Comparison of Two Different Models to Predict Fall Risk in Hospitalized Patients

Area of Science:

• Healthcare Informatics
• Patient Safety
• Clinical Risk Assessment

Background:

• Fall prevention is crucial for patient safety and healthcare economics.
• Automated electronic medical record (EMR) models can aid in predicting and mitigating inpatient falls.
• Validating EMR-based predictive models is essential for improving patient care.

Purpose of the Study:

• To validate the reliability of an EMR-based computerized predictive model (ROF Model) for inpatient falls.
• To compare the predictive accuracy of the ROF Model against the Johns Hopkins Fall Risk Assessment Tool (JHFRAT).
• To assess the feasibility of transitioning to automated fall risk prediction systems.

Main Methods:

• An observational study compared the ROF Model and JHFRAT predictions against actual falls over eight months.
• Descriptive statistics analyzed score distributions and fall risk categorization accuracy.
• Data included 35,709 inpatient encounters, with 329 falls recorded.

Main Results:

• The overall inpatient fall rate was 0.92%.
• The ROF Model identified 60.8% of fallers as high risk (1.82% fall rate), while JHFRAT identified 75.4% (1.39% fall rate).
• ROF Model demonstrated higher specificity (69.7% vs. 49.2%) but lower sensitivity (60.8% vs. 79.3%) than JHFRAT, with similar C-statistics (0.717 vs. 0.702).

Conclusions:

• The ROF Model's performance in predicting inpatient falls is comparable to the JHFRAT.
• These findings support the adoption of more automated fall risk assessment processes.
• Further prospective studies are needed to confirm if implementing the ROF Model reduces inpatient falls.