Specific Motor and Cognitive Performances Predict Falls during Ward-Based Geriatric Rehabilitation in Patients with Dementia

Abstract

The aim of this study was to identify in-hospital fall risk factors specific for multimorbid hospitalized geriatric patients with dementia (PwD) during hospitalization. Geriatric inpatients during ward-based rehabilitation (n = 102; 79.4% females; 82.82 (6.19) years of age; 20.26 (5.53) days of stay) were included in a comprehensive fall risk assessment combining established clinical measures, comprehensive cognitive testing including detailed cognitive sub-performances, and various instrumented motor capacity measures as well as prospective fall registration. A combination of unpaired t-tests, Mann-Whitney-U tests, and Chi-square tests between patients with ("in-hospital fallers") and without an in-hospital fall ("in-hospital non-fallers"), univariate and multivariate regression analysis were used to explore the best set of independent correlates and to evaluate their predictive power. In-hospital fallers (n = 19; 18.63%) showed significantly lower verbal fluency and higher postural sway (p < 0.01 to 0.05). While established clinical measures failed in discriminative as well as predictive validity, specific cognitive sub-performances (verbal fluency, constructional praxis, p = 0.01 to 0.05) as well as specific instrumented balance parameters (sway area, sway path, and medio-lateral displacement, p < 0.01 to 0.03) significantly discriminated between fallers and non-fallers. Medio-lateral displacement and visuospatial ability were identified in multivariate regression as predictors of in-hospital falls and an index combining both variables yielded an accuracy of 85.1% for fall prediction. Results suggest that specific cognitive sub-performances and instrumented balance parameters show good discriminative validity and were specifically sensitive to predict falls during hospitalization in a multimorbid patient group with dementia and an overall high risk of falling. A sensitive clinical fall risk assessment strategy developed for this specific target group should include an index of selected balance parameters and specific variables of cognitive sub-performances.

Keywords: dementia; discriminative validity; fall prediction; hospitalization; instrumented motor assessment; neuro-psychological assessment; predictive validity.

Figure 1

Receiver operating characteristic curve for the two individual and the combined predictors for in-hospital falls in patients with dementia. Presented are receiver operating curves of different models for predicting future in-hospital falls in PwD: Model 1 used ‘constructional praxis’ (AUC: 0.64), model 2 used ‘medio-lateral displacement’ (AUC: 0.74), and model 3 combined these two variables to a combined model (0.75).