Credit CDC
#19,112
The assumption is - for practically every infectious disease - that official case counts are significant undercounts; aka `the tip of the pyramid'. Many cases are mild, asymptomatic - or are misdiagnosed - or occur in medically underserved populations and are therefore never reported.
Over the years we've looked at a number of studies which have attempted to quantify these surveillance/reporting gaps, including:
- Last November's JAMA Open: Asymptomatic Human Infections With Avian Influenza A(H5N1) Virus Confirmed by Molecular and Serologic Testing and 2024's MMWR: Serologic Evidence of Recent Infection with HPAI A(H5) Virus Among Dairy Workers,
- Both EID Journal: Estimation of Severe MERS Cases in the Middle East, 2012–2016 and Presence of Middle East respiratory syndrome coronavirus antibodies in Saudi Arabia: a nationwide, cross-sectional, serological study by Drosten & Memish et al., suggest that far more MERS-CoV cases have occurred than have been reported.
- CID Journal: Estimates Of Human Infection From H3N2v (Jul 2011-Apr 2012) which suggested that the number confirmed infections was 1 out of 200 community cases.
- A 2014 seroprevalence study found antibodies against H9N2 ranged from 5.9% to 7.5% among poultry exposed individuals in Egypt, while a 2016 PLoS One study found a seroprevalence in Southern China ranging from 1.37% to 3.42%.
Today we've a preprint which endeavors to model the number of avian flu infections globally each year, its IFR (Infection Fatality Rate), and how pandemic risks might be lowered through spillover prevention.
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:
With that caveat, the authors - using both recent and historical data - estimate that thousands of unreported avian flu infections occur each year, and that if 20% of those could be prevented, it might delay the next pandemic by nearly a decade.
The authors write:
We estimate that, on average, there are 6,441 annual human infections with AIV worldwide, which is much higher than the 986 human cases reported to date and suggests that many infections are undetected and could be because some humans infected with AIV are asymptomatic or symptomatic but not tested.
Based on our estimate of annual AIV infections and the assumption that all AIV deaths in humans are reported, the IFR of 32 deaths per 10,000 infections is much lower than the reported case fatality rate of 48%
It is obviously impossible to account for all of the variables in our chaotic environment, and so to make a model work, certain assumptions must be made.
As an example, the authors used an average of 38 years between pandemics, based on the last 7 zoonotic pandemics going back 245 years (1781–2026). But the gap between the last 2 pandemics was 11 years, and many studies suggest that the frequency and severity of pandemics is rising.
Using an evolutionary epidemiological model of pandemics to estimate the infection fatality ratio for humans infected with avian influenza viruses
Joshua Mack, Michael Li, Amy Hurford
doi: https://doi.org/10.64898/2026.01.21.26344526
Abstract
The risk of highly pathogenic avian influenza virus infection to humans is challenging to estimate as many human avian influenza virus (AIV) infections are undetected because infections may be asymptomatic, symptomatic but not tested, and difficult to identify through contact tracing, as human-to-human transmission is rare.
We derive equations that consider the evolutionary mechanisms that give rise to pandemics and are parameterized to be consistent with records of past pandemics. We estimate that thousands of human AIV infections occur worldwide in an average year and estimate the infection fatality ratio as 32 deaths per 10,000 infections (95% confidence interval: [9.6, 75]). This estimate is comparable to SARS-CoV-2 during the recent pandemic and higher than seasonal human influenza.
We estimate that preventing animal-to-human influenza spillovers would delay pandemic emergence by several years. Preventing human infections with AIV is necessary given the high risk of severe outcomes to individuals and to reduce the risk of pandemics occurring in the future.
(SNIP)
Efforts to prevent human infections with non-pandemic capable genotypes of HPAI virus are necessary given the high individual risk of severe outcomes (as measured by the IFR), but also to lower the risk of a pandemic emerging in any given year.
We estimate that preventing 20% of animal-to-human AIV spillovers annually would delay pandemic emergence by an average of 9.4 years and preventing 50% of spillovers would delay pandemic emergence by 37.5 years.
Measures that prevent the spillover of HPAI virus to humans include not touching, feeding or handling potentially infected birds or other animals, when contact cannot be avoided wearing gloves and a well-fitted respirator or medical mask, reporting infected animals to the appropriate animal health authority [7], the humane destruction of infected and exposed animals, and strict quarantine and animal movement controls to prevent disease spread [10].
Regardless of the actual numbers, this study suggests that thousands of undetected bird flu infections happen every year, with a significant risk of death. With each infection also comes a small chance for the virus to mutate into a human-adapted pathogen, capable of sparking a pandemic.
Farm workers, vets, hunters, poultry and other livestock handlers, and animal rescue personnel are at particularly high risk of exposure, and their use of proper PPEs (gloves, masks, etc.) and following other biosecurity measures could help lower those risks.
