Falls in older adults Prediction, the role of medication and the use of decision support

The studies described in this thesis focus on personalized falls prevention strategies related to screening tools and optimizing medication use. Existing prediction models for falls were first examined to identify shortcomings of current models. Virtually all examined models were found to have a high risk of bias. Based on identified shortcomings, we developed a new set of prediction models for falls. The discriminative performance of the models was found to be in line with that of previous prediction models. The models’ discriminative performance in geriatric outpatients was found to be similar to that in community-dwelling older adults. Next, we evaluated the screening algorithm that is presented in the World Falls Guidelines (WFG) and compared its predictive performance to that of other commonly used screening tools. The American Geriatrics Society and British Geriatrics Society algorithm and single fall history question showed a similar predictive performance as the WFG algorithm. The 3 Key Questions yielded a higher sensitivity but lower specificity compared to the WFG algorithm. Furthermore, we investigated the associations between potential fall-risk-increasing drugs and falls. Statins and angiotensin-converting enzyme inhibitors were associated with a lower risk of prospective falls. In cross-sectional design, statins and beta blockers were associated with a lower risk of falls, whereas use of proton pump inhibitors was associated with a higher risk of falls. Finally, we describe the design and rationale of the ADFICE_IT trial, which evaluates the use of a clinical decision support systems for optimizing medication use in older patients.