Improving management of Anemia in End Stage Renal Disease using Reinforcement Learning

We present a reinforcement learning approach to elicit individualized dose adjustment policies for patients suffering anemia due to end stage renal disease. Our goal is to achieve stable steady-state anemia management in patients with exhibiting different levels of treatment response. The approach uses Q-learning with parsimonious parametric representation of the state-action value function. We show that this approach achieves stability even in highly responsive patients.