gms | German Medical Science

Introduction: The validity of a hospital-based case-control study rests in part on the assumption that the distribution of exposure in the controls is representative of the study base that gave rise to the cases. The aim of this study is to quantify the bias that arises when controls with hospitalization diagnoses are included that do not represent the exposure prevalence in the study base.

Methods: We assumed a hospital-based case-control with an unbiased odds ratio (OR) of 3.0 for the exposure-disease relation of interest. We assigned several relative weights to the distribution of biasing and valid controls and studied several exposure prevalences of biasing controls within the control group. We restricted the direction of bias for the calculations to upward biases of the exposure prevalence in the control group. To quantify the bias of the OR due to unrepresentative controls, we calculated the differences between the observed and true OR (absolute bias). In addition, we calculated the relative OR bias by dividing this difference by the true OR (relative bias).

Results: The absolute OR bias due to upward biased control prevalences is larger if the true exposure of the study base is low (e.g. 0.10) or high (e.g. 0.80) as compared to a true exposure prevalence of the study base of 0.50. The higher the relative weight of the biased control disease, the larger the bias.

Conclusions: The amount of bias due to unrepresentativeness of hospital controls depends not only on the relative weight of the biasing hospital controls and their exposure prevalence, but also on the underlying true exposure prevalence in the study base.