Search

Summary For a continuous treatment, the generalised propensity score (GPS) is defined as the conditional density of the treatment, given covariates. GPS adjustment may be implemented by including it as a covariate in an outcome regression. Here, the unbiased estimation of the dose–response function assumes correct specification of both the GPS and the outcome‐treatment relationship. This paper introduces a machine learning method, the ‘Super Learner’, to address model selection in this context. In the two‐stage estimation approach proposed, the Super Learner selects a GPS and then a dose–response function conditional on the GPS, as the convex combination of candidate prediction algorithms. We compare this approach with parametric implementations of the GPS and to regression methods. We contrast the methods in the Risk Adjustment in Neurocritical care cohort study, in which we estimate the marginal effects of increasing transfer time from emergency departments to specialised neuroscience centres, for patients with acute traumatic brain injury. With parametric models for the outcome, we find that dose–response curves differ according to choice of specification. With the Super Learner approach to both regression and the GPS, we find that transfer time does not have a statistically significant marginal effect on the outcomes. © 2015 The Authors. Health Economics Published by John Wiley & Sons Ltd.

Many cost-effectiveness analyses (CEAs) use data from observational studies. Statistical methods can only address selection bias if they make plausible assumptions. No quality assessment tool is available for appraising CEAs that use observational studies. We developed a new checklist to assess statistical methods for addressing selection bias in CEAs that use observational data. The checklist criteria were informed by a conceptual review and applied in a systematic review of economic evaluations. Criteria included whether the study assessed the 'no unobserved confounding' assumption, overlap of baseline covariates between the treatment groups and the specification of the regression models. The checklist also considered structural uncertainty from the choice of statistical approach. We found 81 studies that met the inclusion criteria: studies tended to use regression (51%), matching on individual covariates (25%) or matching on the propensity score (22%). Most studies (77%) did not assess the 'no observed confounding' assumption, and few studies (16%) fully considered structural uncertainty from the choice of statistical approach. We conclude that published CEAs do not assess the main assumptions behind statistical methods for addressing selection bias. This checklist can raise awareness about the assumptions behind statistical methods for addressing selection bias and can complement existing method guidelines for CEAs.