Comparison of 3 early warning methods for COVID-19 outbreak in Dalian City based on observation data in the fever clinic

AN Qingyu; WU Jun; GUO Linan

doi:10.3969/j.issn.1007-8134.2024.01.011

Abstract：Objective Based on the data of observation cases in the fever clinic, to identify the best methods for early warning of COVID-19 outbreak in Dalian city, Liaoning province. Methods Poisson distribution, cumulative sum and exponentially weighted moving average were used to establish an early warning model for COVID-19 outbreak in Dalian City and to evaluate and compare the effects of early warning in terms of the sensitivity, specificity, Yoden index and positive predictive value and so on. Results Based on the data from July 1, 2021 to December 19, 2021 and February 24, 2022 to July 21, 2022, 28 medical institutions in Dalian reported observation cases in the fever clinic, three early warning models were established. The result showed that Poisson distribution model when the a=0.03, sensitivity, specificity, positive predictive value and negative predictive value was 94.59％, 41.38％, 33.98％ and 96.00％ respectively, and early warning signals issued 6 days before the 3 actual COVID-19 outbreak. CUSUM model when K=1.5, H=2, sensitivity, specificity, positive predictive value and negative predictive value of C2 was 21.67％, 95.12％, 52％ and 83.27％, respectively, the early warning signals issued 2 days and 1 day before the actual COVID-19 outbreak in November 2021. EWMA model when λ=0.6 or 0.7, K=1, sensitivity, specificity, positive predictive value and negative predictive value was 35.48％, 80.33％, 31.43％ and 83.05％ respectively, the early warning signals issued 6 days and 5 days before the actual COVID-19 outbreak in November 2021 and April 2022. Conclusions Compared with the CUSUM and EWMA model , the Poisson distribution model based on the observation cases in the fever clinic was effective and reliable in early-warning the COVID-19 outbreak in Dalian city. In the practical application process, if the surveillance data of fever clinic is informationized and real-time analysis of the surveillance data is realized, the early warning response can be further improved.