Abstract 17063: Enhanced Prediction of Heart Failure 30-Day Readmission Risk With Four Readily-Available Clinical Variables

Abstract

Introduction: Heart failure (HF) readmissions, with ~25% within 30 days of discharge, can be reduced with costly post-hospitalization interventions. Determining which patients will benefit from these interventions remains a critical challenge. Hypothesis: We aimed to devise a simple model to predict 30-day HF readmission risk. Methods: Patients ≥21 years of age admitted for HF to a single academic center from July 1st, 2008 to December 31st, 2009 were identified by billing codes for HF. Patients were excluded if they died within 90 days, were transferred, discharged to hospice or prison, or later admitted electively or for a surgical emergency not related to HF. Demographics, co-morbidities, laboratories, vitals, and discharge medications (27 variables) were tested as predictors for risk of 30 day readmission. First, independent univariate testing of categorical variables was performed utilizing a retention p-value of < 0.2. Next, continuous variables and retained categorical variables were tested in a multivariable model with retention of variables with p-value < 0.1. False discovery rate was controlled by the method of Benjamini and Hochberg. Statistical analyses were performed using SAS® software version 9.3. Results: Of 1253 HF admissions, 657 (55.8% men, age 60.9 ± 13.8 men, 66.8 ± 14.5 women) met criteria with complete data. Twenty percent were readmitted within 30 days. Odds for readmission increased 1.6% for every 100 pg/ml increase in BNP (p=0.082), 5.6% for each additional discharge medication (p=0.003), 200% if not discharged on an ACE-I/ARB (p=0.001), and 7% for every 10 mmHg increase in systolic blood pressure at discharge compared to admission (p=0.078). Conclusion: One laboratory value and 3 clinical variables were predictive of 30-day HF readmission risk in a multivariate prediction model. The 3 clinical predictors provoke the idea that “in hospital” care is a predictor and potential target for intervention to reduce 30-day readmission risk. Future validation studies should help define standard-of-care interventions that could still be improved to reduce readmission rates in HF patients as well as address the relative cost of more intense hospital interventions over outpatient interventions.