Verification of automatic DPC coding by machine learning using medical articles

Abstract

Many acute care hospitals determine the medical cost of hospitalization by selecting the injury or disease for which the most medical resources were invested during the hospitalization period and determining the diagnosis group classification code assigned to each injury or disease. The series of actions up to the determination of the codes is called DPC coding. In order to reduce the workload of this coding process, we used machine learning to predict coding based on medical articles, and verified the accuracy. Naive Bayes and Transformer were used for model construction. Experimental results showed the feasibility of performing partial automatic DPC coding.