Using Text Mining to Validate Diagnoses of Acute Myocardial Infarction

We applied text mining to a database of discharge letters of patients with acute myocardial infarction for a quality of care-related task: the automatic validation of acute myocardial infarction diagnoses. The system should evaluate if the information contained in the discharge letters was consistent, by medical standards, with the letters’ coded diagnoses of acute myocardial infarction. The system was composed of a text mining tool (GATE) and a set of linguistic resources which were specifically developed from a training set of letters. It was validated on a test set of letters manually annotated by cardiologists and results were satisfactory. Further analyses can be made on the efficiency of the development of the system and on its ongoing effectiveness.