A Pragmatic Preoperative Prediction Score for Nonhome Discharge After Cardiac Operations

Background: Targeted rehabilitation of patients at risk for nonhome discharge (NHD) after an operation is an appealing area for quality improvement. We sought to identify the primary predictors of NHD after cardiac operations to generate a robust preoperative prediction tool for those at greatest risk. Methods: The medical records of 5,253 patients undergoing cardiac operations between January 1, 2012, and March 31, 2016, were reviewed. Two models of NHD were created: a preoperative model using only preoperative predictors and a postoperative model using the same preoperative predictors and including postoperative adverse outcomes and hospital length of stay. We also determined whether NHD also reduced 30-day hospital readmission. Results: A multivariable logistic regression model allowed robust identification of NHD using only preoperative variables of age, sex, marital status, obesity, comorbidities, addictions, psychiatric disease, and planned operation (area under the curve = 0.820, r² = 0.349). Postoperative factors associated with NHD, including hospital length of stay and the occurrence of a neurologic event, were included and improved model performance (area under the curve = 0.860, r² = 0.439), with integrated discrimination improvement of 7.5%. We observed an overall all-cause readmission rate of 12%. Patients with NHD had a higher readmission rate (16% vs 11%; p < 0.0001), as did patients with longer hospital stays, postoperative atrial fibrillation, neurologic event, or infection (all p < 0.0001). Conclusions: We identified preoperative risk factors for NHD after cardiac operations and developed a pragmatic NHD prediction score with high accuracy. Addition of postoperative risk factors for NHD only modestly improved prediction. NHD does not decrease the readmission rate after cardiac operations.