Climate-Driven Perturbations in Influenza Subtype Positivity Before and After the COVID-19 Pandemic: Observational Study

Abstract Background The COVID-19 pandemic significantly altered global influenza transmission patterns. Whether the associations among meteorological conditions, visibility, and influenza subtype (A/H1N1, A/H3N2, and B/Victoria) positivity changed after the pandemic remains unclear. We compared these associations before and after the COVID-19 pandemic in northern China. Objective This study aimed to characterize changes in the associations of meteorological variables and visibility with influenza subtype positivity before and after the COVID-19 pandemic and to identify subtype-specific nonlinear patterns and thresholds. Methods Using 13 years of influenza surveillance and environmental data, we applied Extreme Gradient Boosting enhanced with Shapley additive explanations (SHAP) to characterize the relative importance and direction of associations between environmental variables and influenza subtype positivity. Generalized additive models (GAMs) were subsequently used to characterize nonlinear association patterns and identify thresholds at which the direction of model-estimated associations changed. Results Absolute humidity (AH) remained the most influential environmental factor for A/H3N2 and A/H1N1, accounting for 32.2% and 38.5% of relative feature importance, respectively. Visibility increased in importance after the pandemic, rising from being the fourth to the second most influential environmental indicator. GAM analysis identified thresholds corresponding to points where the direction of the associations changed. For A/H3N2, AH below 4 g/m 3 was associated with positive SHAP values, whereas higher AH was predominantly associated with negative SHAP values. For A/H1N1, the direction of association changed at an AH threshold of 4.5 g/m 3 (95% CI 4.0-4.8) in the postpandemic period. Additionally, relative humidity (RH) demonstrated a complex nonlinear association with A/H3N2 positivity, with transition points at 45.1% (95% CI 43.5%-46.1%) and 69.7% (95% CI 69.2%-70.2%) in the prepandemic period and at 48.7% (95% CI 46.7%-50.4%) and 70.6% (95% CI 66.3%-72.6%) in the postpandemic period. RH values within these ranges were associated with negative SHAP values, whereas values outside these ranges were associated with positive SHAP values. In contrast, visibility showed opposing association patterns for the A/H3N2 and B/Victoria subtypes after the pandemic. Visibility below 17.1 km (95% CI 16.3-17.4) was associated with negative SHAP values for A/H3N2, whereas visibility below 18.4 km (95% CI 18.0-18.9) was associated with positive SHAP values for B/Victoria. Conclusions AH remained a key environmental determinant of influenza A subtype positivity, while visibility became more important after the pandemic and showed opposing association patterns for A/H3N2 and B/Victoria. The identified nonlinear thresholds highlight subtype-specific environmental associations and may inform influenza surveillance and risk assessment under changing environmental conditions.