[Newsdesk] Research in brief
Summary
The number of deaths caused by COVID-19 in the USA during the pandemic may have been underestimated. Researchers used a predictive machine learning algorithm trained on in-patient COVID-19 deaths from March 2020 to December 2021 to predict whether deaths occurring in out-of-hospital settings were likely to be related to COVID-19. The algorithm estimated a total of 995 787 deaths resulted from COVID-19 in the USA in…
The number of deaths caused by COVID-19 in the USA during the pandemic may have been underestimated. Researchers used a predictive machine learning algorithm trained on in-patient COVID-19 deaths from March 2020 to December 2021 to predict whether deaths occurring in out-of-hospital settings were likely to be related to COVID-19. The algorithm estimated a total of 995 787 deaths resulted from COVID-19 in the USA in the study period (95% uncertainty interval [UI] 990 313–1 001 363). This value exceeded the officially reported number of 840 251 deaths by 19% (adjusted reporting ratio 1·19, UI 1·18–1·19), equating to 155 536 unrecognised COVID-19 deaths (UI 150 062–161 112) in the USA during this period.