Article ID: 27020
Background:
Japan’s National Database of Health Insurance Claims (NDB) is used as a data source for selected national cancer control indicators. However, national policy does not explicitly define “patients with cancer” for claims-based measurement. We quantified variations in NDB-based cancer control indicator values across diagnosis-code restrictions.
Methods:
We conducted a repeated cross-sectional study using NDB data from fiscal years (FY) 2020–2022. We assessed 11 indicators using an unrestricted comparator, two all-cancer definitions requiring either any confirmed cancer diagnosis or a principal cancer diagnosis, and a six-site sensitivity analysis. Procedure counts were compared, and retained proportions were calculated relative to the unrestricted algorithm.
Results:
The algorithm requiring any confirmed cancer diagnosis retained 99.2% and 99.1% of procedure counts for cancer-specific indicators, as expected because a cancer diagnosis was required for reimbursement, and 95.1%–97.5% for cancer-predominant indicators. The principal-diagnosis definition and six-site sensitivity analysis retained 71.6%–74.6% and 61.6%–64.5%, respectively, for cancer-specific indicators. For indicators based on procedure codes not specific to cancer care, the algorithm requiring any confirmed cancer diagnosis retained 21.1%–88.4% of unrestricted procedure counts, representing the proportions that met the cancer-diagnosis restriction.
Conclusions:
NDB-based cancer control indicator values varied substantially according to the case-identification algorithm. Among the approaches examined, requiring any confirmed cancer diagnosis was selected post hoc as a reasonable default for national monitoring of these 11 indicators because it avoids reliance on principal-diagnosis coding practices and remains applicable across cancer sites. This pragmatic choice should not be interpreted as evidence of diagnostic accuracy.