Comparison of Influenza Epidemic Trends Based on a Large-scale Claims Database and National Infectious Disease Surveillance in Japan

Clinical Snapshot

Key Findings

• P Value: Not reported

• Effect Size: Rt exceeded 1.0 when cases per sentinel surpassed 0.2–0.3 (specific correlation statistics not reported in abstract). A threshold of 0.25 cases per sentinel was identified as enabling earlier epidemic onset detection.

• Primary Outcome: Epidemic trend curves derived from JMDCdb (diagnosis-based and antiviral prescription-based) showed good concordance with NESID reported case counts across three influenza seasons (2016–2019), with minor discrepancies at epidemic peaks. Rt estimates were consistent across all data sources.

• Nnt Or Sensitivity: The proposed 0.25 cases-per-sentinel threshold enabled epidemic onset detection 4–5 weeks earlier than the current Japanese standard threshold; formal sensitivity, specificity, or predictive values for this threshold are not reported

• Confidence Interval: Not reported

Clinical Application

JMDCdb data access requires institutional agreements with JMDC Inc. and is not freely available; real-time claims data linkage for prospective surveillance would require significant infrastructure investment. The refined sentinel threshold (0.25 cases/sentinel) is immediately implementable within existing NESID infrastructure at no additional cost if validated prospectively. Australia operates distinct but analogous systems: the National Notifiable Diseases Surveillance System (NNDSS), the Australian Sentinel Practices Research Network (ASPREN), and AUSFluWeb (participatory surveillance). The Medicare Benefits Schedule (MBS) and Pharmaceutical Benefits Scheme (PBS) databases represent Australia's equivalent of administrative claims data and have been used for pharmacoepidemiological research. The concept of using PBS antiviral dispensing data (oseltamivir, zanamivir, baloxavir — TGA-approved) as a supplementary influenza surveillance signal is directly translatable and warrants investigation by the Australian Centre for Disease Control (ACDC, established 2023). RACGP sentinel networks could similarly explore refined reporting thresholds analogous to the 0.25 cases/sentinel finding. However, direct threshold values would require Australian-specific derivation and prospective validation before adoption. Public health epidemiologists and infectious disease surveillance officers in Japan; potentially applicable to other countries with comparable administrative claims infrastructure and sentinel surveillance networks

Abstract

BACKGROUND: Seasonal influenza is a recurrent respiratory infection, and timely detection is essential for public health. In Japan, surveillance is conducted through sentinel medical institutions under the National Epidemiological Surveillance of Infectious Diseases (NESID). Recently, access to large claims databases, such as the JMDC claims database (JMDCdb), has increased. While both are sample-based systems, JMDCdb covers a much larger population. We aimed to assess consistency between these sources in estimating influenza cases and the effective reproduction number (Rt) and to explore their utility in epidemic analysis. METHODS: We analyzed data from week 36 of 2016 to week 35 of 2019. Influenza cases were estimated from NESID (reported cases and cases per sentinel) and JMDCdb (cases with influenza-related diagnoses and antiviral prescriptions). Daily infection counts were derived to estimate Rt. RESULTS: Although minor differences appeared at epidemic peaks, estimates from NESID reports aligned well with JMDCdb. Estimates based on cases per sentinel were lower. Rt values were consistent across data sources. Rt exceeded 1.0 when cases per sentinel surpassed 0.2-0.3. Using a threshold of 0.25 cases per sentinel enabled detection of epidemic onset 4-5 weeks earlier than current standards. CONCLUSION: Claims data, such as those from JMDCdb, may be useful for retrospective examination of influenza trends. Moreover, a detailed analysis of the number of cases reported per sentinel suggested the potential to propose threshold values that enable earlier prediction of epidemics than conventional criteria.

References

1. 1.Yamauchi, T., Den, H., Takeuchi, S., Saito, M., Uchida, M., & Kokaze, A. (2026). Comparison of influenza epidemic trends based on a large-scale claims database and national infectious disease surveillance in Japan. Journal of Epidemiology. https://doi.org/10.1007/s40506-016-0085-5