Fig. 3

Effect of physical activity on cognitive frailty according to prespecified subgroups. CI, confidence interval. SPPB, Short Physical Performance Battery. Ordinal logistic regression was used as described in the Method section. For the primary analysis, ordinal variable 1 was created by assigning 0 for no cognitive frailty (i.e., non-frail without mild cognitive impairment [MCI]), 1 for pre-frail without MCI, 2 for frail without MCI, 3 for non-frail with MCI, 4 for pre-frail with MCI, and 5 for cognitive frailty (i.e., frail with MCI). For the secondary analysis, ordinal variable 2 was created by assigning 2 for non-frail with MCI and 3 for frail without MCI while the other values remained unchanged. The models were adjusted for intervention, field center, sex, and baseline levels for cognitive frailty. The interaction between subgroup and intervention were included to determine whether the intervention effect was the same in each subgroup. In a stratified analysis, we determined the odds ratio for the intervention effect and its 95% confidence intervals (CIs) for each subgroup