Background:
Dengue transmission is tightly modulated by monsoon variability, yet booster scheduling remains calendar-fixed and rarely incorporates immune activation or heterotypic exposure history. We aimed to derive a monsoon-responsive booster timing framework integrating rainfall anomaly indices, operational vector surveillance signals, secondary heterotypic serostatus, and circulating CXCL10 dynamics to define a biologically actionable pre-amplification window.
Methods:
Weekly laboratory-confirmed dengue cases from Indonesian Ministry of Health surveillance archives (2013–2025; 119,237 cases across 55 planning areas) were linked to NEA Gravitrap Aedes aegypti Index and Breeding Percentage datasets at subdistrict resolution. Rainfall anomalies were calculated from CHIRPS 0.05° gridded precipitation relative to the 1981–2010 baseline. Immunologic data were obtained from ImmPort dengue immunoprofiling datasets (including SDY1530), comprising 412 PCR-confirmed dengue cases with serial CXCL10 concentrations and plaque-reduction neutralization titers stratified by serostatus. Distributed-lag nonlinear models estimated rainfall–vector–incidence coupling (0–12 week lags). Augmented synthetic control constructed counterfactual epidemic trajectories under alternative booster timing scenarios. Causal forests quantified heterogeneity by CXCL10 tertiles and secondary heterotypic serostatus with doubly robust estimation.
Results:
Rainfall anomalies exceeding +1 SD were associated with a 22% increase in Gravitrap Index within 3 weeks (β=0.21, SE 0.05, p<0.001), followed by a 34% rise in weekly dengue incidence at 6-week lag (incidence ratio 1.34, robust SE 0.03, p<0.001). Subdistricts persistently in the upper quartile of Breeding Percentage showed a median epidemic acceleration of 2.3 weeks (IQR 1.6–3.1) compared with lower quartiles and a 29% higher cumulative seasonal case burden. In the immunologic cohort, secondary heterotypic cases demonstrated significantly elevated peak CXCL10 levels (median 352 pg/mL vs 181 pg/mL in primary infection, p=0.001) and slower neutralizing antibody stabilization, with PRNT50 titers reaching plateau at median day 19 versus day 13 in low-CXCL10 strata. CXCL10 levels correlated inversely with early neutralization kinetics (Spearman ρ=−0.44, p=0.008). Causal forest modeling identified the high-CXCL10 secondary-serostatus stratum as the dominant effect-modifier, with projected peak-intensity reduction of 19.8% under an 8-week pre-amplification booster strategy versus 6.1% in low-CXCL10 groups. Augmented synthetic control across five high-transmission years demonstrated mean seasonal peak attenuation of 11.2%, reduction in epidemic duration by 1.7 weeks, and a 14% decrease in cumulative cases during the primary transmission wave when booster timing was synchronized to rainfall–vector lead indicators.
Conclusions:
Integrating rainfall anomalies, vector surveillance metrics, heterotypic serostatus, and CXCL10-defined immune activation delineates a quantifiable pre-amplification window for dengue booster optimization, producing measurable attenuation of seasonal epidemic intensity and duration.