Marginal Structural Models as a Tool for Standardization

Abstract

In this article, we show the general relation between standardization methods and marginal structural models. Standardization has been recognized as a method to control confounding and to estimate causal parameters of interest. Because standardization requires stratification by confounders, the sparse-data problem will occur when stratified by many confounders and one then might have an unstable estimator. A new class of causal models called marginal structural models has recently been proposed. In marginal structural models, the parameters are consistently estimated by the inverse-probability-of-treatment weighting method. Marginal structural models give a nonparametric standardization using the total group (exposed and unexposed) as the standard. In epidemiologic analysis, it is also important to know the change in the average risk of the exposed (or the unexposed) subgroup produced by exposure, which corresponds to the exposed (or the unexposed) group as the standard. We propose modifications of the weights in the marginal structural models, which give the nonparametric estimation of standardized parameters. With the proposed weights, we can use the marginal structural models as a useful tool for the nonparametric multivariate standardization.

Delayed Internal Fixation of Femoral Shaft Fracture Reduces Mortality Among Patients with Multisystem Trauma

“…In the Appendix, we present an empirical method for assessing the sensitivity of inverse probability of treatment-weighted estimates to the most likely violation of experimental treatment assignment, specifically, that patients who were managed more than twelve hours after admission could not feasibly have been managed earlier. This approach is based on the standardized risk ratio with use of modified inverse probability of treatmentweighted weights 35 , which, in this case, gives an estimate of the effect of delaying surgery for those subjects who were managed early. If subjects managed after twelve hours vary enough from those managed earlier to bias estimates of effect, the standardized risk ratio-with its target population being the t 0 group-should differ from the marginal inverse probability of treatment-weighted estimate.…”

Effect of Massachusetts health insurance reform on mammography use and breast cancer stage at diagnosis

“…We used a standardized mortality ratio propensity score weight for each woman who had a score equal to 1 for Massachusetts women and p/(1 À p) (the propensity odds) for women in California. 19,20 This gives additional weight to California women who most resemble Massachusetts women, so that the weighted distribution of characteristics in the 2 cohorts is well balanced and equal to that of the original Massachusetts cohort. Standardized mortality ratioweighted effects, thus, estimate the likelihood of Abbreviations: SE, standard error.…”

Effectiveness of Rituximab in Patients with Rheumatoid Arthritis: Observational Study from the British Society for Rheumatology Biologics Register

“…The model was adjusted for propensity scores 28 that were calculated using a logistic regression model that included any differences in baseline characteristics between the 2 subgroups of patients, any interaction between baseline variables, and any predictors of the outcome.…”