5 The effectiveness of a nation-wide implemented fall prevention intervention in the Netherlands | 77 Statistical analysis Data was analysed using RStudio (Version 2023.06.0). All analyses were performed according to the intention-to-treat principle. Sample size In the general population of older adults, about one in three persons falls at least once per year (137). To obtain a difference in the number of falls per year of 50% less falls in the intervention than in the control group, a minimum of 106 persons were required per group at a power of 0.80, beta of 0.20 and alpha of 0.05. Taking into account a dropout rate of 20%, the total required sample size was a minimum of 256 participants. Because the optimal number of participants in an In Balance training group is 12, we included a total of 264 participants. Demographic characteristics Data were described using means and standard deviations for normally distributed continuous variables, medians and interquartile ranges for non-normally distributed continuous variables, and numbers and percentages for non-continuous variables. Differences between the intervention and control group were tested by a t-test for normally distributed data and Mann-Whitney U test for non-normally distributed data. Missing data imputation To prevent bias due to selective missingness, we imputed both the primary and secondary outcomes using Multiple Imputation by Chained Equations (MICE) with Predictive Mean Matching using the R ‘MICE’ package (240). The number of falls and injuries were imputed per four months. The secondary outcomes were imputed at M0, M4, and M12. The number of datasets was increased until the fraction of missing information was less than 5%, resulting in 20 imputed datasets. Analyses were done on the imputed datasets and results were pooled thereafter using Rubin’s rules (241). Primary and secondary outcomes The effectiveness of the In Balance intervention in comparison with the control group was analysed using generalized linear models for the primary outcomes, and linear mixed-effects models for the secondary outcomes. We included age and sex as confounders in all analyses (67) and physical activity was included in the analyses of the primary outcomes (49). Other variables were considered confounders if the estimate for the randomization group changed by at least 10% after adding that variable to the model. The number of falls and injuries were analysed using generalized linear models with a negative binomial regression. In case the Intraclass Correlation Coefficient (ICC) was 0.2 or higher,
RkJQdWJsaXNoZXIy MjY0ODMw