Extracting clinical information from electronic medical records

Abstract

As the adoption of Electronic Medical Records (EMRs) rises in the healthcare institutions, these resources are each day more important because of the clinical data they contain about patients. However, the unstructured textual data in the form of narrative present in those records, makes it hard to extract and structure useful clinical information. This unstructured text limits the potential of the EMRs, because the clinical data these records contain, can be used to perform important operations inside healthcare institutions such as searching, summarization, decision support and statistical analysis, as well as be used to support management decisions or serve for research. These operations can only be done if the clinical data from the narratives is properly extracted and structured. Usually this extraction is made manually by healthcare practitioners, what is not efficient and is error-prone. The present work uses Natural Language Processing (NLP) and Information Extraction(IE) techniques in order to develop a pipeline system that can extract clinical information directly from unstructured texts present in Portuguese EMRs, in an automated way, in order to help EMRs to fulfil their potential.