Readmission Risk Trajectories for Heart Failure Patients Using a Dynamic Prediction Approach (Preprint)
BACKGROUND Patients hospitalized with heart failure suffer the highest rates of 30-day readmission among any clinically-defined patient populations in the United States. Investigation into the predictability of 30-day readmissions can lead to clinical decision-support tools and targeted interventions that can help care providers to improve individual patient care and reduce readmission risk. OBJECTIVE We developed a dynamic readmission risk prediction model that yields daily predictions for hospitalized heart failure patients toward identifying risk trajectories over time. We identified clinical predictors associated with different patterns in readmission risk trajectories. METHODS A two-stage predictive modeling approach combining logistic and beta regression was applied to electronic health record (EHR) data accumulated daily to predict 30-day readmission for a cohort of 534 heart failure patient encounters over 2,750 patient-days. Unsupervised clustering was performed on predictions to uncover time-dependent trends in readmission risk over the patient’s hospital stay. RESULTS Readmission occurred in 107 (20.0%) encounters. The out-of-sample AUC for the two-stage predictive model was 0.73 (±0.08). Dynamic clinical predictors capturing lab results and vital signs had the highest predictive value compared to demographic, administrative, medication and procedural data included. Unsupervised clustering identified four risk trajectory groups: decreasing risk (24.5% encounters), high risk (21.2%), moderate risk (33.1%), and low risk (21.2%). The decreasing risk group demonstrated change in average probability of readmission from admission (0.69) to discharge (0.30), while the high risk (0.75), moderate risk (0.61), and low risk (0.39) maintained consistency over the hospital course. Clinical predictors that discriminated groups included lab measures (hemoglobin, potassium, sodium), vital signs (diastolic blood pressure), and the number of prior hospitalizations. CONCLUSIONS Dynamically predicting readmission and quantifying trends over patients’ hospital stay illuminated differing risk trajectory groups. Identifying risk trajectory patterns and distinguishing predictors may shed new light on indicators of readmission and the isolated effects of the index hospitalization.