Abstract 208: Improvement in Prediction of Readmission after CABG with Cause-Specific Readmission Models

Background: Hospital readmissions after cardiac surgery occur in up to 20% of cases and are associated with significant cost and morbidity. Most statistical models for readmission have poor predictive power, and hence very little clinical value to identify at-risk patients. We hypothesized that the predictive value of readmission models could be improved upon when cause-specific readmission is examined. Methods and Results: Data from a large statewide registry was queried for all patients who underwent isolated coronary artery bypass grafting (CABG) between Oct 1, 2005 and September 30, 2015. A total of 32,694 patients were included. There was a total of 4,745 (14.5%) hospital readmissions. Seven hundred and eighty-four (16.5%) readmissions were due to infection, 680 (14.3%) due to heart failure, and 3,281 (69.2%) due to other causes. A total of 80 variables were included in the analysis. Multivariate logistic regression analysis was utilized for all-cause readmission and for cause-specific readmission (heart failure and infection). Risk factors for all-cause readmission were: Older age, female gender, ASA 4/5, low hemoglobin (<11.1 g/dL), diabetes with complications, chronic lung disease, chronic liver disease, heart failure, history of stroke, sleep apnea and earlier year of surgery. AUC was 0.6418. Risk factors for readmission for heart failure were: Older age, diabetes with complications, chronic lung disease, history of heart failure, and high glucose (>135 mg/dL). AUC was 0.7252. Risk factors for readmission for infection were: Older age, female gender, chronic lung disease, morbid obesity, high glucose (>135 mg/dL) and hypertension. AUC was 0.6292. Conclusions: The predictive value for readmission after isolated CABG improved when readmission for heart failure was examined, but not for infection. However, data from cause-specific models was more likely to yield clinically relevant risk factors than with all-cause readmission.