Confounding and Biasing Variables in Observational and Intervention Studies in Rehabilitation

Abstract

To prove and determine effects of rehabilitation treatments, appropriate study designs have to be applied which allow inferring that differences in the outcome variables (e .g. health state) are causally determined by rehabilitation treatments. In observational studies confounding variables may affect or distort the association between treatment and outcome variables. Preference for a treatment, motivation for treatment or level of impairment before rehabilitation are typical confounders in rehabilitational intervention studies. These potential confounders are especially important if patients chose treatment voluntarily or the treatment has been suggested because of specific impairment characteristics (external program effects). Appropriately controlled intervention studies (especially randomized control trials, RCTs) allow controlling for such biasing effects. But also for appropriate controlled intervention studies, biasing effects may arise due to the implementation of the study design (e. g. selection of participants, applying assessment instruments, communication of participants of different study groups; design effects, treatment contaminations) and may deteriorate the unambiguous interpretability of empirical results. It is shown which confounding and distortional effects have to be regarded that may affect the interpretability of associations. Recommendations for planning, conducting and analyzing empirical studies are given to avoid such biasing effects in the best possible way.