Getting the best of both worlds: an approach for maximising internal and external validity in cost-effectiveness studies

Abstract: Randomised controlled trials (RCTs) tend to include atypical patients and centres, and the accompanying cost-effectiveness analyses (CEA) may be subject to sample selection bias. To reduce this bias, observational data can be used to reweight the trial-based estimates. I will present an approach assessing the assumptions that underlie any such reweighting strategy.

In the paper I decompose sample selection bias into observable or unobservable differences between the RCT and the target population. I consider alternative ways of reweighting the RCT estimates, to the population’s characteristics. The first estimation strategy, reweights
according to Inverse Probability of Treatment Weighting (IPW), where ‘treatment’ is inclusion in the RCT. The second strategy uses maximum entropy (MaxEnt) reweighting to match the RCT patients to those treated in the population. Either approach makes the  identifying assumption that selection into the RCT is conditional on observable characteristics. I assess this assumption with novel placebo tests that contrast reweighted endpoints following treatment in the RCT, versus in the target population. The placebo test is only passed if the identifying assumption holds, and there is sufficient power to detect a difference in endpoints across settings.

I consider this approach in a CEA of Pulmonary Artery Catheterisation (PAC) by using patient-level data from a large observational study to reweight the RCT estimates.