Code for a reinforcement learning model applied to the management of intravenous fluids and vasopressors in patients with sepsis in intensive care.
Author: Dr Matthieu Komorowski, Imperial College London, 2016-2017 - m.komorowski14@imperial.ac.uk
This repository contains PostgreSQL and Matlab code to:
- define cohorts of patients fulfilling the sepsis-3 definition in two databases: MIMIC-III (https://mimic.physionet.org/) and eICU-RI (not publicly available in full, subset available here: http://eicu-crd.mit.edu/)
- extract the data of interest from both databases
- build a Markov Decision Process model from the MIMIC-III dataset and identify an optimal policy
- test the optimal policy identified on the eICU-RI dataset
- computes the main results and key figures
References: Singer, JAMA 2016 http://jamanetwork.com/journals/jama/fullarticle/2492881