Evaluating Flexible Modeling of Continuous Covariates in Inverse Weighted Estimators.

Evaluating Flexible Modeling of Continuous Covariates in Inverse Weighted Estimators. Am J Epidemiol. 2019 Jan 10;: Authors: Kyle RP, Moodie EEM, Klein MB, Abrahamowicz M Abstract Correct specification of the exposure model is essential for unbiased estimation in marginal structural models with inverse-probability-of-treatment weights. However, although flexible modeling is commonplace when estimating effects of continuous covariates in outcome models, its use is less frequent in estimation of inverse probability weights. Using simulations, we assess the accuracy of the treatment effect estimates and covariate balance, obtained with different exposure model specifications when the true relationship between a continuous possibly time-varying covariate Lt and the logit of the probability of exposure is non-linear. Specifically, we compare four approaches to modeling the effect of Lt when estimating inverse probability weights: a linear function, the covariate-balancing propensity score, and two easy-to-implement flexible methods that relax the assumption of linearity: cubic regression splines, and fractional polynomials. In two empirical studies, we compare linear versus flexible exposure models to estimate the effect of sustained virologic response to hepatitis C virus treatment on liver fibrosis progression. Our simulation results demonstrate that ignoring important non-linear relationships when fitting the exposure model may provide...
Source: Am J Epidemiol - Category: Epidemiology Authors: Tags: Am J Epidemiol Source Type: research