Using data from registries like the TraumaRegister DGU((R)) for effectiveness evaluations

In some situations like a rare disease or an emergency intervention, a randomized controlled trial is rather difficult or even impossible to conduct. In these cases, the use of observational studies or registries for the assessment of effectiveness is discussed. Results from registries, however, are subject to the same methodological limitations as any non-randomized comparison. Specifically, the comparability of selected patient groups is not ensured. Using three examples from the TraumaRegister DGU, a nationwide registry for severely injured patients in Germany, the advantages as well as the risks of outcome evaluation with registry data will be discussed: 1) whole-body computed tomography in the early in-hospital phase, 2) pre-hospital volume resuscitation, and 3) surgical treatment according to the principle of damage control. The most important prerequisite for any successful outcome evaluation using registry data is the documentation and availability of so-called confounding variables (prognostic factors), the absence of an accepted standard treatment, i.e., both interventions are routinely applied in similar situations, and, finally, the use of sophisticated statistical methods for multivariate adjustment of results.