Credit MRC Centre
#15,528
One of the most contentious, and least precise, metrics from this (or any other) pandemic is how many lives have been lost to the virus. The debate over the death toll from the pandemics of 1918, 1957, 1968 and 2009 continues in academia, with 1918 listed as anywhere from 40 million to 100 million deaths.
The CDC's recently published Burden Of The 2019-2020 Flu Season estimated the number of flu related deaths as 22,000, but that was averaged from a range of 17,970 - 29,053.Even during non-pandemic flu seasons the number of deaths, illnesses, and hospitalization attributed to regular influenza can only be broadly estimated.
Even for something as specific, and unusual as pediatric influenza-related deaths - which are required by law to be reported to the CDC - the CDC estimates that their surveillance captures only about 1/3rd of the cases.
As unsatisfying as it might be, the only recourse is to use mathematical modeling to try to estimate the burden of a disease - whether it be seasonal flu, the incidence of malaria in Africa, or a COVID-19 pandemic.
While imperfect, we use computer models every day to try to mathematically simulate real- world events; everything from climate change to Hurricane Forecasting to traffic engineering.
The late George E. P. Box (18 October 1919 – 28 March 2013) - Professor Emeritus of Statistics at the University of Wisconsin - is often credited with coining the familiar adage - “All models are wrong, but some models are useful.”
While no one modeling approach can be assumed to be reliable, when multiple approaches produce similar results, we can begin to get more comfortable with their output.
Like other estimates we've seen, this study puts the likely IFR in the vicinity of 1% in high income countries - which tend to have a greater concentration of elderly individuals - and substantially lower in countries with a younger demographic.
This the full 18-page report is available, but will be of most interest to those with a strong statistical background. I've provided the link and the author's summary below.
Report 34: COVID-19 Infection Fatality Ratio: Estimates from Seroprevalence
Nicholas F Brazeau 1 , Robert Verity 1 , Sara Jenks 2 , Han Fu 1 , Charles Whittaker 1 , Peter Winskill 1 , Ilaria Dorigatti 1 , Patrick Walker 1 , Steven Riley 1 , Ricardo P Schnekenberg 3 , Henrique Hoeltgebaum 4 , Thomas A Mellan 1 , Swapnil Mishra 1 , H Juliette T Unwin 1 , Oliver J Watson 1 , Zulma M Cucunubá 1 , Marc Baguelin 1 , Lilith Whittles 1 , Samir Bhatt 1 , Azra C Ghani 1 , Neil M Ferguson 1 , Lucy C Okell 1 十 .
Summary
The infection fatality ratio (IFR) is a key statistic for estimating the burden of coronavirus disease 2019 (COVID-19) and has been continuously debated throughout the current pandemic. Previous estimates have relied on data early in the epidemic, or have not fully accounted for uncertainty in serological test characteristics and delays from onset of infection to seroconversion, death, and antibody waning.
After screening 175 studies, we identified 10 representative antibody surveys to obtain updated estimates of the IFR using a modelling framework that addresses the limitations listed above.
We inferred serological test specificity from regional variation within serosurveys, which is critical for correctly estimating the cumulative proportion infected when seroprevalence is still low. We find that age-specific IFRs follow an approximately log-linear pattern, with the risk of death doubling approximately every eight years of age.
- Using these age-specific estimates, we estimate the overall IFR in a typical low-income country, with a population structure skewed towards younger individuals, to be 0.23% (0.14-0.42 95% prediction interval range).
- In contrast, in a typical high income country, with a greater concentration of elderly individuals, we estimate the overall IFR to be 1.15% (0.78-1.79 95% prediction interval range).
We show that accounting for seroreversion, the waning of antibodies leading to a negative serological result, can slightly reduce the IFR among serosurveys conducted several months after the first wave of the outbreak, such as Italy. In contrast, uncertainty in test false positive rates combined with low seroprevalence in some surveys can reconcile apparently low crude fatality ratios with the IFR in other countries. Unbiased estimates of the IFR continue to be critical to policymakers to inform key response decisions. It will be important to continue to monitor the IFR as new treatments are introduced.
The code for reproducing these results are available as a R Research Compendium on Github: `mrc- ide/reestimate_covidIFR_analysis`.
(SNIP)
In summary, we estimate that the overall COVID-19 IFR ranges from 0.14 - 0.42% in low income countries to 0.78 - 1.79% in high income countries, with the differences in those ranges reflecting the older demography of high income settings. The IFR is also likely to vary depending on available healthcare and underlying health conditions.
Our results suggest that the overall risk of death from COVID-19 doubles with approximately every eight years of age. Our estimates of the IFR of COVID-19 are consistent with early estimates and remain substantially higher than IFR estimates for seasonal influenza (<0.1% in the USA) 40 .
To the best of our knowledge, this is the first study accounting for seroreversion as part of IFR estimation as well as simultaneously accounting for uncertainty in serological test characteristics and delays from infection to death and seroconversion. As the pandemic progresses, it will be important to continue to update these estimates to capture changes driven by improvements in care and potential genetic mutations in the virus.
While I expect the IFR debate to continue for years to come, most of the analyses we've seen have centered around that 1% mark, making COVID-19 many times deadlier than seasonal influenza.
