Predicting outcomes over time in patients with heart failure, left ventricular systolic dysfunction, or both following acute myocardial infarction

Background: Most studies of risk assessment or stratification in patients with myocardial infarction (MI) have been static and fail to account for the evolving nature of clinical events and care processes. We sought to identify predictors of mortality, cardiovascular death or nonfatal MI, and cardiovascular death or nonfatal heart failure (HF) over time in patients with HF, left ventricular systolic dysfunction, or both post‐MI. Methods and Results: Using data from the VALsartan In Acute myocardial iNfarcTion (VALIANT) trial, we developed models to estimate the association between patient characteristics and the likelihood of experiencing an event from the time of a follow‐up visit until the next visit. The intervals are: hospital arrival to discharge or 14 days, whichever occurs first; hospital discharge to 30 days; 30 days to 6 months; and 6 months to 3 years. Models were also developed to predict the entire 3‐year follow‐up period using baseline information. Multivariable Cox proportional hazards modeling was used throughout with Wald chi‐squares as the comparator of strength for each predictor. For the baseline model of overall mortality, the 3 strongest predictors were age (adjusted hazard ratio [HR], 1.35; 95% CI, 1.28–1.42; P < 0.0001), baseline heart rate (adjusted HR, 1.17; 95% CI, 1.14–1.21; P < 0.0001), and creatinine clearance (≤100 mL/min; adjusted HR, 0.86; 95% CI, 0.84–0.89; P < 0.0001). According to the integrated discrimination improvement (IDI) and net reclassification improvement (NRI) indices, the updated model had significant improvement over the model with baseline covariates only in all follow‐up periods and with all outcomes. Conclusions: Patient information assessed closest to the time of the outcome was more valuable in predicting death when compared with information obtained at the time of the index hospitalization. Using updated patient information improves prognosis over using only the information available at the time of the index event