How Consistent are Clinicians? Evaluating the Predictability of Sepsis Disease Progression with Dynamics Models
This work addresses the challenge of optimizing sepsis treatment policies in intensive care, but it is incremental as it focuses on evaluating predictability rather than proposing a new solution.
The paper investigated whether clinician actions in sepsis treatment are predictable enough to improve disease progression models, finding that incorporating action information did not significantly enhance performance, suggesting limited variability in clinician actions.
Reinforcement learning (RL) is a promising approach to generate treatment policies for sepsis patients in intensive care. While retrospective evaluation metrics show decreased mortality when these policies are followed, studies with clinicians suggest their recommendations are often spurious. We propose that these shortcomings may be due to lack of diversity in observed actions and outcomes in the training data, and we construct experiments to investigate the feasibility of predicting sepsis disease severity changes due to clinician actions. Preliminary results suggest incorporating action information does not significantly improve model performance, indicating that clinician actions may not be sufficiently variable to yield measurable effects on disease progression. We discuss the implications of these findings for optimizing sepsis treatment.