Long‐ and Short‐Run Asymmetric Effects of Meteorological Parameters on Hemorrhagic Fever with Renal Syndrome in Heilongjiang: A Population‐Based Retrospective Study

Abstract

Examining both long‐term and short‐term effects can enhance the precision and reliability of time series analysis. This study aimed to delve into the asymmetric effects of weather conditions on hemorrhagic fever with renal syndrome (HFRS) in the long and short terms and build a forecasting system. Data comprising monthly HFRS incidents and weather factors in Heilongjiang from January 2004 to December 2019 were extracted. Subsequently, the long‐ and short‐term asymmetric impacts were examined using the autoregressive distributed lag (ARDL) and nonlinear ARDL (NARDL) models. Next, the samples were partitioned into training and testing subsets to evaluate the predictive potential of both models. From 2004 to 2019, HFRS exhibited a declining trend (average annual percentage change = −6.744%, 95% CI: −13.52%–0.563%) and a dual seasonal pattern, with a prominent peak in June and a secondary one in October–December. This study identified long‐term asymmetric effects of rainfall (Wald long‐run asymmetry (WLR) = 3.292, p = 0.001), wind velocity (WLR = −3.271, p = 0.001), and air pressure (WLR = −6.453, p < 0.001) on HFRS. Additionally, this study observed short‐term asymmetric impacts of relative humidity (Wald short‐run symmetry (WSR) = −1.547, p = 0.001), rainfall (WSR = −1.984, p = 0.049), and air pressure (WSR = −2.33, p = 0.021) on HFRS. A unit increase in relative humidity, sunshine hours, and air pressure resulted in about 10.9%, 1.9%, and 13.6% decreases in HFRS, respectively; a unit decrease in relative humidity, rainfall, and sunshine hours led to about 6.7%, 1.8%, and 2% decreases in HFRS, respectively. When temperature increased and decreased by one unit, the HFRS incidence increased by 11.6% and 22.5%, respectively. HFRS also varied significantly with the positive and negative changes in differenced (D) temperature, D (relative humidity), D (wind velocity), D (rainfall), D (air pressure), and D (sunshine hours) at 0−3‐month delays over the short term. The NARDL model exhibited notably lower error rates in forecasting compared to the ARDL model. Meteorological parameters affect HFRS in both the long and short term, often showing asymmetric effects. The NARDL model, capable of incorporating various weather parameters, proves to be valuable in predicting HFRS epidemic and guiding strategies for prevention and control.