journal of biomedical informatics
All submissions of the EM system will be redirected to Online Manuscript Submission System. Authors are requested to submit articles directly to Online Manuscript Submission System of respective journal.

Automatic Selection of Diagnosis Procedure Combination Codes Based on Partial Treatment Data Relative to the Number of Hospitalization Days

Author(s): Kazuya Okamoto*, Toshio Uchiyama, Tadamasa Takemura, Naoto Kume, Tomohiro Kuroda and Hiroyuki Yoshihara

Objectives: The authors developed and evaluated a method of selecting accurate diagnosis procedure combination (DPC) codes based on standardized treatment information relative to the number of hospitalization days.

Methods: The authors used machine learning methods to generate DPC codes based on treatment data. The machine learning methods utilized were the Naïve Bayes method, the SVM method, and a combined method of the two methods. We prepared DPC code data and standardized treatment data corresponding to cases occurring in fiscal year 2008 at Kyoto University Hospital. To produce classification models, machine learning methods require a moderate amount of data corresponding to each DPC code; accordingly, we selected 166 DPC codes that were each related to at least 20 cases. The number of cases with these DPC codes was 10,123.

Results: DPC code selection was attempted using the Naïve Bayes method, the SVM method, and the combined method of the two; of these, the combined method yielded the best results, producing accurate DPC codes in 73.8% of cases. In addition, we were able to improve the precision of DPC code selection to 76.5% by utilizing partial treatment data gathered up to the 11th day of each hospitalization.

Conclusion: The present study confirmed the feasibility of automatic DPC code selection through machine learning methods based on treatment information. Our future work will include the construction of a system to select DPC codes automatically and the evaluation of this system to determine whether it can reduce doctors’ workloads.