Multicomponent Intervention and Long-Term Disability in Older Adults: A Nonrandomized Prospective Study.

Abstract

To evaluate the long-term association between a multicomponent intervention program and disability in socioeconomically vulnerable older adults. This was a nonrandomized prospective intervention trial. The setting was a community. Participants included older Koreans living alone or receiving government assistance from a low-income program. The intervention was a 24-week multicomponent program compromising group exercise, nutritional supplementation, management of depression, deprescribing, and home hazard reduction (n = 187) versus usual care (n = 196). The number of dependencies in 17 basic and instrumental activities of daily living was measured every 3 months for 30 months (range: 0-17; greater values indicated worse disability). Inverse probability weighting Poisson regression was used to model the number of dependencies to adjust for confounding bias and higher dropout rates of those with greater disability. The study population had a mean age of 76 years, and 26% were men. During the 30-month follow up, 17 died (n = 8, intervention; n = 9, control), 62 (n = 16, intervention; n = 46, control) were institutionalized or received nursing home care, and 34 (n = 15, intervention; n = 19, control) were lost to follow up. After inverse probability weighting, the mean number of dependencies at baseline was 1.21 and 1.29 for the intervention group and the control group, respectively (P = .80). The intervention group had fewer dependencies than the control group, but the difference was attenuated over time: 1.08 versus 1.60 at 6 months (P = .04), 1.29 versus 1.87 at 12 months (P = .03), 1.62 versus 2.17 at 18 months (P = .06), 2.08 versus 2.51 at 24 months (P = .18), and 2.73 versus 2.90 at 30 months (P = .67). A 24-week multicomponent intervention was associated with a slower progression of disability; however, the diminishing association from 24 months and beyond suggests that reassessment and intervention may be necessary. Due to a lack of randomization, our findings should be interpreted with caution.