Objectives: To investigate the effect of an exercise program on falls in intermediate and high-level long-term care (LTC) residents and to determine whether adherence, physical capacity, and cognition modified outcomes.

Design: Randomized controlled trial. Setting and Participants: Residents (n = 520, aged 84 ± 8 years) from 25 LTC facilities in New Zealand. Methods: Individually randomized to Staying UpRight, a physical therapist–led, balance and strength group exercise program delivered for 1 hour, twice weekly over 12 months. The control arm was dose-matched and used seated activities with no resistance. Falls were collected using routinely collected incident reports.

Results: Baseline fall rates were 4.1 and 3.3 falls per person-year (ppy) for intervention and control groups. Fall rates over the trial period were 4.1 and 4.3 falls ppy respectively [P = .89, incidence rate ratio (IRR) 0.98, 95% CI 0.76, 1.27]. Over the 12-month trial period, 74% fell, with 63% of intervention and 61% of the control group falling more than once. Risk of falls (P = .56, hazard ratio 1.08, 95% CI 0.85, 1.36) and repeat falling or fallers sustaining an injury at trial completion were similar between groups. Fall rates per 100 hours walked did not differ between groups (P = .42, IRR 1.15, 95% CI 0.81, 1.63). Program delivery was suspended several times because of COVID-19, reducing average attendance to 26 hours over 12 months. Subgroup analyses of falls outcomes for those with the highest attendance (≥50% of classes), better physical capacity (Short Physical Performance Battery scores ≥8/12), or cognition (Montreal Cognitive Assessment scores ≥ 18/30) showed no significant impact of the program.

Conclusions/Implications: In intermediate and high-level care residents, the Staying UpRight program did not reduce fall rates or risk compared with a control activity, independent of age, sex, or care level. Inadequate exercise dose because of COVID-19–related interruptions to intervention delivery likely contributed to the null result.