Abstract
Simulation-based approaches to disease progression allow us to make counterfactual predictions about the effects of an untried series of treatment choices. However, building accurate simulators of disease progression is challenging, limiting the utility of these approaches for real world treatment planning. In this work, we present a novel simulation-based reinforcement learning approach that mixes between models and kernel-based approaches to make its forward predictions. On two real world tasks, managing sepsis and treating HIV, we demonstrate that our approach both learns state-of-the-art treatment policies and can make accurate forward predictions about the effects of treatments on unseen patients.
Highlights
Despite progress in machine learning methods for clinical decision support (e.g. [1,2,3]), machine learning algorithms usually operate as uninterpretable black-boxes which clinicians are often hesitant to trust and adopt as tools
We present a quantitative evaluation of the policy and the forward simulation
The choice of time horizon is made on the basis of how frequently an HIV patient visits the hospital for treatment, medical guidelines and drugs available
Summary
Despite progress in machine learning methods for clinical decision support (e.g. [1,2,3]), machine learning algorithms usually operate as uninterpretable black-boxes which clinicians are often hesitant to trust and adopt as tools. [1,2,3]), machine learning algorithms usually operate as uninterpretable black-boxes which clinicians are often hesitant to trust and adopt as tools. Given this context, simulation-based approaches to managing disease progression are appealing because they allow us to make counterfactual predictions about the possible future outcomes associated with different treatment options. An intensivist may see a physiologically implausible blood-pressure trajectory accompanying a treatment recommendation and correctly decide to ignore the recommendation In this way, simulations provide a complementary context than a set of guidelines or recommendations
Sign-in/Register to view highlights & summary 
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call