Background
Although health insurance claims data can address questions that clinical trials cannot answer, the uncertainty of disease names and the absence of stage information hinder their use in gastric cancer (GC) research. This study aimed to develop and validate a claims-based algorithm to identify and determine the progression phases of incident GC cases in Japan.
Methods
The gold standard for validation in this retrospective observational study was medical records of patients with incident GC who underwent specific treatments, defined by the claim codes associated with GC treatment. The algorithm was developed and refined using a cohort from two large tertiary care medical centers (April–September 2017 and April–September 2019) and subsequently validated using two independent cohorts: one from different periods (October 2017–March 2019 and October 2019–March 2021) and the other from a different institution (a community hospital). The algorithm identified incident cases based on a combination of the International Classification of Diseases, 10th Revision diagnosis codes for GC (C160-169), and claim codes for specific treatments, classifying them into endoscopic, surgical, and palliative groups. Positive predictive value (PPV), sensitivity of incident case identification, and diagnostic accuracy of progression phase determination were evaluated.
Results
The developed algorithm achieved PPVs of 90.0% (1119/1244) and 95.9% (94/98), sensitivities of 98.0% (1119/1142) and 98.9% (94/95) for incident case identification, with diagnostic accuracies of 94.1% (1053/1119) and 93.6% (88/94) for progression phase determination in the two validation cohorts, respectively.
Conclusions
This validated claims-based algorithm could advance real-world GC research and assist in decision-making regarding GC treatment.