Fall prediction using decision tree analysis in acute care units.

Area of Science:

• Gerontology
• Clinical Nursing
• Health Informatics

Background:

• Falls are a significant risk for inpatients, leading to injury and increased healthcare costs.
• Accurate fall risk assessment is crucial for effective prevention strategies in acute care settings.

Purpose of the Study:

• To develop and validate a straightforward fall prediction system using decision tree analysis.
• To integrate data from fall assessment sheets and the Berg Balance Scale (BBS) for enhanced prediction accuracy.

Main Methods:

• A retrospective analysis of inpatient data from acute care units was conducted.
• Decision tree analysis was employed using extracted fall risk factors and BBS scores.
• Key risk factors identified included fall history, muscle weakness, mobility aid use, transfer assistance needs, medication use (narcotics), behavior, and self-reliance.

Main Results:

• The decision tree model achieved an area under the curve of 0.7919, indicating good predictive performance.
• Five key fall risk factors were identified by the analysis.
• Patients with no prior fall history and no need for transfer assistance, or a BBS score of 51 or higher, demonstrated a low incidence of falls.

Conclusions:

• Decision tree-based fall prediction offers a useful and uncomplicated approach for clinical settings.
• Identifying patients with no fall history, minimal transfer assistance needs, or a high BBS score can effectively identify low-risk individuals.