Epidemic models can fit total deaths while missing inequalities
A peer-reviewed modelling study tests how well epidemic simulations reproduce differences in disease burden across social groups and places. Similar overall death curves can conceal unequal accuracy between groups. A hypothetical vaccination comparison also improves equity and total protection together. These findings come from one computational study with simplified intervention assumptions, rather than a measured vaccination campaign.
Science··Evening
Matching the total can conceal uneven accuracy
Epidemic simulations that reproduce similar overall death curves can differ in how accurately they represent the burden on particular groups. A single peer-reviewed computational study evaluates this gap using social and geographical inequalities. Its definition of model fairness concerns reproducing observed differences; deciding which avoidable inequalities to reduce is a separate public-health judgment.[1]
Examples include COVID-19 comparisons in London and New York, Santiago’s mobility patterns and Colombia’s Zika outbreak. Adding detail does not consistently improve the representation of unequal outcomes. In London, accounting for age-related mortality risk can matter more than elaborating contact patterns.[1]
A hypothetical allocation improves both outcomes
A separate vaccination exercise assumes doses sufficient for 20 per cent of the population. The median reduction in deaths is 68.2 per cent with population-proportional allocation and 73.9 per cent with an inequality-aware comparison. In this model example, the latter strategy improves median outcomes for every group relative to the baseline, alongside better aggregate protection.[1]
Different groupings need different inequality measures
The framework uses a concentration index for ordered groups such as income categories, and a Theil index for categories lacking a natural ordering. The vaccination scenario assumes complete protection and leaves out waning immunity, hesitancy and delivery barriers. These are simulated outcomes rather than measured benefits from an actual campaign. The broader framework also does not identify every social mechanism behind unequal exposure and illness. Its results depend on the groups, epidemic setting and comparison used.[1]
Related columns
For more information on this topic, you can read the related columns.
