Fall prediction using decision tree analysis in acute care units

[Purpose] To present an accurate and straight-forward system of fall prediction by performing decision tree analysis using both the fall assessment sheet and Berg balance scale (BBS). [Participants and Methods] The participants in this retrospective study were inpatients from acute care units. We extracted the risk factors for falls from the fall assessment and performed a decision tree analysis using the extracted fall risk factors and BBS score. [Results] "History of more than one fall in the last 1 year", "Muscle weakness", "Use of a walking aid or wheelchair", "Requires assistance for transfer", "Use of Narcotics", "Dangerous behavior", and "High degree of self-reliance" were fall risk factors. The decision tree analysis extracted five fall risk factors, with an area under the curve of 0.7919. Patients with no history of falls and who did not require assistance for transfer or those with a BBS score ≥51 did not fall. [Conclusion] Decision tree-based fall prediction was useful and straightforward and revealed that patients with no history of falling and those who did not require assistance for transfer or had a BBS score ≥51 had a low risk of falling.