Case ascertainment in stroke studies: the risk of selection bias

OBJECTIVES: The purpose was to compare the completeness of case ascertainment in two stroke registers, one local population-based, the other a national quality register (Riks-Stroke), and to examine if patient characteristics could be affected by selection bias. MATERIAL AND METHODS: By the way of linking and matching computer files, the completeness of case ascertainment was evaluated. RESULTS: In the local stroke incidence study 377 patients were included. Of them, 63% were reported to the hospital-based national quality register. The case-fatality was lower in the national register. A larger proportion of the patients in the national register appeared to have been treated in a stroke unit and undergone rehabilitation, and computerized tomography seemed to have been performed in a larger proportion. CONCLUSIONS: Because of selection bias, outcome data get skewed when case ascertainment does not embrace all stroke cases. A community-based stroke register is the golden standard when measuring stroke incidence.