This analysis used disease estimates from the Global Burden of Disease (GBD) 2021 study together with global varicella vaccination policy information to quantify the burden of varicella among children aged 5–9 years and to forecast trends through 2030. The primary outcomes estimated were incidence, prevalence, and disability-adjusted life years (DALYs) for the target age group. An Extreme Gradient Boosting (XGBoost) forecasting model was trained on available inputs to predict 2030 metrics, and Shapley Additive exPlanations (SHAP) were applied to identify the most influential predictors for model outputs.
Using GBD 2021 inputs, the study reports that from 1990 to 2021 the global rate of varicella incidence among children aged 5–9 years increased slightly from 1,028.46 per 100,000 to 1,048.24 per 100,000. Prevalence rose modestly over the same period from 30.81 per 100,000 to 31.07 per 100,000. In contrast, the burden measured in DALYs declined, falling from 21.09 per 100,000 in 1990 to 12.36 per 100,000 in 2021. These observed trends reflect changes captured in the GBD dataset and summarized by the authors.
The trained XGBoost model produced projections for 2030 indicating declines in all three primary measures for the 5–9-year-old cohort. Projected values for 2030 are an incidence of 897.01 per 100,000, prevalence of 28.96 per 100,000, and DALYs of 11.71 per 100,000. The model thus anticipates a reduction in incidence and prevalence compared with 2021 values as well as a further modest decline in DALYs relative to 2021.
Across the 1990–2030 timeframe examined, notable regional differences were reported. Southeast and South Asia had the highest incidence of varicella in children aged 5–9 years. The highest prevalence estimates occurred in Asian and Pacific Island nations. The largest DALY burdens were concentrated in sub-Saharan Africa and South Asia. The authors emphasize that although global measures decline, these geographic disparities are projected to persist into 2030.
SHAP analysis was used to rank predictors driving the model outputs. Population size emerged as the strongest single predictor, with mean absolute SHAP values reported as 0.167 for incidence and 0.103 for prevalence. The analysis also indicated that a higher vaccination program tier may be associated with lower DALYs, suggesting that policy-level vaccination program characteristics relate to reductions in disease burden as captured by the model.
The combination of GBD-derived estimates and XGBoost forecasts provides a global view of varicella dynamics in young school-age children and highlights persistent regional inequities. The findings suggest potential value in strengthening vaccination programs where feasible: higher vaccination program tier in the analysis correlated with lower DALYs. The authors note that these results can inform public health planning and prioritization by indicating regions where varicella burden remains comparatively high and where interventions could have the greatest impact.
The authors explicitly note important limitations in the underlying data that affect interpretation. Specifically, the absence of granular vaccination coverage and cohort-level immunization exposure data limited the ability to attribute observed trends and projected changes directly to vaccination effects. Because the study relied on GBD inputs and policy-tier indicators rather than detailed coverage histories or cohort exposure, the vaccine effect estimates are constrained. The authors recommend that future studies incorporate both vaccination coverage and cohort-level exposure information to strengthen policy evaluation and improve the capacity to link immunization strategies with observed changes in varicella burden.
Overall, the study reports a mixed picture: while measured DALYs for varicella in children aged 5–9 years have declined and model projections predict further reductions by 2030, incidence and prevalence trends and clear regional disparities indicate that targeted public health actions and improved data on vaccination coverage and cohort exposure are needed to fully assess and optimize policy impact.