Abstract 18061: Predictors of Hospitalization or Death Due to Heart Failure in Diabetic Patients by Gender in the ACCORD Trial Using Random Survival Forests

Background: Heart failure is an important cause of hospitalization and mortality in patients with type 2 diabetes (T2D). In a prior secondary analysis of publicly available ACCORD trial (NCT00000620) data, we noted increased hospitalization or death due to heart failure (hdHF) with intensive vs. standard glucose-lowering therapy in women, particularly in those with established cardiovascular disease (CVD). The objective of this exploratory analysis was to identify important predictive factors for hdHF in the ACCORD trial using random survival forests (RSF), a machine-learning methodology that effectively analyzes large numbers of covariates. Methods: A total of 240 variables, including demographic, clinical and laboratory data, and their change from baseline during follow-up, were analyzed as potential predictors of hdHF, using RSF in women and men. An RSF of 500 trees were constructed, with each tree constructed on a different bootstrap sample from the original data. The top 20 predictors were included i...