193

David Fisman.October 31, 2026 at 3:09 PM EDT

New paper out in Lancet Regional Health Americas, with Alicia Grima, Natalie Wilson, Simran Mann, Ashleigh Tuite and Clara Lee. How many Ontario deaths are attributable to respiratory viruses, and how does SARS-CoV-2 compare to flu? 🧵 doi.org

Death certificates undercount viral deaths. Flu kills people through heart attacks, strokes and secondary bacterial infections that never get coded as flu. So we used 30+ years of monthly all-cause deaths (1993-2025) for Ontario and asked how mortality moves when each virus circulates. Pre-pandemic, influenza A was associated with 1.8% of all deaths in Ontario, about 1600 a year. That fits the familiar ~3500 Canadian flu deaths per year, scaled to Ontario. Influenza B showed no detectable signal. From March 2020 to Feb 2025, SARS-CoV-2 was associated with 6.5% of all deaths. In a joint model that's 3.4-fold the pre-pandemic flu A burden (95% CI 2.2-4.9). And that's with vaccines, antivirals, NPIs and lots of prior immunity in play. A bit of external validation: our model attributed 18,052 deaths to SARS-CoV-2 (Mar 2020-Mar 2024). Ontario's independently adjudicated COVID deaths over the same period: 18,594. Within 3%. Ecological attribution studies rarely get a benchmark like that. The outsized impact on attributable death by SARS-2 didn’t end with the end of the pandemic. After the PHEIC ended (May 2023-Feb 2025), SARS-CoV-2 was still associated with 9.8% of deaths (95% CI 1.1-17.7). Wide interval, not different from the PHEIC era, but clearly not zero. (And of course it will be possible to update this analysis on a rolling basis now. We stop in 2025 because those were the most recent data when the paper was submitted). Here is the above information in graphical form:

Why would excess attributable deaths persist? Our method doesn't care what's on the death certificate. Post-COVID PE, other vascular events and invasive bacterial infection deaths get coded as something else. Booster uptake has also fallen, and protection wanes *sorry, PE is pulmonary embolism. More on that here: sciencedirect.com

A methods lesson: seasonal adjustment matters a lot for strongly seasonal viruses. RSV went from a nonsensical negative association to ~1.9% of deaths when Fourier (seasonal) terms were dropped. Flu A went from 1.8% to 2.9%. SARS-CoV-2 was stable either way (6.5% vs 8.0%). Bottom line: COVID is not "just like flu." It has carried a population mortality burden several times larger than flu A, and that burden continued past the emergency phase. Sustained surveillance and prevention still matter. (It is odd to me that clinicians who emphasize the importance of flu shots tell people not to worry about COVID boosters) All data and code are open: github.com (Zenodo archived). This specific study was unfunded but our group is supported by funds from CIHR for which we are very grateful.

1 / 1