Reinforcement Learning for Warfarin Dosing

Overview

This is a clinical study designed to test the hypothesis that a computer model for dosing warfarin is superior to current clinical practice. Subjects will be randomized to two groups based on how warfarin dose is determined and followed for 6 months. The primary outcome is the percent of the time that the INR is maintained in the effective range.

This will be a single-center, open-label, randomized prospective study. Primary outcome will be percent of time within a specific INR range. Subjects will be randomized to to control and treatment groups and stratified 1:1 based on sex. The control group will have warfarin doses adjusted by experts in the treatment of patients with warfarin. The treatment group will have warfarin doses determined using a clinical support tool based on reinforcement learning. Based on simulations of the experimental design with an expectation that the percent of INR values within the target range increase by 20%, 70 subjects per group are required for statistical significance. Based on an attrition rate of 15%, the investigators will enroll 80 subjects per group. Statistical analysis will compare the percent of patients within the target INR range between groups as the primary outcome with number of adverse events between groups as the safetly outcome.

Conditions

Interventions

Eligibility

Outcomes