Models as the base for counter-factual evidence
The premis for Model Estimated Control is that a model can be used to estimate a patient's outcome and that this estimate can be compared against their observed outcome
Model Estimated Controls - Methods
Concept
Average Treatment Effects obtained by averaging across individual effect
To estimate the average treatment effect over a group of patients, we can average over all individual effects.
Estimation
Two procedures exist for estimating treatment effects using MEC - a simulation and a likelihood approach
Simulation
Use the model to directly simulate 'synthetic' patients. For each observed patient we can then use basic analytical methods to compare the observed and the expected outcomes
Likelihood based approach
Define a likelihood where a term is added to measure the direct difference between the counter factual model and the new cohort of observations
Further Details
MECs for Binary/Continuous Outcomes
Understand how Model Estimated Controls work with MEC based on Generalised Linear Models
MECs for Survival Outcomes
Understand how Model Estimated Controls work with MEC based on Parametric Survival Models
Bayesian Estimation of MECs
Get more details on the Bayesian estimation procedures along with MCMC algorithms