Abstract 89: Validating The Identification Of Patients At High Risk For Readmission By Examining Hospitalization History
OBJECTIVES: To identify patients at high risk of readmission by validating a simple predictive tool based solely on hospitalization history. BACKGROUND: There is a federal mandate to reduce preventable readmissions. Predicting hospital readmission risk is of great interest to identify which patients would benefit most from transition interventions. Current models perform poorly. Mount Sinai Hospital (MSH) has implemented the Preventable Admissions Care Team (PACT), which has achieved significant results for patients not targeted by other transitional programs. PACT, a social worker-led transitional program, decreased 30-day readmission rate from 30% to 12%, ED visits by 63%, and achieved a 90% primary care show rate at 7-10-days post-discharge. Patients are identified for PACT solely by readmission history: one readmission in 30 days or 2 in 6 months, prior to the index hospitalization. Thus, our objective here was to determine the concordance of predictions based on hospitalization history with a more formal risk model based on factors that characterize patients through demographics and comorbidities. METHODS: Using logistic regression, we developed a risk prediction model for readmission within 30-days. The model, which used patient demographics and co-morbidities (alcohol abuse, AMI, afib, breast cancer, CKD, COPD, CVA, depression, hip fracture, or osteoporosis), was developed in a cohort of Medicare FFS beneficiaries with a high proportion of cardiovascular disease, hospitalized at MSH. The higher the risk score, the higher risk of readmission. Scores of 0-2 had a 7% risk of readmission; scores of 3 or 4 and above 5 had 30-day readmission rates of 19%, and 29%, respectively. We then applied this risk scoring model to patients enrolled in PACT to determine how many of them would have been identified as high risk for readmission based on the regression model. RESULTS: A total of 393 patients were enrolled in PACT in a year and completed a 5 week intervention. Eighty seven percent had 1 cardiac comorbid illness (76% CAD, 66% CHF, and 17% Afib). Readmission data was available through 2010 thus, the analysis was completed for 111 patients. Ninety-five percent of PACT enrollees had a risk score greater than 3: 19 patients (17.1%) had a risk score of 3-4, and 87 patients (78.4%) had a risk score of 5 or greater. CONCLUSIONS: Hospitalization history alone is a reasonable proxy to more formal multivariable regression models in predicting 30-day readmission risk. If substantiated through further study, this could have national implications for real time high risk patient identification for transitional services.