A full-text information retrieval system for an epidemiological registry

Case finding for epidemiologic registries still relies mainly on a manual process. In this paper, we show that retrieval information tools could be a complementary way to identify cases for a pediatric malformation registry. We developed a full-text and metadata search engine plugged to a clinical documents repository and used it to identify Epi/Hypospadia and Spina bifida cases. The queries were enriched with Snomed terminologies. We compared the performances of this prototype versus the hospital DRG database (classical method). The best precisions of prototype for identification of Spina bifida and Epi/Hypospadia were respectively 73% and 87%. The prototype overlap with the DRG system was 83% and 97%. Compared to DRG, 13 new not referenced and 2 miscoded cases were detected. This free full- text retrieval system prototype allows efficiently reusing clinical documents for case finding for an epidemiologic purpose.