Artificial Intelligence in Fall Risk Assessment: A Systematic Literature Review

Several studies have investigated the use of Artificial Intelligence (AI) in fall risk assessment.However, these studies lack a clear systematic evaluation of the optimal Machine Learning (ML) methods for clinicians to model fall risk applications.This study aims to provide a literature review of the advantages, disadvantages, and applications of AI.ML in fall risk assessment within inpatient and outpatient settings.We conducted a systematic review of 15 primary research articles covering both clinical datadriven and sensory data-driven modalities.Here, we report consistent advantages in using clinical Electronic Medical Record (EMR) data for ML algorithms.These studies resulted in the highest quality ML models with superior predictive accuracies and model performance.On the other hand, kinematic sensory data is prone to variable outcomes as ML classifiers are often sensitive to small changes in that sensory data.These data collection methods are more time-intensive and less applicable to the real world.Furthermore, Adaptive Boosting and Random Forest ML classifiers were the most common in producing the highest-performing algorithms.This study suggests that, by using ML, clinicians can reliably streamline and predict fall risk at higher accuracies than current fall risk assessments.Further research may leverage these ML methods to assess variables in fall prevention, bridging the gap between fall risk assessment and prevention strategies.