Medical coding is a challenging job, performed by specialized coders and increasingly demanding given the current regulatory environment. Currently, and particularly in Portuguese hospitals, a great deal of human time and effort is expended in assigning codes to diseases and other health-related conditions described in medical records.
In order to assign codes to a medical record, the coder must read the content of each clinical document and select from thousands of codes the smallest set appropriate to a given situation. Since these codes determine mainly reimbursement, it is important to accomplish this task as accurately and as fast as possible.
This thesis presents a coding system capable of automatically process the content of the clinical reports, extract the clinical terms referred in the free-text clinical documents and map between these terms and the corresponding International Classification of Diseases 9th revision with clinical modification (ICD-9-CM) codes.
The developed coding system contains six main components: Document Pre-Processing; Document Reader; Natural
Language Processing; Name Entity Recognition; ICD-9-CM Module; Assignment Code Result. This system is developed on top off the Unstructured Information Management Architecture (UIMA).
For the evaluation of the system, 22 clinical notes were created by two physicians specialized in cardiology. The results of the system indicate a precision of 96.81% on the extraction of clinical terms and a precision of 90.43% in the assignment of unique identifier to clinical term. The assignment of ICD-9-CM code was evaluated with a precision of 86.17%.
Author: Carla Abreu
Type: MSc thesis
Partner: Faculdade de Engenharia da Universidade do Porto