Incorporating vaccine effects into epidemiological models: common pitfalls and solutions
Casey E. Middleton, Oliver Eales, James M. McCaw, Freya M. Shearer · 2026
Read Incorporating vaccine effects into epidemiological models: common pitfalls and solutions by Casey E. Middleton, Oliver Eales, James M. McCaw, Freya M. Shearer (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-18571.txt.
Introduction
Casey Middleton and three co-authors examine, in a 2026 paper, pitfalls in adding vaccine effects to epidemiological models. Real-world efficacy numbers often fail to match model parameters directly, because both the vaccine's action and the estimation method matter. For leaky vaccines, which partly rather than fully cut infection risk, naive use of efficacy numbers understates population impact. The authors offer a new formula matching modeled effects to real-world numbers, predicting fewer infections and lower herd immunity thresholds.
Catalog record
| CREATOR | Casey E. Middleton, Oliver Eales, James M. McCaw, Freya M. Shearer |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 60528 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.18571
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Incorporating vaccine effects into epidemiological models: common pitfalls and solutions Casey E. Middleton, Oliver Eales, James M. McCaw, Freya M. Shearer Abstract Incorporating vaccination into mathematical models appears deceptively simple: models integrate vaccine-derived protections, such as reduced susceptibility to infection, using parameters informed by empirical estimates of vaccine efficacy or effectiveness ( VE ). In practice, however, empirical VE estimates often do not correspond directly to the parameters of epidemiological models. Here, we extend previous work to demonstrate that in order to accurately parameterize a model, one must consider both a vaccine’s mechanism of action and the statistic used to infer VE from empirical data. When a vaccine confers leaky protection — that is, vaccination partially rather than completely reduces individual infection risk — we show that common empirical VE estimation methods do not provide directly applicable values for model parameters. Naïve (i.e. direct) incorporation of these VE estimates into models results in an underestimate of population-level vaccine impact. To make progress when these estimates are the only available sources for VE , we introduce a parameterization approach which more accurately aligns the modeled effect of vaccination with empirical estimates. Under this adjusted parameterization approach, models predict fewer total infections and lower herd immunity thresholds for leaky vaccines than would be predicted under current parameterization practices. Our parameterization guidelines and adjustment approach can be used to improve accuracy in models that are used in vaccine decision making and public health planning. 1 Introduction Vaccination is one of the most successful public health interventions to fight transmissible diseases [ 51 ] , and mathematical models are an integral tool used to guide vaccine decision making. Including the effects of vaccination in models of disease transmission empowers public health practitioners to weigh decisions on vaccine allocation policies [ 6 , 33 , 25 , 46 , 9 , 1 ] , predict herd immunity thresholds [ 34 , 5 , 10 , 35 ] , and project expected future disease burden with (versus without) vaccination [ 24 , 49 , 2 , 31 , 42 ] . To incorporate vaccination into transmission models, an assumption must be made about the vaccine’s mechanism of action. All-or-nothing vaccines provide perfect protection (i.e. sterilizing immunity) to some fraction of vaccine recipients, while the remaining fraction receive no protective benefit [ 45 ] . Leaky vaccines provide some level of imperfect protection to all vaccinated individuals. In-between these two extremes exist so-called “heterogeneous” vaccines, where vaccine-derived protection varies among vaccine recipients [ 12 , 36 , 26 , 15 ] . For any assumed vaccine mechanism, developing a mathematical model relies on understanding the vaccine effect parameter(s), which specify the extent to which a vaccine reduces infection risk. Modelers face myriad potential pitfalls when parameterizing such models, because the effects of vaccination as measured by clinical trials and observational studies rarely correspond directly to the parameters of classic epidemiological models. While the existing literature demonstrates that VE estimates may vary depending on both the statistical approach used to infer VE from data and the vaccine mechanism of action [ 16 , 45 , 36 , 43 ] , it does little to guide researchers in best practices for parameterizing mathematical models of vaccination. Here, we describe the VE sources which should preferentially be used to parameterize vaccine effects in mathematical models of disease transmission, depending on a vaccine’s assumed mechanism of action. We review common statistical methods for estimating VE , and demonstrate how naïve (i.e., direct) use of these estimates for parameterization of transmission models is incorrect for some vaccine mechanisms. We then introduce a parameterization approach to align model parameters with available VE estimators. Using a simulation study, we demonstrate that our approach can be used to improve model accuracy in most scenarios and show that theoretical herd immunity thresholds are lower under this adjusted parameterization approach. Models that use this parameterization approach will be more accurate in their predictions for vaccine decision making and public health planning. 2 Measuring vaccine effects Quantifying a vaccine’s direct effect at preventing infection or disease is a fundamental need for public health decision making. During phase 3 of vaccine clinical trials, a first estimate is made of vaccine-derived protection in trial participants, called vaccine efficacy [ 44 , 39 , 17 ] . Once vaccines are available to the general public, “real-world” protection estimates are obtained from observational studies [ 19 , 3 , 7 ] . These estimates are called vaccine effectiveness. Both vaccine efficacy and effectiveness are measurements of vaccine-derived risk reduction, and in this manuscript we use VE interchangeably to refer to either metric. Clinical trials and observational studies can measure VE against a range of outcomes and often have a clinical endpoint. Here, we focus on VE against infection acquisition given its importance in transmission model dynamics. In order to estimate VE from data, it must be defined mathematically. The simplest way to define VE is the reduction in cumulative infection risk for vaccinated individuals, as compared to unvaccinated individuals, measured using the cumulative attack rate ratio (ARR) [ 16 , 39 , 44 ] . We refer to these cumulative risk-based estimates as \textit{VE}_{\text{measured,R}} . Other studies define VE using hazards ratio (HR) estimators [ 37 , 13 , 44 ] , which we refer to as \textit{VE}_{\text{measured,H}} . HR estimates require data on transmission chains [ 37 ] or infection time [ 13 , 44 ] to calculate instantaneous risk, or hazard, conditioned on susceptibility. ARR estimates require less stringent monitoring and are thus the target estimated by most studies, including clinical trials [ 44 , 39 , 17 ] and test-negative designs with no time component [ 19 , 28 ] . These statistical definitions may yield different estimates of VE depending on a vaccine’s mechanism of action [ 45 , 36 ] . 2.1 Measured VE depends on vaccine mechanism of action and study statistic We consider three mechanisms of action to describe the distribution of vaccine-derived protection amongst vaccine recipients: all-or-nothing, leaky, and heterogeneous. Smith, et. al. showed that ARR-based VE estimates may differ from HR-based estimates for all-or-nothing and leaky vaccines [ 45 ] . They find that for all-or-nothing vaccines, VE estimates can accurately recover the direct effect of vaccination using ARR estimators, but will vary with time and converge to \textit{VE}_{\text{measured,H}}\to 1 (indicating perfect protection) using HR estimators. For leaky vaccines, HR estimators can be used to recover the direct effect, but ARR estimators will converge to \textit{VE}_{\text{measured,R}}\to 0 (indicating zero protection) as study follow-up time increases, since an individual vaccinated with a leaky vaccine is guaranteed to be infected given infinite exposure opportunities. Thus, ARR estimators may— due to differential depletion of susceptibles in vaccinated versus unvaccinated groups—incorrectly indicate waning of protection for a leaky vaccine through time, even when true protection is constant [ 20 , 21 , 11 , 45 ] . For heterogeneous vaccines, the ability of HR estimators to recover the direct effect has been shown to depend on the population-level distribution of vaccine-derived protection [ 36 ] . To quantify the temporal variation in \textit{VE}_{\text{measured,R}}\ estimates, we developed a simple model of exposure events as a series of Bernoulli