Back

Cremation based estimates suggest significant under- and delayed reporting of COVID-19 epidemic data in Wuhan and China

He, M.; Li, L.; Dehner, L. P.; Dunn, L.

2020-06-06 infectious diseases
10.1101/2020.05.28.20116012 medRxiv
Show abstract

BackgroundChinas COVID-19 statistics fall outside of recognized and accepted medical norms. Here we estimated the incidence, death and starting time of the COVID-19 outbreak in Wuhan and China based on cremation related information. MethodsData sources included literature on COVID-19 in China, official Chinese government figures, state-run and non state-run media reports. Our estimates are based on investigative media reports of crematory operations in Wuhan, which is considered as a common data end point to life. A range of estimates is presented by an exponential growth rate model from lockdown (Jan 23,2020) until the intervention started to show effects, which was estimated 14.5 days after lockdown. ResultsFor the cumulative infections and total deaths, under different assumptions of case fatality rates (from 2.5% to 10%) and doubling time 6.4 days, the estimates projected on February 7, 2020 in Wuhan range from 305,000 to 1,272,000 for infections and from 6,811 to 7,223 for deaths - on the order of at least 10 times the official figures (13,603 and 545). The implied starting time of the outbreak is October 2019. The estimates of cumulative deaths, based on both funeral urns distribution and continuous full capacity operation of cremation services up to March 23, 2020, give results around 36,000, more than 10 times of the official death toll of 2,524. ConclusionsOur study indicates a significant under-reporting in Chinese official data on the COVID-19 epidemic in Wuhan in early February, the critical time for response to the COVID-19 pandemic.

Matching journals

The top 8 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.