A ‘case-mix’ approach to understand adherence trajectories for a falls prevention exercise intervention: A longitudinal cohort study

Objective: We identified adherence-based case-mixes from participants’ longitudinal adherence to falls prevention exercise interventions over 12 months. Second, we identified modifiable baseline predictors (cognition, mobility and functional status) based on participants' case-mix adherence trajectories. Study design and outcome measures: This study was a 12-month longitudinal secondary analysis of data from 172 participants who received the Otago Exercise Program (OEP) in a randomized controlled trial. Adherence to the OEP was ascertained monthly via self-report. Case-mixes, groups of individuals who followed similar adherence trajectories, were visually defined using 12-month longitudinal trajectories; we used latent class growth modeling. Baseline predictors of adherence were examined for the following categories: 1) cognition, 2) mobility and 3) functional status. Results: Four distinct case-mixes were identified. The “non-adherent” case-mix (18 %) was distinguished by a non-adherent and decreasing adherence trajectory over time. The “low adherence” case-mix (45 %) did not have complete adherence or consistent adherence over the 12-month follow-up. The “moderate adherence” case-mix (27 %) was characterized by a stable (i.e., non-variable) adherence trajectory with a slightly increasing pattern at midpoint. The “high adherence” case-mix (10 %) demonstrated consistent and high adherence over the 12-month follow-up. For individuals with “moderate adherence”, the Digit Symbol Substitution Test (DSST) significantly predicted adherence (relative risk ratio (RRR) = 1.12 (0.95 CI: 1.0–1.26); p = 0.049). For individuals with “high adherence”, the Digits Forward minus Digits Backward (RRR = 0.43 (0.95 CI: 0.23−0.79); p = 0.002) and Instrumental Activities of Daily Living (RRR = 0.36 (0.95 CI: 0.16−0.81); p = 0.01) significantly predicted adherence. Conclusions: Cognitive profile and activities of daily living at baseline may predict the longitudinal pattern of adherence.