LAUSR

This study evaluated the use of chief complaint data from emergency departments (EDs) to detect the increment of influenza cases identified from the nationwide medical service usage and developed a forecast model to predict the number of patients with influenza using the daily number of ED visits due to fever. The National Health Insurance Service (NHIS) and the National Emergency Department Information System (NEDIS) databases from 2015 to 2019 were used. The definition of fever included having an initial body temperature ≥ 38.0 °C at an ED department or having a report of fever as a patient’s chief complaint. The moving average number of visits to the ED due to fever for the previous seven days was used. Patients in the NHIS with the International Classification of Diseases-10 codes of J09, J10, or J11 were classified as influenza cases, with a window duration of 100 days, assuming the claims were from the same season. We developed a forecast model according to an autoregressive integrated moving average (ARIMA) method using the data from 2015 to 2017 and validated it using the data from 2018 to 2019. Of the 29,142,229 ED visits from 2015 to 2019, 39.9% reported either a fever as a chief complaint or a ≥38.0 °C initial body temperature at the ED. ARIMA (1,1,1) (0,0,1)7 was the most appropriate model for predicting ED visits due to fever. The mean absolute percentage error (MAPE) value showed the prediction accuracy of the model. The correlation coefficient between the number of ED visits and the number of patients with influenza in the NHIS up to 14 days before the forecast, with the exceptions of the eighth, ninth, and twelfth days, was higher than 0.70 (p-value = 0.001). ED-based syndromic surveillances of fever were feasible for the early detection of hospital visits due to influenza.