In-hospital fall prediction using machine learning algorithms and the Morse fall scale in patients with acute stroke:...

Area of Science:

• Medical informatics
• Clinical neurology
• Machine learning in healthcare

Background:

• Falls are a common and dangerous accident in medical settings, particularly for acute stroke patients.
• Predicting and preventing falls is crucial for patient safety and improving prognosis.
• Existing tools like the Morse Fall Scale (MFS) are used for fall risk assessment.

Purpose of the Study:

• To develop and evaluate a machine learning (ML) model for predicting falls in acute stroke patients.
• To compare the accuracy of the developed ML model against the established Morse Fall Scale (MFS).

Main Methods:

• A retrospective nested case-control study involving 8462 acute stroke patients.
• Six machine learning algorithms were employed: regularized logistic regression, support vector machine, naïve Bayes (NB), k-nearest neighbors, random forest, and extreme-gradient boosting (XGB).
• 156 fall events were matched with 934 control cases for analysis.

Main Results:

• The Morse Fall Scale (MFS) showed an AUROC of 0.76 for fall prediction.
• Extreme-gradient boosting (XGB) achieved the highest AUROC of 0.85, outperforming MFS.
• XGB and Naïve Bayes (NB) demonstrated the highest F1 score of 0.44.

Conclusions:

• Machine learning algorithms, particularly XGB, show high accuracy in predicting fall risk among acute stroke patients.
• ML models offer comparable or superior performance to the MFS for fall risk screening.
• These findings suggest ML can facilitate accurate and efficient fall screening in clinical practice.