Hurley, Nathan C.

Abstract Acute myocardial infarction complicated by cardiogenic shock (AMI-CS) is a heterogeneous clinical syndrome associated with substantial morbidity and mortality. We developed a machine learning-based mortality model to identify features and patient subgroups associated with the largest change in mortality risk when evaluating treatment with Impella devices or intra-aortic Balloon Pump (IABP). Our cohort comprised 369 sites and 15,796 patient visits to the cardiac catheterization laboratory from the National Cardiovascular Data Registry. The estimated population mean excess mortality effect of treatment with Impella devices vs IABP was 10.4 ± 0.8%. However, we identified clinical subgroups of 282 patients for whom a decreased risk of mortality was associated with use of Impella as compared with IABP. Those patients were on average younger, presented with higher systolic blood pressure, higher rate of salvage percutaneous coronary intervention, higher initial creatinine, and lower hemoglobin. While Impella devices were associated with higher mortality risk overall, certain clinical profiles were associated with lower risk, illustrating heterogeneity of treatment effects.

Observational medical data present unique opportunities for analysis of medical outcomes and treatment decision making. However, because these datasets do not contain the strict pairing of randomized control trials, matching techniques are to draw comparisons among patients. A key limitation to such techniques is verification that the variables used to model treatment decision making are also relevant in identifying the risk of major adverse events. This article explores a deep mixture of experts approach to jointly learn how to match patients and model the risk of major adverse events in patients. Although trained with information regarding treatment and outcomes, after training, the proposed model is decomposable into a network that clusters patients into phenotypes from information available before treatment. This model is validated on a dataset of patients with acute myocardial infarction complicated by cardiogenic shock. The mixture of experts approach can predict the outcome of mortality with an area under the receiver operating characteristic curve of 0.85 ± 0.01 while jointly discovering five potential phenotypes of interest. The technique and interpretation allow for identifying clinically relevant phenotypes that may be used both for outcomes modeling as well as potentially evaluating individualized treatment effects.