No	3,721	40.6
Yes	5,453	59.4
Living with children
No	2,618	28.5
Yes	6,556	71.5
Education
None	4,183	45.6
Below 5 years	1,884	20.5
6–10 years	2,308	25.2
11+ years	799	8.7
Wealth quintile
Lowest	1,756	19.1
Second	1,837	20.0
Middle	1,821	19.8
Fourth	1,844	20.1
Highest	1,916	20.9

Outcomes

This study examined six outcomes, collected by tools previously used in the Indian context, to create both binary and linear measures of self-rated health, psychological well-being, subjective well-being, cognitive ability, ability to do activities of daily living (ADLs) and ability to do instrumental activities of daily living (IADLs) [20–23]. Self-rated health was elicited using a five-point scale. Responses were recoded as 0 (representing ‘poor’ or ‘fair’) or 1 (representing ‘good’, ‘very good’ or ‘excellent’). Psychological well-being was measured through a 12-item General Health Questionnaire about recent experiences with stressful symptoms [24]. Each item was coded as 0 (less than usual or no more than usual) or 1 (more than usual or much more than usual). Responses were summed; higher values represented higher psychological well-being. Scores of 6+ were recoded as 1, indicating higher psychological well-being, and score of 5 or less as 0 [25]. Subjective well-being was measured with a nine-item inventory [26]. Responses were coded such that respondents reporting a negative experience were assigned a score of 0 and any other response as 1. Item scores were summed; higher scores indicated higher levels of subjective well-being. Scores of 6+ were recoded as 1 (representing higher subjective well-being) and scores of 5 or less were recoded as 0.

Memory function was measured by number of words recalled [27]. Five or more words represented higher memory function. Basic functioning was measured by whether assistance was needed to do six ADLs including bathing, dressing, feeding, getting in and out of a bed/chair, toileting and continence. Not needing any assistance for an ADL was given a score of 1 and 0 otherwise. Scores across were summed; higher scores implied a higher level of independence. A score of 6 was recoded as 1 representing full independence in ADLs and 5 or less was recoded as 0. Complex functioning was measured through eight IADLs where 1 was given for higher functionality or a 0 otherwise. Scores were summed. Scores of 6+ were recoded as 1 for high-IADL individuals and scores 5 or less as 0 for low-IADL individuals.

Explanatory variables

Five questions about frequency of personal community involvement (Appendix A) were used to create an individual-level measure of structural social capital. Scores were summed to create a continuous variable representing personal community involvement (Cronbach’s alpha = 0.77). Separately, the question ‘do you have someone you can trust and confide in?’ was used to provide a binary, individual-level measure of cognitive social capital, referred to as personally having someone to trust. We then created two continuous social capital variables at the village/ward level (hereafter referred to as village level) by computing the mean community involvement score in the respondent’s village and the proportion of people in the respondent’s village who personally had someone to trust.

Age was categorised as 60–64, 65–69, 70–74, 75–79 and 80+ years. Marital status was currently married vs. never married, widowed and divorced/separated. Dummy variables indicated whether the respondent was living with children, and whether the respondent was currently working. Social group was recorded as scheduled tribe (ST), scheduled caste (SC), other backward classes (OBC) and others. Religion was recorded as Hindu, Muslim, Christian, Sikh or other. Education was categorised as none, 1–5, 6–10 and 11 or more years. Information on 30 assets and housing characteristics was used to construct household wealth quintiles, a proxy for household economic status [19]. Sex, location (rural vs. urban) and state were also included.

Statistical analysis

Using MLwiN and the runmlwin module in STATA, we ran two multilevel logistic regression models for all binary outcomes (main analysis) and then ran two multilevel linear regression models for all continuous outcomes (sensitivity analysis). In addition, we fit models with everyone and then stratified by sex and location, separately. Model 1 uses the following equation for estimation:

$${\mathrm{Y}}_{\mathrm{ijk}}{=\mathrm{\alpha }}_{\mathrm{0}}{+\mathrm{\alpha }}_{\mathrm{1}}{\mathrm{S}}_{\mathrm{ijk}}{+\mathrm{\alpha }}_{\mathrm{2}}{\mathrm{X}}_{\mathrm{ijk}}{+\mathrm{\mu }}_{\mathrm{j}}{+\mathrm{\nu }}_{\mathrm{k}}{+\mathrm{\epsilon }}_{\mathrm{ijk}}$$ (1)

Y is the dependent variable for individual i (level 1) in village j (level 2) in district k (level 3), S represents the two individual-level social capital variables, and X is the vector of other explanatory variables. The α’s are the fixed parameters to be estimated, whereas μ_j and ν_k are village-specific and district-specific random effects. ε_ijk is the random component of the error term.

Model 2 uses the following equation for estimation:

$${\mathrm{Y}}_{\mathrm{ijk}}{=\mathrm{\beta }}_{\mathrm{0}}{+\mathrm{\beta }}_{\mathrm{1}}\left({\mathrm{S}}_{\mathrm{ijk}}{-\mathrm{S}}_{\mathrm{jk}}\right){+\mathrm{\beta }}_{\mathrm{2}}{\mathrm{S}}_{\mathrm{jk}}{+\mathrm{\beta }}_{\mathrm{3}}{\mathrm{X}}_{\mathrm{ijk}}{+\mathrm{\mu }}_{\mathrm{j}}{+\mathrm{\nu }}_{\mathrm{k}}{+\mathrm{\epsilon }}_{\mathrm{ijk}}$$ (2)

S represents both the personal- and village-level social capital indicators. For this analysis, the continuous village-level indicators are ‘group-centred’ (i.e. S_ijk − S_jk) to address collinearity inherent in individual- and village-level social capital terms [16]. The parameter β₁ can then be interpreted as the association between personal social capital and the outcome above and beyond any village-level association. Moreover, the equation can be rewritten as follows:

$${\mathrm{Y}}_{\mathrm{ijk}}{=\mathrm{\beta }}_{\mathrm{0}}{+\mathrm{\beta }}_{\mathrm{1}}{\mathrm{S}}_{\mathrm{ijk}}+\left({\mathrm{\beta }}_{\mathrm{2}}{–\mathrm{\beta }}_{\mathrm{1}}\right){\mathrm{S}}_{\mathrm{jk}}{+\mathrm{\beta }}_{\mathrm{3}}{\mathrm{X}}_{\mathrm{ijk}}{+\mathrm{\mu }}_{\mathrm{j}}{+\mathrm{\nu }}_{\mathrm{k}}{+\mathrm{\epsilon }}_{\mathrm{ijk}}$$ (3)

The Wald test (of the difference between parameters β₂ and β₁) indicates whether there is an independent association of village-level social capital with the outcome beyond the personal associations.

Results

Supplementary Tables 3 and 4, available at Age and Ageing online show the distribution of the six outcomes and both individual-level and village-level social capital variables across socio-demographic characteristics. Table 2 presents the estimates of associations between social capital variables and good self-rated health and low psychological distress from sex-stratified and place-stratified models. Personally having someone to trust was not associated with good self-rated health in any model for any stratum. For men and for urban people, there was a positive association between personal community involvement and good self-rated health. The village proportion having someone to trust was associated with good self-rated health for all strata except for urban dwellers. Personal community involvement and personally having someone to trust were both positively associated with low psychological distress for all strata. The village proportion having someone to trust was associated with low psychological distress only for urban dwellers with an independent effect as indicated by a Wald test.

Table 2.

Logistic regression estimates of the association between measures of social capital and good self-rated health, and separately low psychological distress, among older adults (60+ years) in rural and urban locations in India

	Good self-rated health				Low psychological distress
	Model 1		Model 2		Model 1		Model 2
	AOR	95% CI	AOR	95% CI	AOR	95% CI	AOR	95% CI
Men
Personal community involvement	1.05*	(1.00–1.09)	1.07**	(1.02–1.12)	1.12***	(1.05–1.18)	1.14***	(1.08–1.21)
Personally having someone to trust	1.08	(0.88–1.33)	0.95	(0.76–1.19)	1.30*	(1.03–1.64)	1.27	(0.99–1.63)
Average level of community involvement in one’s village			0.88b	(0.76–1.01)			0.89a	(0.75–1.06)
Proportion having someone to trust in one’s village			3.14***,b	(1.69–5.83)			1.71	(0.86–3.40)
Women
Personal community involvement	1.02	(0.97–1.06)	1.02	(0.97–1.07)	1.07*	(1.01–1.13)	1.08*	(1.02–1.14)
Personally having someone to trust	1.07	(0.90–1.28)	0.98	(0.81–1.18)	1.43***	(1.19–1.72)	1.40**	(1.16–1.70)
Average level of community involvement in one’s village			1.01	(0.89–1.15)			1.01	(0.86–1.19)
Proportion having someone to trust in one’s village			3.23***,b	(1.75–5.98)			1.82	(0.93–3.55)
Rural
Personal community involvement	1.03	(0.99–1.07)	1.03	(0.99–1.08)	1.11***	(1.06–1.17)	1.13***	(1.07–1.19)
Personally having someone to trust	1.14	(0.95–1.37)	1.04	(0.86–1.25)	1.34**	(1.11–1.62)	1.33**	(1.10–1.62)
Average level of community involvement in one’s village			1.00	(0.86–1.16)			0.91a	(0.76–1.07)
Proportion having someone to trust in one’s village			4.31***,b	(2.18–8.52)			1.52	(0.75–3.06)
Urban
Personal community involvement	1.06*	(1.01–1.11)	1.07**	(1.02–1.12)	1.10**	(1.03–1.17)	1.10**	(1.03–1.17)
Personally having someone to trust	0.90	(0.74–1.11)	0.87	(0.71–1.07)	1.50**	(1.19–1.89)	1.43**	(1.13–1.81)
Average level of community involvement in one’s village			0.93	(0.77–1.11)			1.08	(0.85–1.38)
Proportion having someone to trust in one’s village			1.70	(0.72–4.01)			4.36**,a	(1.55–12.2)

*P < 0.05; **P < 0.01; ***P < 0.001.^a and ^b denote Wald test significance at 5 and 1% levels. It discerns whether there is a statistically significant difference between parameters of village level and individual level of the social capital variables.

All models adjusted for age, marital status, living arrangement, social group, religion, education, work, wealth quintile and state. Sex was adjusted for in the place-stratified model and place was adjusted for in the sex-stratified model. All estimates accounted for survey design by using a three-level random intercepts model.

Table 3 presents estimates of associations between social capital variables and high levels of subjective well-being and good memory function from sex-stratified and place-stratified models. Personally having someone to trust and personal community involvement were both associated with highly subjective well-being across all strata. For men [AOR 1.19 (1.01; 1.39)] and for rural dwellers [AOR 1.25 (1.07; 1.46)], the village average level of community involvement was associated with high levels of subjective well-being, although the Wald test did not indicate a statistically significant independent effect for either group. In contrast, the village proportion having someone to trust was associated with high levels of subjective well-being across all strata. These associations represented independently statistically significant effects as indicated by Wald tests. Although both personal social capital variables were associated with good memory function across all strata, neither village-level social capital variable was associated with good memory function across any strata.

Table 3.

Logistic regression estimates of the association between measures of social capital and highly subjective well-being, and separately good cognitive ability, among older adults (60+ years) in rural and urban locations in India

	Highly subjective well-being				Good cognitive ability
	Model 1		Model 2		Model 1		Model 2
	AOR	95% CI	AOR	95% CI	AOR	95% CI	AOR	95% CI
Men
Personal community involvement	1.20***	(1.14–1.26)	1.20***	(1.13–1.27)	1.10***	(1.05–1.15)	1.13***	(1.08–1.18)
Personally having someone to trust	1.52***	(1.21–1.91)	1.36*	(1.07–1.74)	1.32**	(1.06–1.64)	1.34*	(1.06–1.69)
Average level of community involvement in one’s village			1.19*	(1.01–1.39)			0.91b	(0.80–1.04)
Proportion having someone to trust in one’s village			3.14**,b	(1.64–5.98)			1.26	(0.71–2.24)
Women
Personal community involvement	1.13***	(1.07–1.20)	1.13***	(1.06–1.19)	1.12***	(1.06–1.17)	1.13***	(1.07–1.19)
Personally having someone to trust	1.51***	(1.25–1.82)	1.35**	(1.11–1.64)	1.39**	(1.15–1.69)	1.37**	(1.12–1.67)
Average level of community involvement in one’s village			1.14	(0.98–1.32)			1.02	(0.89–1.17)
Proportion having someone to trust in one’s village			5.21***,b	(2.77–9.81)			1.65	(0.87–3.13)
Rural
Personal community involvement	1.17***	(1.11–1.23)	1.16***	(1.10–1.22)	1.10***	(1.05–1.15)	1.12***	(1.07–1.17)
Personally having someone to trust	1.57***	(1.30–1.89)	1.46***	(1.20–1.77)	1.34**	(1.11–1.63)	1.39**	(1.14–1.71)
Average level of community involvement in one’s village			1.25**	(1.07–1.46)			0.93a	(0.81–1.06)
Proportion having someone to trust in one’s village			3.55***,b	(1.84–6.85)			0.91	(0.50–1.66)
Urban
Personal community involvement	1.18***	(1.11–1.25)	1.19***	(1.12–1.26)	1.14***	(1.09–1.20)	1.15***	(1.09–1.21)
Personally having someone to trust	1.35**	(1.08–1.70)	1.26	(1.00–1.59)	1.37**	(1.10–1.70)	1.28*	(1.02–1.60)
Average level of community involvement in one’s village			1.12	(0.91–1.38)			1.15	(0.96–1.38)
Proportion having someone to trust in one’s village			4.63**,a	(1.77–12.1)			4.41**,b	(1.84–10.6)

*P < 0.05; **P < 0.01; ***P < 0.001. ^a and ^b denote Wald test significance at 5 and 1% levels. It discerns whether there is a statistically significant difference between parameters of village level and individual level of the social capital variables.

All models adjusted for age, marital status, living arrangement, social group, religion, education, work, wealth quintile and state. Sex was adjusted for in the place-stratified model and place was adjusted for in the sex-stratified model. All estimates accounted for survey design by using a three-level random intercepts model.

Table 4 presents the estimates of associations between social capital variables and high-ADL functioning and high-IADL functioning. Personally having someone to trust was associated with high-ADL functioning for men [AOR 1.57 (1.05; 2.35)] and rural dwellers [AOR 1.51 (1.10; 2.07)] only, whereas personal community involvement was associated with high-ADL functioning across all strata. The village average level of community involvement was associated with high-ADL functioning for women only [AOR 1.24 (1.02; 1.52)], though no independent effect was indicated by a Wald test. The village proportion of having someone to trust was associated with high-ADL functioning for men only [AOR 2.67 (1.02; 6.99)], though no independent effect was indicated by a Wald test. Personal community involvement was associated with high-IADL functioning for all strata, but personally having someone to trust was not. The village proportion having someone to trust was associated with high-IADL functioning for both men and women, and for rural dwellers but not urban dwellers, separately, according to the stratified analyses. Independent effects were indicated by Wald tests for the statistically significant associations. For men only, a higher average village level of community involvement was associated with lower IADL functioning, which appeared to be an independent association.

Table 4.

Logistic regression estimates of the association between measures of social capital and highly subjective well-being, and separately good cognitive ability, among older adults (60+ years) in rural and urban locations in India

	High-ADL function				High-IADL function
	Model 1		Model 2		Model 1		Model 2
	AOR	95% CI	AOR	95% CI	AOR	95% CI	AOR	95% CI
Men
Personal community involvement	1.47***	(1.33–1.62)	1.55***	(1.40–1.72)	1.10***	(1.05–1.15)	1.14***	(1.08–1.19)
Personally having someone to trust	1.65**	(1.14–2.39)	1.57*	(1.05–2.35)	0.99	(0.79–1.24)	0.90	(0.71–1.14)
Average level of community involvement in one’s village			1.04b	(0.82–1.32)			0.80**,b	(0.69–0.92)
Proportion having someone to trust in one’s village			2.67*	(1.02–6.99)			2.45**,b	(1.25–4.79)
Women
Personal community involvement	1.33***	(1.21–1.46)	1.34***	(1.22–1.48)	1.20***	(1.14–1.26)	1.24***	(1.17–1.31)
Personally having someone to trust	1.17	(0.87–1.57)	1.12	(0.81–1.53)	1.18	(0.97–1.43)	1.04	(0.85–1.28)
Average level of community involvement in one’s village			1.24*	(1.02–1.52)			0.96b	(0.83–1.10)
Proportion having someone to trust in one’s village			1.55	(0.69–3.49)			4.39***,b	(2.31–8.34)
Rural
Personal community involvement	1.36***	(1.25–1.48)	1.39***	(1.27–1.52)	1.15***	(1.10–1.20)	1.18***	(1.12–1.24)
Personally having someone to trust	1.58**	(1.17–2.13)	1.51*	(1.10–2.07)	1.12	(0.93–1.36)	0.97	(0.79–1.19)
Average level of community involvement in one’s village			1.17	(0.94–1.45)			0.88b	(0.76–1.02)
Proportion having someone to trust in one’s village			2.33	(0.96–5.62)			7.72***,b	(3.93–15.2)
Urban
Personal community involvement	1.47***	(1.32–1.64)	1.50***	(1.34–1.67)	1.20***	(1.14–1.26)	1.23***	(1.16–1.29)
Personally having someone to trust	1.09	(0.75–1.58)	1.04	(0.70–1.54)	0.86	(0.69–1.07)	0.83	(0.67–1.05)
Average level of community involvement in one’s village			1.27	(0.97–1.66)			0.91b	(0.76–1.09)
Proportion having someone to trust in one’s village			1.63	(0.51–5.25)			1.34	(0.56–3.18)

*P < 0.05; **P < 0.01; ***P < 0.001. ^a and ^b denote Wald test significance at 5% and 1% levels. It discerns whether there is a statistically significant difference between parameters of village level and individual level of the social capital variables.

Notes: All models adjusted for age, marital status, living arrangement, social group, religion, education, work, wealth quintile and state. Sex was adjusted for in the place-stratified model and place was adjusted for in the sex-stratified model. All estimates accounted for survey design by using a three-level random intercepts model.

Estimates of associations between outcomes and all variables included in all models are presented in Supplementary Tables 5–10 (binary outcomes) and 11-16 (linear outcomes), available at Age and Ageing online.

Discussion

Structural and cognitive social capital, as captured by community involvement and trust, were positively associated with several indicators of health and well-being among older adults in India. However, more associations existed between the outcomes and personal social capital indicators as compared to the number of associations between outcomes and village-level social capital indicators, though results depended on the stratum (men or women, and rural or urban residence). These findings suggest that social capital may have uneven impact across sex and location.

Analyses revealed that women’s psychological well-being was positively associated with the contextual measure of trust while this finding was not shown for men. An overall environment of greater trust may provide greater opportunities for women’s social engagement thus creating a favourable pathway for psychological well-being through mutual sharing and interactions. In contrast, personal structural capital among men was positively related with good self-rated health. No such association was found for women. Perhaps older women have fewer opportunities for social engagements with their community than men. Structural social capital at the personal level therefore may not necessarily bear a direct influence on how women rate their own health. Contextualising these findings within known rural/urban health differences in India as well as strong gender norms due to a paternalistic culture and traditional kinship systems [28–30] is important to consider when developing effective interventions for social capital and health in different communities in India.

This study is the first to have examined the association between social capital and several health-related outcomes among older adults in India from a population-based perspective as well as explored the robustness of associations by sex and rural–urban location. Furthermore, this study offers novel information about whether social capital may differently influence health when it is measured at the individual level vs. a higher population level, and whether a contextual measure of social capital could render an independent effect which is beyond the effect attributable to the individual measure of the same social capital indicator. However, there are a few primary limitations to the present analysis. First, the data do not permit studying causal direction. Second, they represent only seven states though information is from regions in India with the greatest ageing populations. Third, it cannot be assumed that associations between social capital and health-related outcomes are generalisable across developed and developing countries [10–12]. Future research should assess the mechanisms through which social capital may differently impact health outcomes across sex, place and population level.

Conclusion

This study provides a comprehensive appraisal of the associations between several health-related outcomes among older adults in India and structural and cognitive social capital. Personal measures of community involvement and trust were positively associated with most domains of subjective, psychological and functional health, but not always. In addition, associations with contextual measures of community involvement and trust varied by outcome as well as by sex and location. Given the highly gendered and often resource-constrained context, this evidence provides insights about the potential need to tailor the design of social capital-based interventions targeting healthy ageing among older adults in India.

Key points.

• Personal and contextual levels of structural and cognitive social capital may indicate older adults at risk of poor health.

• Community involvement and having someone to trust can identify vulnerable groups of older adults in India depending on sex.

• Higher social capital within a village exhibits varying associations with health-related outcomes for older adults in India.

• Interventions promoting health ageing through community involvement and interpersonal trust may need tailoring by subgroup

Funding

Research reported in this publication was supported by the National Institute on Aging of the National Institutes of Health under Award Number P30AG024409. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Conflicts of interest

None.