A dynamic risk model for inpatient falls.

Area of Science:

• Gerontology
• Clinical Pharmacy
• Health Informatics

Background:

• Inpatient falls are a significant safety concern, particularly for patients receiving fall risk-increasing drugs (FRIDs).
• Existing fall risk assessment tools may not adequately capture risks associated with FRID use.
• Developing a precise, automated tool is crucial for effective fall prevention.

Purpose of the Study:

• To construct and validate a fall risk prediction model specifically for inpatients exposed to FRIDs.
• To compare the predictive performance of the new model against the established Morse Fall Scale (MFS).

Main Methods:

• A retrospective cohort study analyzed 75,036 admissions over 22 months with FRID exposure within the first 5 hospital days.
• Logistic regression identified factors influencing fall risk.
• The developed model was internally validated and compared with the MFS model.

Main Results:

• The study evaluated 220,904 patient-days of FRID exposure, with oxycodone, morphine, and hydromorphone being the most common.
• The new model captured 30.9% of falls within the 90th percentile of risk scores (C statistic, 0.69), outperforming the MFS model (20.2%, C statistic, 0.62).
• Key predictors included a history of falling, overestimation of ambulation ability, and a composite comorbidity measure.

Conclusions:

• The novel risk model for inpatient falls demonstrated superior predictive performance compared to the MFS.
• All identified risk factors were derived from electronic health record data, enabling automated, real-time identification of high-risk patients.