The Field Guide · No. 52
How are lives saved estimated?
A lives-saved figure compares observed outcomes with a counterfactual, using deaths, intervention coverage and effectiveness to estimate the difference.
Updated
A lives-saved estimate starts with an ending we can see and compares it with one we cannot. No one can identify each person who survived only because of a seat belt or vaccine. The model estimates how many deaths the same population would have had under a specified alternative, such as no seat belts or no historical vaccination. That imagined comparison is the counterfactual.
There is no single formula for every intervention. A model names the people and years it covers, the deaths that could be affected, how well the intervention works, how many people received it, and what share of deaths the intervention can prevent. The answer changes with those choices. A treatment's effect in a trial cannot tell you by itself how many deaths it prevented across a country; coverage and the starting risk matter too.
The Lives Saved Tool (LiST), built at the Johns Hopkins Bloomberg School of Public Health, shows the moving parts in a rotavirus vaccine example from a 2021 WHO modelling guide. Start with 10,000 deaths from diarrhoea. Increase vaccine coverage from 0% to 50%; suppose rotavirus accounts for 20% of those deaths and the vaccine is 50% effective. LiST multiplies 10,000 × 0.5 × 0.2 × 0.5 and estimates 500 deaths averted. That is one model's recipe, not a formula for every intervention.
Seat-belt estimates use a different calculation. NHTSA's 2009 FAQ works one example from its Fatality Analysis Reporting System: in 2007 it recorded 760 deaths of passenger-car drivers who wore three-point belts and had no frontal air bag. NHTSA used an effectiveness estimate of 0.48 and calculated 0.48 × 760 ÷ (1 − 0.48) = 702 lives saved. The 702 is the model's estimate of additional deaths in those same crash scenarios without the belts. NHTSA repeats the calculation across crash types and adds the scenario estimates. The agency says exact crash details cannot be known for every fatal or potentially fatal crash, so the result is an estimate, not a count of named people.
The 2024 Lancet study by Shattock and colleagues combined a suite of mathematical and statistical models. It estimated deaths averted by routine and supplementary vaccination against 14 pathogens from 1974 through 2024, compared with a hypothetical history without vaccination. The estimate was 154 million deaths averted, including 146 million among children younger than five and 101 million among infants younger than one. WHO described the result as conservative because it covered only 14 diseases, not every possible effect of vaccination.
Do not add unrelated lives-saved headlines into one grand total. The studies may cover different years, populations, causes and counterfactuals. Even two interventions aimed at the same disease can prevent the same death. WHO's guide to LiST says calculating each in isolation would overcount, so the model subtracts deaths already averted before estimating the next intervention. A total only means something beside the assumptions that made it.
What to remember
- A lives-saved estimate compares observed outcomes with a specified counterfactual, or what the model says would have happened without the intervention.
- Models need a defined population and period, baseline deaths, intervention effectiveness and coverage; some also estimate what fraction of deaths the intervention can affect.
- Estimates for interventions that overlap in population or cause of death cannot simply be added without accounting for deaths already averted.
- A published lives-saved number is a model estimate, not a list of identifiable people.
From the record
The equation which relates these three variables is the following: LS = E*F/(1-E).
Asked often
Are lives saved counted one by one?
No. NHTSA says its estimates are not exact counts because no crash database can show what would have happened in every crash if an occupant had not used a belt. The agency applies effectiveness estimates to crash data and calculates a total.
What is a counterfactual in a lives-saved estimate?
It is the model's version of what would have happened to the same population under a different condition, such as no seat belts or no historical vaccination. The lives-saved estimate is the difference between the model's outcomes in the two scenarios.
Can you add lives-saved estimates from different interventions?
Not safely when they cover overlapping people, causes of death, years or assumptions. A model such as LiST accounts for overlap by subtracting deaths already averted before it estimates another intervention's effect.
Read the news better, every morning.
We send you the day's real progress in science, medicine and beyond, told plainly and sourced. Free, forever.
Free for good. One click to leave.