Credit: CC0 Public Domain
You've likely seen the "25 million lives saved by COVID vaccines" number floating around again recently. It is reporting with new studies, new press releases, and new headlines - like it is fact. But the thing is, it is still based on models, not real-world data.
The original estimate was out of a 2022 simulation completed by Watson et al. and now another group is following up with a 2025 report (Jamal et al.) where they are pushing the same number again. Both rely on predictive models that make assumptions such as: how deadly the virus would have been without vaccines, how many people would've been infected by then, and how effective the vaccines were in every scenario.
They aren't measuring real-world deaths avoided, they are generating "what if" worlds using computer code and best guesses.
Change the inputs a bit (e.g. the infection fatality rate, or excess mortality baseline), and the entire outcome will change. The nuance is usually lost when it is being reported. You only get the headline, 25 million saved.
It is not that they are not useful - they just aren't real. They are simulations built on multiple layers of assumptions.
The next time you come across that statistic, you may want to ask yourself: was that from real-world evidence - or was it from computer simulation?
References:
https://jamanetwork.com/journals/jama-health-forum/fullarticle/2836434
https://www.thelancet.com/journals/laninf/article/PIIS1473-3099(22)00320-6/fulltext