Association between community-level social capital and frailty onset, based on a multilevel logistic regression analysis with multiple imputation approach
| Null model | Model 1 | Model 2 | Model 3 | Model 4 | ||||||
| OR (95% CI) | P value | OR (95% CI) | P value | OR (95% CI) | P value | OR (95% CI) | P value | |||
| Fixed effects | ||||||||||
| Community-level variables | ||||||||||
| Social capital* | ||||||||||
| Civic participation | 0.91 (0.88 to 0.94) | <0.001 | 0.94 (0.90 to 0.97) | 0.001 | 0.94 (0.91 to 0.98) | 0.003 | 0.94 (0.91 to 0.98) | 0.003 | ||
| Social cohesion | 1.01 (0.98 to 1.05) | 0.462 | 0.98 (0.94 to 1.02) | 0.295 | 0.99 (0.95 to 1.03) | 0.483 | 0.99 (0.95 to 1.03) | 0.498 | ||
| Reciprocity | 1.02 (0.98 to 1.06) | 0.277 | 1.03 (0.99 to 1.07) | 0.093 | 1.04 (1.00 to 1.08) | 0.075 | 1.04 (1.00 to 1.08) | 0.073 | ||
| Individual-level variables | ||||||||||
| Frequency of contact with friends (ref: more than once a week) | Less than once a week | 1.19 (1.00 to 1.41) | 0.051 | 1.18 (0.99 to 1.41) | 0.060 | |||||
| Participation in community groups (ref: no) | Yes | 0.74 (0.64 to 0.86) | <0.001 | 0.75 (0.65 to 0.88) | <0.001 | |||||
| Perception of community social cohesion (ref: no) | Yes | 0.59 (0.47 to 0.73) | <0.001 | 0.59 (0.48 to 0.74) | <0.001 | |||||
| Social support (ref: low) | High | 0.59 (0.46 to 0.75) | <0.001 | 0.59 (0.46 to 0.76) | <0.001 | |||||
| Alcohol consumption (ref: never/past) | Currently | 0.88 (0.75 to 1.04) | 0.136 | |||||||
| Smoking history (ref: never/past) | Currently | 1.35 (1.06 to 1.70) | 0.013 | |||||||
| Walking time (ref: <60 min/day) | ≥60 | 0.85 (0.75 to 0.98) | 0.020 | |||||||
| Random effects | ||||||||||
| Community-level variance | 0.194 | 0.147 | 0.063 | 0.066 | 0.064 | |||||
| Median OR | 1.52 | 1.44 | 1.27 | 1.28 | 1.27 | |||||
Model 1 was a crude model; model 2 was adjusted for age, gender, living arrangement, marital status, education, equivalent income, self-rated health, present illness and community-level characteristics (urbanisation, ageing rate and rate of low-level education); model 3 was adjusted for frequency of contact with friends, participation in community groups, perception of community social cohesion and social support in addition to model 2 covariates; model 4 was adjusted for alcohol consumption, smoking history and walking time in addition to model 3 covariates.
All information about the results is provided in online supplemental table 3.
*Estimation per 10 percentage points.