Different types of community-based intervention activities may have differential effects in improving the intrinsic capacity (IC) of older people. This study aims to (i) identify subgroups of older people based on their IC impairments, (ii) examine the differential associations between different types of activity participations and change in IC across subgroups, and (iii) assess whether the activity participation patterns of older people align with the way that would benefit them the most.

Participants were community-dwelling older people aged 60 years or above. They were screened for IC impairments at baseline, and their participation records of different types (cognitive, physical, nutritional, mental, and social) of intervention activities were collected for one year. An aggregated IC score was created based on four IC domains including cognitive (self-rated memory), locomotor (self-rated difficulties in walking), vitality (self-rated weight loss), and psychological (subjective well-being). Cluster analysis was used to group homogenous participants. Mixed-effects regression was used to examine the associations between activity counts (i.e., number of sessions participated) and change in IC. Activity participation patterns were also compared across subgroups.

Data were obtained from 7,357 participants (mean age = 74.72 years). Four clusters were identified, including those who were relatively robust (cluster 1, N = 4,380, 59.5%), those who had cognitive decline (cluster 2, N = 2,134, 29.0%), those who had impaired mobility and vitality (cluster 3, N = 319, 4.3%), and those with poor psychological well-being (cluster 4, N = 524, 7.1%). Overall, activity count was associated with IC improvement (β = 0.073, 95% CI [0.037, 0.108]). However, as regards the cluster-specific results, different types of activities were associated with IC improvement for different specific clusters. For instance, cognitive activity count was associated with IC improvement only for cluster 2 (β = 0.491, 95% CI [0.258, 0.732]). Notably, none of the activity types were associated with IC improvement for cluster 1. Regarding the activity participation patterns, there were no significant differences across the four clusters (Wilk's Λ = 0.997, F = 1.400, p =.138).

IC improvement depended on the activity types and IC status of older people. In view of this, a people-centred and targeted approach should be adopted to maximize the overall benefits of intervention activities.