An AHP-multiple logistic regression model for risk assessment of highly pathogenic avian influenza

Quantitative risk assessment of highly pathogenic avian influenza ( HPAI ) is crucial to timely formulate and adjust effective prevention and control strategies for epidemic prevention agencies. However, the risk assessment models based on risk level classification are very few. This study established a quantitative risk assessment model for early risk warning of HPAI. Through collecting data on reported cases of HPAI from July 2007 to July 2022, the outbreak characteristics of HPAI were analyzed in terms of the outbreak time, virus subtype, and outbreak location. The risk factor weights were determined by the analytic hierarchy process ( AHP ), and the risk levels of HPAI reports were quantified using a multi-indicator composite score. Results showed that the highest frequency of HPAI outbreaks were from November to April. H5N8 and H5N1 subtypes were the most susceptible accounting for 48.3 % and 30.9 % of outbreaks, respectively. H5N3, H5N9, H7N2, and H7N1 subtypes had the highest mortality of about 100 %. Chinese Taipei, France, Germany, Hungary, and Nigeria had the largest number of HPAI reports over the past decade, which had more than 500 reports. The risk assessment model based on AHP-multiple logistic regression was built with an accuracy rate of 93.3 %. The quantitative analysis showed that R ≥ 0, −0.4 ≤ R < 0, −0.6 ≤ R < −0.4, and R < −0.6 could be used to classify as four risk levels of significant, high, medium, and low, respectively. The comparison showed that the Monte Carlo simulation risk assessment model based on @Risk software couldn't satisfy the requirements of HPAI data analysis due to the crossed risk ranking thresholds and inaccurate risk ranking. Therefore, the risk assessment model based on AHP-multiple logistic regression can be used to assess the risk of HPAI occurrence, which has advantages including the operable method, the accurate evaluation of the epidemic situation, and the risk factors significantly correlated to risk results.