Analysis of Observational Studies in the Presence of Treatment Selection Bias

Abstract

N THE FACE OF THE FINANCIAL, practical, and ethical challenges inherent in undertaking randomized clinical trials (RCTs), investigators often use observational data to compare the outcomes of different therapies. These comparisons may be biased due to prognostically important baseline differences among patients, often as a result of unobserved treatment selection biases. Unmeasurable clinical and social interactions in the diagnostic-treatment pathway, and physicians' knowledge of unmeasured prognostic variables, may affect treatment decisions and outcomes. Physicians are frequently risk averse in case selection, performing interventions on lower-risk patients despite greater clinical benefit to higher-risk patients. [1][2][3] In some cases, especially when data are collected on detailed clinical risk factors, these differences can be controlled using standard statistical methods. In other cases, when unmeasured patients characteristics affect both the decision to treat and the outcome, these For editorial comment see p 314.

Comparison of Functional Status Improvements Among Patients With Stroke Receiving Postacute Care in Inpatient Rehabilitation vs Skilled Nursing Facilities

“…However, the inability of more analytical techniques to eliminate the differences in the control outcome of all-cause mortality between 30 and 365 days suggests that those approaches did not eliminate selection biases. This pattern is consistent with prior comparative effectiveness studies using observational data 7,8,9 and reinforces the view that such techniques should be avoided in the face of strong selection bias.…”

Survival Outcomes Associated With Cytoreductive Nephrectomy in Patients With Metastatic Clear Cell Renal Cell Carcinoma

“…It is likely that this discrepancy reflects the fact that surgical indication for cytoreductive nephrectomy is primarily driven by factors that are not commonly measured or available in observational data sets. These findings are similar to those identified in a landmark analysis by Stukel et al, who demonstrated that instrumental variable analysis more closely approximated clinical trial outcomes than conventional adjustments for measured confounding in an observational study assessing the benefit of cardiac catheterization among patients with acute myocardial infarction.…”

Hospital-Based Palliative Care Consultation: Effects on Hospital Cost

“…Such biases in observational studies can be substantial and instrumental variable methods are better than standard analyses. 23 Thus, our finding of cost savings lends additional credibility to similar findings in the literature that did not account for unmeasured selection. Moreover, the magnitude of the effect on cost savings from our study is likely to be more accurate.…”