Mining discharge letters for diagnoses validation and quality assessment

We present two projects where text mining techniques are applied to free text documents written by clinicians. In the first, text mining is applied to discharge letters related to patients with diag-noses of acute myocardial infarction (by ICD9CM coding). The aim is extracting information on diagnoses to validate them and to integrate administrative databases. In the second, text mining is applied to discharge letters related to patients that received a diagnosis of heart failure (by ICD9CM coding). The aim is assessing the presence of follow-up instructions of doctors to patients, as an aspect of information continuity and of the continuity and quality of care. Results show that text mining is a promising tool both for diagnoses validation and quality of care as-sessment.