Estimating treatment effects with partially observed covariates using outcome regression with missing indicators

Abstract

Missing data is a common issue in research using observational studies to investigate the effect of treatments on health outcomes. When missingness occurs only in the covariates, a simple approach is to use missing indicators to handle the partially observed covariates. The missing indicator approach has been criticized for giving biased results in outcome regression. However, recent papers have suggested that the missing indicator approach can provide unbiased results in propensity score analysis under certain assumptions. We consider assumptions under which the missing indicator approach can provide valid inferences, namely, (1) no unmeasured confounding within missingness patterns; either (2a) covariate values of patients with missing data were conditionally independent of treatment or (2b) these values were conditionally independent of outcome; and (3) the outcome model is correctly specified: specifically, the true outcome model does not include interactions between missing indicators and fully observed covariates. We prove that, under the assumptions above, the missing indicator approach with outcome regression can provide unbiased estimates of the average treatment effect. We use a simulation study to investigate the extent of bias in estimates of the treatment effect when the assumptions are violated and we illustrate our findings using data from electronic health records. In conclusion, the missing indicator approach can provide valid inferences for outcome regression, but the plausibility of its assumptions must first be considered carefully.

Practice Patterns and Outcomes Associated With Anticoagulation Use Following Sepsis Hospitalizations With New-Onset Atrial Fibrillation

“…31,32 The missingness indicator approach was used to address missing baseline and time-dependent covariate values. 30,33,34 Longitudinal data were updated daily for 365 days after hospital discharge and structured to respect temporal ordering between covariates, exposure, and outcome using the LtAtStructuR R package. 35 Multiple sensitivity analyses were conducted to assess the impact of specific assumptions on the results.…”

Early initiation of prophylactic anticoagulation for prevention of coronavirus disease 2019 mortality in patients admitted to hospital in the United States: cohort study

“…We included a missing category for covariates with missing data given the variables with most missingness do not drive the decision to administer prophylactic anticoagulation, but rather are markers for general health severity. Under the additional assumption that associations between fully observed covariates and receipt of prophylactic anticoagulation did not differ across missingness patterns, this approach produces unbiased estimates 2021. Each patient was weighted by the inverse probability of receiving the exposure of interest, with the goal of balancing observable characteristics, including missingness patterns,22 between treatment groups.…”

STEM Pathways of Rural and Small-Town Students: Opportunities to Learn, Aspirations, Preparation, and College Enrollment

“…In all block-entry hierarchical regression models, missing indicators were created and included in the models to account for missing data (Blake et al, 2020;Chiou et al, 2019). 5 Joint hypotheses tests (F tests) were conducted to determine whether a given block of predictors has explanatory power in predicting a specific outcome, net of baseline or restricted models (Wooldridge, 2019).…”