An ecological approach to fall risk factors for preventive interventions design: a pilot study.

Abstract

Background Recent literature reinforces that interventions for fall prevention should include multimodal training [1]. However, even multimodal training tends to focus on exercises separately in single physical, cognitive or environ- mental hazards variables. An ecological approach to explain phenome- na’s such as fall occurrence, underlines not only the accumulative effect of isolated variables but also interactions between different variables. Objective To reduce a set of correlated variables to a smaller number that may explain fall occurrence. Methods 187 older adults aged 65 to 96 years were assessed for falling risk factors. Principal component analysis (PCA) was performed including data from the 6-minute walk test (6MWT) [2], Gait Scale [3], Fullerton Advanced Bal- ance Scale (FAB) [4], body composition - fat body mass percentage (FBM %), Mini-Mental State Examination (MMSE) [5], Environmental Hazards Scale (EH) [6], health conditions (HC), time up and go test (TUG) [2] and the Epworth Sleepiness Scale (ESS) [7]. Factors with eigenvalues of at least 1.0 were retained and a varimax rotation was used to produce inter- pretable factors. A binary regression analysis was performed using the forward stepwise (conditional) technique to identify the most significant components explaining fall occurrence. Receiver operating characteristics (ROC) curves were used to assess the discriminative ability of the logistic model. Results Three principal components were identified. In component 1, the domin- ant variables concerned physical and cognitive fit (6MWT, Gait Scale, FAB, MMSE, TUG), in component 2 dominant variables concerned health and environmental conditions (FBM %, EH, HC), whereas in component 3, the dominant variable concerned alertness (ESS). These components ex- plained cumulatively 37%, 56% and 70% of the variance in fall occur- rence. Logistic regression selected components 1 (OR: 0.527; 95% CI: 0.328–0.845) and 2 (OR: 1.614; 95% CI: 1.050–2.482) as predictive of falls. The cut-off level yielding the maximal sensitivity and specificity for pre- dicting fall occurrence was set as 0.206 (specificity = 72.7%, sensitivity = 47.7%, and the area of the ROC curve was computed as 0.660 (95% CI: 0.564-0.756). Conclusions This pilot study showed that multiple correlated variables for fall risk as- sessment can be reduced to three uncorrelated components character- ized by: physical and cognitive fit; health and environmental conditions; and alertness. The first two were the main determinants of falls. Recom- mendations: Interventions for fall prevention should privilege multimodal training including tasks that work simultaneously physical fitness, cogni- tive fitness and alertness, considering participant’s specific health and en- vironmental conditions.