Reported pertussis cases in China climbed dramatically from 6,658 in 2015 to 476,690 in 2024, representing a 71.6-fold increase. The series contained a marked decline in 2020 during COVID-19 non-pharmaceutical interventions and then an explosive resurgence in 2023–2024 despite sustained reported DTaP3 coverage above 95%. The surveillance record therefore shows both interruption and rebound in transmission over the decade.
The study team fitted a seasonally forced SEIARV compartmental model to national monthly pertussis case counts covering January 2015 through December 2024. Waning immunity was included in the model formulation; the authors reported annual parameter fitting rather than a single static parameter set for the entire decade. The source did not provide additional methodological details beyond the use of seasonal forcing, compartment labels SEIARV, and inclusion of a waning term in the vaccine protection compartment.
Investigators reconstructed the time-varying effective reproduction number Re across the study interval. Re ranged from a low of 0.88 in 2020 to a peak of 1.37 in 2023. Pre-pandemic Re values fell between 0.98 and 1.07. The fitted baseline transmission rate β0 increased from 8.96 month−1 in 2020 to 13.82 month−1 in 2023, a 29% rise versus the pre-COVID-19 mean reported by the authors. Annual in-sample fitting performance was strong: reported annual fitting errors were below 5% and the in-sample root-mean-square error (RMSE) was 2.6%.
The authors ran a counterfactual analysis to estimate the indirect effects of pandemic-era interventions on the susceptible population. Their estimate indicated that COVID-19 non-pharmaceutical interventions increased the effective susceptible population by approximately 14% by early 2023. The paper explicitly characterizes that 14% figure as an upper-bound estimate; the increase cannot be fully disentangled from contemporaneous shifts in vaccine acceptance or hesitancy during the pandemic period. In other words, the model cannot fully separate susceptibility increases driven by reduced exposure from those driven by reductions in vaccine uptake.
Forecasting used 500-member stochastic ensembles to project incidence for 2025–2027 across four transmission scenarios defined by the authors. Under a strengthened-control scenario, median projected annual cases fell to near zero by 2027 (median 0; 95% prediction interval 0–155). By contrast, scenarios that assumed continued vaccine waning or a return to the historical-maximum transmission produced far larger burdens: median annual case counts of 9.3 million and 10.8 million, respectively. The authors reported outbreak probabilities of 97–100% for these high-transmission or sustained-waning scenarios. The source text did not provide detailed breakouts of the four scenario definitions beyond these summary projections.
Sensitivity analysis within the modeling framework indicated that waning of acellular-vaccine protection materially amplified transmission. The authors reported a sensitivity index of Re to the waning rate of +0.79, signifying substantial influence of the waning parameter on transmission potential in the model. The fitted baseline transmission increase and the effect of waning both contributed to the resurgence signal; the study team emphasized that, in a homogeneous national model, their separate numerical contributions cannot be formally decomposed.
The authors acknowledged several inherent limitations that arose directly from the modeling choices and data. First, the homogeneous national model structure prevents formal decomposition of the separate contributions of increased baseline transmission versus waning vaccine-derived immunity. Second, the counterfactual estimate attributing a roughly 14% rise in effective susceptibility to COVID-19 non-pharmaceutical interventions is explicitly presented as an upper-bound and may conflate with pandemic-era vaccine hesitancy; the source did not report analyses that fully separate these factors. Third, while the model fit statistics (annual errors <5%, in-sample RMSE 2.6%) are reported, the source did not provide granular methodological detail on parameters, priors, or compartment transition rates beyond noting seasonal forcing, waning immunity, and annual parameter fitting. Finally, the authors used 500-member stochastic ensembles for forecasting but did not publish the full scenario definitions or the precise stochastic model implementation in the abstract; those details were not reported in the source abstract.
These limitations map onto concrete unanswered questions left by the analysis: how much of the 2023–2024 resurgence is attributable to behavioral changes or increased transmissibility versus immunological waning at the individual level? What are the regional heterogeneities that a national homogeneous model cannot resolve? Which policy interventions (for example, booster schedules or targeted catch-up campaigns) would most effectively shift the high-burden forecast scenarios toward the strengthened-control projection? The source documents these gaps and frames them as constraints on interpretation rather than resolved findings.