Using EHR data to predict hospital-acquired pressure ulcers: A prospective study of a Bayesian Network model

Background: Hospital-acquired pressure ulcers (HAPU) are common among inpatients and create substantial morbidity, mortality, and costs, but prevention strategies have been only variably effective. Objectives: To develop and assess the impact of a decision support intervention to predict HAPU on the prevalence of ulcers and length of stay in an intensive care unit (ICU), and on the user adoption rate and attitudes. Methods: We compared the HAPU prevalence before and after introducing the intervention, and surveyed the users. We used a Bayesian Network model that was validated in previous studies and linked to the electronic health record system in an application called Pressure Ulcer (PU) Manager. The intervention group included 866 at-risk patients in the surgical ICUs of a tertiary teaching hospital over a 6-month period in 2009 and 2010; the controls were 348 patients from a 6-month baseline period in 2006 and 2007. Results: In the intervention group, the overall HAPU prevalence rate fell from 21% to 4.0% and the ICU length of stay shortened from 7.6 to 5.2 days. After adjustment for primary diagnoses and illness severity, the intervention group was significantly less likely than the baseline group to develop HAPU [odds ratio (OR) = 0.1, p < 0.0001] and had a shorter ICU length of stay (OR = 0.67, p < 0.0001). Data entry regarding ulcer severity and body site increased, and the participants used PU Manager more than once a day for over 80% of eligible cases. Attitudes toward PU Manager were positive. Conclusions: This decision support approach reduced the prevalence of HAPU tenfold and the ICU length of stay by about one-third. Furthermore, the nurses had favorable attitudes toward using it. (C) 2013 Elsevier Ireland Ltd. All rights reserved.