Bayesian causal inference models for healthcare treatment using real world patient data
AssigneeChildren's Hospital Medical Center
InventorBin Huang
Life of this patent
Estimating the term…
Abstract
Computer implemented methods, systems, and computer readable medium are provided for performing causal inference analyses to determine the more effective treatment among alternative treatments in the healthcare setting using real world observational data. Both binary treatment and adaptive treatment strategies are considered in the analysis. The methods comprise generating a Bayesian marginal structural model and performing a single step of Bayesian regression that incorporates matching, weighting, and estimation processes and in which the matching process is performed using a Guassian process (“GP”) prior covariance function.
Claims36
A computer implemented Bayesian nonparametric causal inference method for determining the more effective treatment among two alternative treatments administered to a plurality of patients, the method comprising generating a marginal structural model and performing a single step of Bayesian regression on patient data of the plurality of patients, wherein the method incorporates matching, weighting, and estimation processes into the single regression step and wherein the matching process is performed using a Guassian process (“GP”) prior covariance function.
A computer implemented method for generating a marginal structural model to determine a treatment effect based on patient data, the method comprising
- receiving patient data of a plurality of patients, each patient in the plurality having been diagnosed with the same disease or condition and administered an intervention treatment or a comparator treatment for treating the disease or condition, the patient data comprising a set of observed covariates for each patient including demographic data, disease characteristics, and biological data;
- specifying a set of baseline covariates from the set of observed covariates;
- generating a marginal structural model in which the potential outcomes are modeled nonparametrically using a Guassian process (“GP”) prior covariance function to generate a covariance matrix;
- fitting the model using Bayesian Markov chain Monte Carlo (MCMC) sampling;
- calculating matching weights based on the GP covariance function;
- and
- estimating the treatment effect and potential outcomes using the matching weights.
A system for generating a marginal structural model to determine a treatment effect based on patient data, the system comprising
- a data receiving module configured to receive patient data from an electronic medical records system, the data comprising demographic data, disease characteristics, and biological data for a plurality of patients consisting of at least 200 or at least 300 patients, each patient in the plurality having been diagnosed with the same disease or condition and administered an intervention treatment or a comparator treatment for treating the disease or condition;
- a processor-implemented module configured to generate a marginal structural model by initializing a treatment effect parameter and generating a GP prior variance-covariance matrix;
- fitting the model for a continuous bounded or unbounded outcome using Bayesian Markov chain Monte Carlo (MCMC) sampling;
- calculating matching weights based on the GP prior variance-covariance matrix;
- and estimating an averaged treatment effect.
Dependencies are inferred from the claim text.