In older persons, falls are the leading cause of injuries, often resulting in emergency room visits, serious injuries, and possibly even death. Medications are a major risk factor for falls. Because we lack tools to assess individualized risks, general practitioners (GPs) struggle with fall related medication management for seniors, and senior patients are not properly equipped to engage in the joint management of their medications. Our aim in this project is to develop and evaluate a comprehensive data-driven science approach for valid prediction of personalized risk of falling that effectively supports joint medication management between seniors and GPs. The project has two objectives. First, we aim to develop and validate prediction models from electronic health records for assessing individualized risk of medication-related falls. Data science challenges include free text analysis; accounting for missing values; searching medication hierarchies; engineering new predictors, and understanding limitations of our approach. Second, we aim to develop and evaluate a joint medication management strategy for older patients and GPs, consisting of a clinical decision support system (CDSS) and a patient portal. We evaluate the effects of this strategy on changes in the quality of shared decision-making during a medication review consultation, medication management, and patient outcomes. The learnings from this project and the architecture underpinned by predictive modelling to support both GPs and patients can also be applied to other major health problems in the future.
Westerbeek et al. (Mon,) studied this question.