Validity of initial cancer diagnoses in the Diagnosis Procedure Combination data in Japan

Background: In Japan, the Diagnosis Procedure Combination (DPC) data have been used as a nationwide administrative hospital discharge database for clinical studies. However, few studies have evaluated the validity of recorded diagnoses of cancer in the database. Methods: We compared the DPC data with hospital-based cancer registries in Osaka Prefecture, Japan to assess the validity of the recorded cancer diagnoses in the DPC data. Fifteen types of cancer were included in the analysis. Cancer stage with tumor-node-metastasis (TNM) classification was assessed for eight cancer types with >400 patients. We evaluated concordance and positive predictive value of cancer diagnosis, and concordance of cancer stage between the DPC data and the hospital-based cancer registry. Results: In total, we identified 29,180 eligible patients. The five types of cancer with the highest number of patients were as follows: 6,765 (23.2 %) colorectal, 6,476 (22.2 %) stomach, 4,862 (16.7 %) breast, 4,445 (15.2 %) lung, and 2,257 (7.7 %) liver. Concordance of diagnosis ranged from 63.9 %-99.5 %, and twelve of the fifteen types of cancers had concordance of over 90 %. Positive predictive values of diagnosis ranged from 86.8 %-100 %. Regarding cancer stage, the overall degree of concordance was 67.2 % in all patients and the concordance was over 70 % in four types of cancers. Conclusions: The DPC data had high validity of cancer diagnosis. However, the potential impact of the misclassifications and low concordance in cancer stage among specific type of cancers in the DPC data should be considered.

Automated summary

This study compared the validity of cancer diagnoses in Japan's nationwide administrative hospital discharge database (DPC data) with hospital-based cancer registries in Osaka Prefecture. Results showed high concordance and positive predictive values of cancer diagnosis in the DPC data, but some specific types of cancer had low concordance in cancer stage, which may impact the results.