Applying Text Mining and Natural Language Processing to Electronic Medical Records for Extracting and Transforming Texts into Structured Data

The recording of patients' data in electronic patient records (EPRs) by healthcare providers is usually performed in free text fields, allowing different ways of describing that type of information (e.g., abbreviation, terminology, etc.). In scenarios like that, retrieving data from such source (text) by using SQL (Structured Query Language) queries becomes an unfeasible issue. Based on this fact, we present in this paper a tool for extracting comprehensible and standardized patients' data from unstructured data which applies Text Mining and Natural Language Processing techniques. Our main goal is to carry out an automatic process of extracting, clearing and structuring data obtained from EPRs belonging to pregnant patients from the Januario Cicco maternity hospital located in Natal - Brazil. 3,000 EPRs written in Portuguese from 2016 e 2020 were used in our comparison analysis between data manually retrieved by health professionals (e.g., doctors and nurses) and data retrieved by our tool. Moreover, we applied the Kruskal-Wallis statistical test in order to statically evaluate the obtained results between manual and automatic processes. Finally, the statistical results have showed that there was no statistical difference between the retrieval processes. In this sense, the final results were considerably promising.