Forecasting Respiratory Illnesses at the Milano Cortina 2026 Winter Olympic and Paralympic Games Using Structured Expert Elicitation

Background: Respiratory illness is consistently the most common cause of medical encounters among athletes at the Olympic and Paralympic Games, yet illness burden varies substantially across Games depending on environmental conditions and prevention practices.Forecasting respiratory illness risk for Team USA is challenging because each Games represents a singular, non-repeatable event, limiting the predictive utility of historical surveillance data alone.The purpose of this study was to develop a prospective Bayesian framework for forecasting respiratory illness burden among Team USA athletes at the Milano Cortina 2026 Winter Olympic and Paralympic Games that formally incorporates expert uncertainty alongside historical benchmarks.Methods: Structured expert elicitation was conducted using the Sheffield Elicitation Framework to quantify clinician expectations of respiratory illness burden under defined prevention-adherence scenarios.Two independent expert panels (one Olympic, 1 one Paralympic) were assembled that each consisted of five clinicians with direct experience providing care at multiple prior Games.Experts provided probabilistic estimates of expected respiratory illness counts under baseline, permissive/low-adherence, and strict/highadherence prevention scenarios.These elicited distributions were integrated into Bayesian forecasting models to generate prospective predictive distributions for illness counts and incidence.In parallel, benchmark forecasts were constructed using Bayesian hierarchical Poisson models fitted to published global Winter Olympic and Paralympic respiratory illness surveillance data.Results: Expert-based forecasts of respiratory illness burden varied widely by prevention-adherence scenario and Games type.For the Olympic Games, mean forecasted illness counts ranged from 13.9 cases under strict prevention to 37.5 cases under permissive conditions (3.38-9.13illnesses per 1,000 athlete-days), with a baseline estimate of 30.9 cases (7.5 per 1,000 athlete-days).For the Paralympic Games, corresponding estimates ranged from 3.52 to 9.07 cases (5.61-14.4per 1,000 athlete-days), with a baseline estimate of 6.24 cases (9.91 per 1,000 athlete-days).In contrast, historical benchmark models predicted substantially lower illness burden, with mean incidence of 2.63 per 1,000 athlete-days (10.8 cases) for the Olympic Games and 3.56 per 1,000 athlete-days (2.2 cases) for the Paralympic Games.Across both approaches, forecasted incidence was consistently higher for the Paralympic Games, and expert-based forecasts exhibited markedly wider uncertainty than historical models.Conclusions: This work demonstrates the feasibility of using structured expert elicitation within a Bayesian framework to support prospective medical planning and enable transparent post-Games evaluation of illness forecasts and prevention strategies.