Development of a predictive inpatient falls risk model using machine learning

Area of Science:

• Gerontology
• Health Informatics
• Machine Learning in Healthcare

Background:

• Traditional fall risk assessments focus on risk factors, potentially overlooking crucial mitigating elements.
• Machine learning offers advanced capabilities for analyzing diverse variables to enhance patient risk identification.

Purpose of the Study:

• To develop a predictive model for identifying patients at risk of falls, incorporating a fall prevention variable.
• To evaluate the impact of including a fall prevention variable on the model's predictive performance.

Main Methods:

• A retrospective cohort study involving 22,515 adult internal medicine patients.
• Application of machine learning techniques, specifically the Two-Class Bayes Point Machine algorithm.
• Extraction of variables from electronic medical records.

Main Results:

• The model incorporating a fall prevention variable (Model-A) demonstrated superior performance.
• Model-A achieved higher sensitivity (0.74 vs. 0.71), specificity (0.82 vs. 0.74), and Area Under the Curve (AUC) (0.82 vs. 0.78) compared to the model without it (Model-B).

Conclusions:

• Fall prevention is a critical variable for accurate fall risk detection.
• The developed model serves as a decision-support tool for nurses, improving patient fall risk identification and decreasing workload through electronic medical record integration.