Estimating the case fatality ratio of the COVID-19 epidemic in China
Xing Wang,Zihui Ma,Yi Ning,Chen Chen,Rujin Chen,Qiwen Chen,Heng Zhang,Chunming Li,Yan He,Tao Wang,Cheng Tong,Junqing Wu,Yuyan Li,Handong Ma,Shaodian Zhang,Hongxin Zhao
DOI: https://doi.org/10.1101/2020.02.17.20023630
2020-01-01
MedRxiv
Abstract:Background: Corona Virus Disease 2019 (COVID-19) due to severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) emerged in Wuhan city and rapidly spread throughout China since late December 2019. Crude case fatality ratio (CFR) with dividing the number of known deaths by the number of confirmed cases does not represent the true CFR and might be off by orders of magnitude. We aim to provide a precise estimate of the CFR of COVID-19 using statistical models at the early stage of the epidemic. Methods: We extracted data from the daily released epidemic report published by the National Health Commission P. R. China from 20 Jan 2020, to 20 Feb 2020. Competing risk model was used to obtain the cumulative hazards for death, cure, and cure-death hazard ratio. Then the CFR was estimated based on the slope of the last piece in joinpoint regression model, which reflected the most recent trend of the epidemic. Results: As of 20 Feb 2020, totally 75,570 cases were diagnosed as COVID-19 in China. The CFR of COVID-19 were estimated to be 8.71% (95% CI: 7.95%-9.64%) in Hubei province, including Wuhan, the epicenter, and 1.21% (95% CI: 1.07%-1.40%) in other areas of China, respectively. We observed obvious decreasing trends of CFR for COVID-19, with three distinct turning points on January 30, February 6, and February 14 for Hubei province, and one turning point on February 7 for other areas, respectively. Conclusions: Based on analyses of public data, we found that the CFR in Hubei was much higher than that of other regions in China, over 7 times in all estimation. The CFR would follow a downwards trend based on our estimation from recently released data. Nevertheless, at early stage of outbreak, CFR estimates should be viewed cautiously because of limited data source on true onset and recovery time.