Population Attributable Mortality Associated with Respiratory Viruses in Ontario

This study estimates that SARS-CoV-2 caused a population mortality burden in Ontario approximately three to four times higher than seasonal influenza A between 2020 and 2025, with virus-attributable deaths remaining elevated above pre-pandemic levels even after the public health emergency ended, while highlighting that influenza and RSV mortality estimates are highly sensitive to seasonal modeling adjustments.

The Gist

Imagine the human body and the population of Ontario as a giant, bustling city. Every month, people in this city get sick and, unfortunately, some pass away. For decades, the city's health officials have tried to figure out exactly who is responsible for these deaths.

The problem is like trying to find a specific thief in a crowded market by only looking at the empty spots left behind. Often, death certificates just say "heart failure" or "old age," hiding the fact that a virus like the flu or RSV was the real culprit pulling the strings.

This study is like a team of detectives who decided to stop guessing and start using mathematical telescopes to see the invisible patterns. They looked at 30 years of data (from 1993 to 2025) to answer a simple question: If we could magically stop a specific virus from circulating, how many fewer people would die?

Here is the story of what they found, broken down into simple concepts:

1. The "Seasonal Ghost" (The Old Normal)

Before the big pandemic (1993–2020), the city had a predictable rhythm. Every winter, a ghost named Influenza A would visit.

• The Analogy: Think of Influenza A as a seasonal storm. It comes every year, knocks over a few trees, and causes about 1.8% of all the deaths in the city.
• The Mystery: The detectives also looked for RSV (a virus that hurts babies and the elderly) and Influenza B. But when they used their "seasonal telescope" (a math tool that filters out regular winter patterns), RSV seemed to disappear! It looked like it wasn't causing any deaths at all.
• The Twist: The detectives realized their telescope was too powerful. It was filtering out the virus along with the winter weather. When they turned the telescope off (removed the seasonal filter), RSV suddenly reappeared, showing it was actually responsible for about 2% of deaths. It turns out RSV was there all along, just hiding behind the winter fog.

2. The "New Giant" (The Pandemic Era)

Then, in 2020, a new, massive storm arrived: SARS-CoV-2 (COVID-19).

• The Scale: This wasn't just a seasonal storm; it was a tsunami. Even with vaccines, masks, and treatments, this virus was responsible for 6.1% of all deaths during the pandemic years.
• The Comparison: That is roughly four times worse than the old seasonal flu storms. Even though the city had better tools to fight this new enemy, it still caused a massive amount of loss.
• The Validation: The detectives checked their math against the official death counts reported by the government. Their numbers matched perfectly, proving their "mathematical telescope" was working correctly.

3. The "Aftermath" (Life After the Emergency)

The study didn't stop when the official "Emergency" ended in 2023. They kept watching.

• The Finding: Even after the emergency declaration was over, SARS-CoV-2 didn't just vanish. It settled in as a permanent resident, still causing about 10% of deaths in the later years of the study (though the numbers vary depending on how you count).
• The Flu's Return: Interestingly, the old seasonal flu (Influenza A) didn't suddenly get worse. It went back to its normal, smaller role. The "new normal" isn't that the flu got stronger, but that we now have to live with a much bigger, persistent virus (COVID) alongside it.

4. The "Math Lesson" (Why the Tools Matter)

One of the most important parts of this story is about how you measure things.

• The Analogy: Imagine you are trying to hear a violin in an orchestra.
  • If you use a filter that blocks out all the background noise (the seasonal adjustment), you might accidentally block out the violin if it plays at the same time as the drums.
  • If you don't use a filter, you hear the violin clearly, but you might also hear the drums and think the violin is louder than it is.
• The Takeaway: The study showed that for viruses like the Flu and RSV, the "filter" you choose changes the answer. For the big new virus (COVID), the answer was the same no matter which filter you used. This teaches us that when scientists count deaths, how they do the math matters just as much as the data itself.

The Bottom Line

This paper tells us that while we are used to thinking of the flu as the main winter killer, SARS-CoV-2 has become a much heavier burden on the population, causing 3 to 4 times more deaths than the flu ever did, even with all our modern defenses.

It also warns us that RSV is still a silent killer that we need to pay attention to, especially for the very young and very old. Finally, it reminds us that the pandemic didn't just end; it changed the landscape of our health, and we need to keep our "telescopes" focused on these viruses to protect the city in the future.

Technical Summary

1. Problem Statement

Respiratory viruses (Influenza A/B, RSV, SARS-CoV-2) impose a significant mortality burden, but traditional cause-of-death coding often undercounts their impact due to misclassification and inaccuracies in death certificates. To accurately prioritize public health interventions (such as vaccination), researchers need robust estimates of the Population Attributable Fraction (PAF)—the proportion of deaths that could be prevented if the viral exposure were eliminated.

Existing methods often rely on cause-specific mortality, which is prone to error. This study addresses the need for an indirect, regression-based approach to quantify the true population-level mortality burden of respiratory viruses in Ontario, Canada, over a 32-year period (1993–2025), specifically examining how modeling choices (seasonal adjustment) and the transition from pandemic to post-pandemic eras affect these estimates.

2. Methodology

The study utilized an ecological regression design analyzing monthly all-cause mortality data against laboratory surveillance indicators for viral circulation.

• Data Sources:
  • Mortality: Monthly all-cause death counts from Statistics Canada (1991–1993) and the Ontario Deaths Registry (1994–Feb 2025).
  • Viral Surveillance: FluWatch data for Influenza A/B and RSV (1993–present); SARS-CoV-2 data combined from Ontario's Case and Contact Management System (adjusted case counts, 2020–2022) and FluWatch percent positivity (2022–2025).
  • Population: Monthly population denominators from Statistics Canada.
• Statistical Modeling:
  • Primary Model: Negative binomial regression (chosen over Poisson due to overdispersion) fitted to monthly data.
  • Covariates: Viral indicators (positivity rates or adjusted case counts), linear/quadratic secular trends, and Fast Fourier Transforms (FFT) (sine/cosine terms) to account for seasonal periodicity. Population size was included as a log-offset.
  • Stratification: Analyses were split into Pre-pandemic (Jan 1993–Feb 2020) and Pandemic (Mar 2020–Feb 2025) periods.
• PAF Calculation:
  • PAF was derived by comparing model-predicted deaths under observed viral circulation against a counterfactual scenario where viral coefficients were set to zero.
  • Confidence intervals were generated using the punaf command in Stata.
• Sensitivity Analyses:
  • Seasonal Adjustment: Models were re-run without Fourier terms to test if seasonal covariates were competing with viral predictors for explanatory power.
  • Temporal Stratification: The pandemic period was split into the PHEIC period (Mar 2020–Apr 2023) and Post-PHEIC period (May 2023–Feb 2025).
  • Early Pandemic Exclusion: March–June 2020 was excluded to assess the influence of the initial high-mortality phase.
  • Meta-Analysis: Random-effects meta-analyses (DerSimonian and Laird) were performed to pool PAF estimates across modeling approaches and time periods, assessing heterogeneity ($I^2$).

3. Key Contributions

• Extended Time Horizon: The study provides the first long-term (1993–2025) quantification of virus-attributable mortality in Ontario, covering the pre-pandemic baseline, the acute pandemic phase, and the emerging post-PHEIC endemic phase.
• Modeling Sensitivity: It formally demonstrates that the inclusion of seasonal adjustment terms (Fourier transforms) significantly alters PAF estimates for highly seasonal viruses (Influenza A and RSV), often masking their true mortality impact by attributing variance to general seasonality rather than specific viral circulation.
• Post-PHEIC Characterization: It offers early evidence on the mortality burden of SARS-CoV-2 after the formal end of the Public Health Emergency of International Concern (PHEIC), a period with limited data in existing literature.
• Validation: The model's SARS-CoV-2 attributable death estimates closely matched reported COVID-19 death counts from Public Health Ontario, validating the regression-based approach.

4. Key Results

Pre-Pandemic Period (1993–2020)

• Influenza A: Consistently associated with mortality.
  • With seasonal adjustment: PAF 1.8% (95% CI 1.4–2.3%).
  • Without seasonal adjustment: PAF 2.9% (95% CI 2.5–3.3%).
• Influenza B: No detectable impact on all-cause mortality in any model.
• RSV:
  • With seasonal adjustment: Showed a non-significant inverse association (PAF -0.45%), likely due to statistical competition with seasonal terms.
  • Without seasonal adjustment: Showed a significant positive association with a PAF of 1.9% (95% CI 1.3–2.4%).

Pandemic Period (2020–2025)

• SARS-CoV-2: The dominant driver of mortality.
  • Combined Pandemic PAF (with seasonal adjustment): 6.1% (95% CI 4.2–8.0%).
  • This is approximately 4-fold higher than the pre-pandemic Influenza A burden, despite widespread vaccination, antivirals, and NPIs.
  • Post-PHEIC (May 2023–Feb 2025): SARS-CoV-2 remained a significant predictor with a PAF of 9.8% (95% CI 1.1–17.7%), suggesting a sustained endemic burden.
• Other Viruses: Influenza A, B, and RSV showed no significant association with mortality during the pandemic period, consistent with the suppression of seasonal circulation due to NPIs and SARS-CoV-2 dominance.

Meta-Analysis & Heterogeneity

• Heterogeneity: Substantial heterogeneity was found for Influenza A ($I^2=92.8\%$) and RSV ($I^2=89.1\%$) across modeling approaches, driven largely by the inclusion/exclusion of seasonal terms.
• Robustness: SARS-CoV-2 showed minimal heterogeneity ($I^2=5.5\%$), indicating its mortality signal is robust regardless of modeling choices.
• Combined Burden: Total respiratory virus PAF ranged from 1.5% (pre-pandemic, adjusted) to 10.8% (pandemic, unadjusted).

5. Significance and Implications

• Public Health Burden: SARS-CoV-2 has imposed a mortality burden 3–4 times greater than historical seasonal influenza, even in the presence of advanced countermeasures. This highlights the virus's continued threat in the post-pandemic era.
• Methodological Caution: The study underscores that modeling choices are critical. Excluding seasonal adjustment reveals a hidden mortality burden for RSV and Influenza A that is otherwise masked. Researchers must carefully consider whether seasonal terms are "over-adjusting" and removing viral signal.
• Policy Recommendations:
  • SARS-CoV-2: Surveillance and prevention strategies must be sustained beyond the formal emergency phase, as the virus continues to drive significant mortality.
  • RSV: The ~2% mortality burden (revealed in unadjusted models) supports the urgent implementation of new prevention strategies, such as vaccines for older adults and monoclonal antibodies for infants.
  • Influenza: Continued seasonal vaccination remains vital, though Influenza B appears to have a negligible population-wide mortality footprint compared to A.

In conclusion, this study provides a robust, data-driven framework for quantifying respiratory virus mortality, revealing that the post-pandemic landscape involves a sustained, high-burden endemic phase for SARS-CoV-2 and a previously under-quantified burden for RSV.