A nationwide ecological time-series analysis found no stable temporal or predictive relationship between dengue and COVID-19 surveillance records in Mexico from 2020 to 2025

Using national surveillance records, we evaluated whether COVID-19 dynamics explain, anticipate, or improve prediction of weekly dengue surveillance-record burden in Mexico during 2020–2025. We analysed 1,085,869 dengue records and 20,632,470 COVID-19 records. Audit of daily symptom-onset coverage identified 10 incomplete dengue weeks clustered around year-end/source snapshot boundaries; these weeks were treated as missing rather than zero, leaving 8,695 state-week observations from 32 entities for inferential analyses. Primary weekly counts were transformed as log10(records + 1), with a population-standardized sensitivity analysis using fixed 2020 INEGI populations. We examined pooled, year-specific, pandemic/post-pandemic, state-specific seasonal, nonlinear time-adjusted and lagged associations, and performed a direct predictive ablation. The audited pooled association was inverse (Pearson r = -0.331, bootstrap 95% CI -0.349 to -0.312; Spearman rho = -0.311), but its direction and magnitude varied by year, from r = 0.096 in 2021 to r = -0.312 in 2023. Population standardization yielded r = -0.386, whereas removal of both state and calendar-week effects reduced the association to r = -0.049. After state-specific adjustment for year, cyclic epidemiological-week structure and nonlinear time, the pooled residual correlation was r = -0.055; after additional AR(4) prewhitening, the largest absolute lag correlation was 0.053. In the direct ablation, adding COVID-19 lags to dengue lags, seasonal week and state effects changed test RMSElog10 from 0.1753 to 0.1749 (delta = -0.0004; block-bootstrap 95% interval − 0.0019 to 0.0011). These findings support no stable direct temporal or predictive coupling between the two surveillance series; the pooled inverse association is primarily a feature of calendar-time and epidemic-regime structure.