Selection bias modeling using observed data augmented with imputed record-level probabilities

Abstract

PURPOSE Selection bias is a form of systematic error that can be severe in compromised study designs such as case-control studies with inappropriate selection mechanisms or follow-up studies that suffer from extensive attrition. External adjustment for selection bias is commonly undertaken when such bias is suspected, but the methods used can be overly simplistic, if not unrealistic, and fail to allow for simultaneous adjustment of associations of the exposure and covariates with the outcome, when of interest. Internal adjustment for selection bias via inverse-probability-weighting allows bias parameters to vary with levels of covariates but has only been formalized for longitudinal studies with covariate data on patients up until loss-to-follow-up. METHODS We demonstrate the use of inverse-probability-weighting and externally obtained bias parameters to perform internal adjustment of selection bias in studies lacking covariate data on unobserved participants. RESULTS The ‘true’ or selection-adjusted odds ratio for the association between exposure and outcome was successfully obtained by analyzing only data on those in the selected stratum (i.e. responders) weighted by the inverse probability of their being selected as function of their observed covariate data. CONCLUSIONS This internal adjustment technique using user-supplied bias parameters and inverse-probability-weighting for selection bias can be applied to any type of observational study.

Physical activity and 22‐year all‐cause and coronary heart disease mortality

“…This study provides new evidence that elevated OPA in terms of relative aerobic workload (%VO 2 max or %VO 2 Res) predicts increased CHD and all‐cause mortality in middle‐aged working men, even at levels of relative aerobic workload that are below the recommended maximum. The findings further support the notion that inconsistencies in the literature on the cardiovascular health effects of OPA may be due to alternative choices of analytic strategies, exposure misclassification, health‐based selection bias or healthy worker effect, uncontrolled confounding, and complex interactions between cardiovascular health status, cardiorespiratory fitness, and physical job demands. Worksite primary prevention of CVD and premature death requires approaches that take the individual worker CVD health status, aerobic capacity, and the energy demands of the job into account.…”

Variety of gambling activities from adolescence to age 30 and association with gambling problems: a 15‐year longitudinal study of a general population sample

“…The trajectories from the selected best fitting model were compared in terms of the type of gambling activities, the frequency of gambling, and gambling problems. Moreover, the association between trajectories and problem gambling at age 30 [52][53] was applied to the GLIMMIX procedure examining the association between trajectories and gambling problems at age 30.…”

Weight gain during incarceration: systematic review and meta‐analysis

“…Subsequently, adjustment or matching on important factors potentially associated with both weight change and imprisonment could be done, as was done by Baldwin et al (24), who found that weekly weight gain in their prison population was higher than race and age-matched gain in the general population. Nonetheless, future studies should be subjected to rigorous design and analysis to minimize and assess sensitivity to bias due to unmeasured confounders, measurement error and selective nonresponse (65)(66)(67)(68). The presence of publication bias suggests that many published studies focusing on weight change in prison tend to show a weight gain rather than a normal distribution of weight changes during incarceration.…”