Abstract
ABSTRACT: Readmission is an increasingly important focus for improvement regarding quality, value, and patient burden in our surgical patient population. We hypothesized that inpatient harm events increase the likelihood of readmission in surgical patients. We created a system-wide inpatient registry with 30-day readmission. A surgical subset was created, and harm events were tracked through the electronic health record system. Between 2015 and 2017, 37,048 surgical patient encounters met inclusion criterion. A total of 2,887 patients (7.69%) were readmitted. After multiple logistic regression of the highly significant harm measures, seven harm measures remained statistically significant (p < .05). Those with the three highest odds ratios were mucosal pressure ulcer, Clostridium difficile, and glucose <40. Incorporating harm measures to the traditional risk, predictive model for 30-day readmission improved our model performance (area under the ROC curve from 0.68 to 0.71). This study demonstrated that inpatient hospital-based harm events can be electronically monitored and used to predict 30-day readmission.