Standardization of case definition and development of early-warning model for acute respiratory infection syndromes based on Yinzhou Regional Health Information Platform.

BACKGROUND

Acute respiratory infection syndromes (ARIs) pose major public health challenges due to their high infectivity, rapid transmission, and the lack of standardized definitions balancing sensitivity and specificity in current surveillance systems.

OBJECTIVE

Using data from Yinzhou Regional Health Information Platform (YRHIP), we refined ARIs definition, improved classical epidemic criteria and designed a comprehensive graded early-warning model to enhance early response capabilities.

METHODS

We optimized ARIs definition based on laboratory-confirmed cases and evaluating screening performance with clinical diagnoses. Anomaly detection methods, including historical limits method (HLM), moving percentile method (MPM), cumulative sum control chart (CUSUM), and exponentially weighted moving average (EWMA), were employed to develop a graded early-warning model. Syndrome selection and parameter tuning were guided by Youden's index, agreement rate and F1-score.

RESULTS

The refined ARIs definition includes: Acute-phase fever with at least one typical respiratory symptoms; or acute-phase fever with at least two atypical respiratory symptoms; or at least one typical respiratory symptoms combined with at least two atypical respiratory symptoms. Furthermore, we demonstrate that ARIs outperform ILIs definition in early screening due to their broader symptom scope. By leveraging multidimensional time series data, we developed a robust epidemic criteria framework for early-warning models. The optimal early-warning parameters included configurations of HLM ( = 0.8), MPM (85th percentile), CUSUM( = 0.7,  = 5), and EWMA ( = 3,  = 0.05). The graded early-warning system revealed: Red early-warnings (all four models triggered) had the highest specificity; Orange early-warnings (at least three models triggered) demonstrated the best overall performance; Amber early-warnings (at least two models triggered) captured subtle trends; Green early-warnings (at least one model triggered) provided the highest sensitivity.

CONCLUSION

This study establishes an optimized, multi-model-based framework for ARIs early-warning that balances sensitivity and specificity to strengthen public health management against diverse pathogens.