Pruning the Path to Optimal Care: Identifying Systematically Suboptimal Medical Decision-Making with Inverse Reinforcement Learning

TL;DR

This paper introduces a novel IRL-based method to identify suboptimal medical decisions in ICU data by comparing clinician actions to peer consensus, revealing insights into clinical priorities and disparities.

Contribution

It presents a two-stage IRL approach with trajectory pruning to detect suboptimal decisions, advancing understanding of clinical decision-making from observational data.

Findings

Suboptimal actions vary by disease

Removing suboptimal actions impacts demographic groups differently

IRL can uncover clinical priorities from observational data

Abstract

In aims to uncover insights into medical decision-making embedded within observational data from clinical settings, we present a novel application of Inverse Reinforcement Learning (IRL) that identifies suboptimal clinician actions based on the actions of their peers. This approach centers two stages of IRL with an intermediate step to prune trajectories displaying behavior that deviates significantly from the consensus. This enables us to effectively identify clinical priorities and values from ICU data containing both optimal and suboptimal clinician decisions. We observe that the benefits of removing suboptimal actions vary by disease and differentially impact certain demographic groups.