|
Abstract:
|
When making treatment decisions, clinicians and patients desire accurate predictions of mean residual life (MRL) that leverage all available patient information, including longitudinal biomarker data. Biomarkers are variables that reflect disease progression and are often measured repeatedly on patients in the clinical setting. Dynamic prediction methods leverage accruing biomarker measurements to improve performance, providing updated predictions as new measurements become available. We introduce two methods for dynamic prediction of MRL using longitudinal biomarkers. In both methods, we use long short-term memory networks to construct encoded representations of the biomarker trajectories, referred to as "context vectors." We then dynamically predict MRL, treating the context vectors as covariates. In our first method, we conduct prediction via a transformed MRL model. In our second method, we use a feed-forward neural network. We demonstrate the improved performance of both proposed methods relative to competing methods via a data application, where we dynamically predict the restricted MRL of septic patients in the intensive care unit from electronic medical record data.
|
