Augmenting Braden Scores With Predictive Modeling to Reduce Hospital-Acquired Pressure Injuries: A Clinical Pilot Study.

OBJECTIVE: To evaluate the efficacy of a machine learning predictive model integrating the Braden Scale with patient demographic and care-related features, in aiding risk identification and guiding targeted interventions aimed at preventing hospital-acquired pressure injuries (HAPIs). METHODS: This clinical pilot study was conducted at a regional medical center in the midwestern United States and included admitted patients ages 18 and older. A gradient boosting trees algorithm was trained using historical patient data and deployed in the electronic health record. Following a validation period, a 6-month pilot study on 2 medical floors involving 3808 admissions evaluated the model's impact. Patients identified by the model as high-risk received supplemental rounding and preventive care. Patients on control floors received standard care. RESULTS: The primary outcome measures were HAPI events per 1000 census days and the distribution of injury severities. A statistically significant shift in the proportion of deep tissue injuries to full-thickness PIs was observed on pilot floors (odds ratio, 0.09; 95% CI, 0.01-0.88; P =.03). A reduction in the odds of census days with HAPI events to those without was also observed but was not significant in the span of the pilot (odds ratio: 0.69; 95% CI, 0.34-1.39; P =.38). CONCLUSIONS: Predictive modeling shows promise for enhancing HAPI risk identification and guiding targeted prevention in a clinical setting. Further research is needed to confirm that the results are generalizable across diverse health care systems.