de Wit, Kerstin2014-06-172014-06-1720142014http://hdl.handle.net/10393/31190http://dx.doi.org/10.20381/ruor-3786Background Bleeding can be an adverse side effect from hospital treatment. The aim was to develop an electronic identification method for patients who are bleeding within The Ottawa Hospital. Methods A retrospective exploratory cohort (N=1000) was used to identify potential candidate markers for bleeding. Electronic data were extracted to evaluate candidate identifiers. Data which were associated with bleeding events were assessed in a model derivation cohort (N=700). Multivariate analysis was used to establish the best model for identifying all bleeding events and in-hospital bleeding events. Results Overall 38% of the exploratory cohort had bleeding. In the model derivation set 29% had bleeding. The model predicting all bleeding included number of transfusions, admitting specialty, re-operation and endoscopy (C-statistic 0.82, 95%CI 0.79-0.86). The model predicting in-hospital bleeding included number of transfusions, admitting specialty and re-operation (C-statistic 0.78, 95% CI 0.73-0.84). Conclusion We have developed two models for identifying hospital bleeding events from The Ottawa Hospital electronic medical records. These should be validated prospectively on the hospital-wide population.enbleedingelectronic medical recordselectronic identificationThesis