tetano
Editor, Senior Moderator
Front Vet Sci
. 2022 May 31;9:897763.
doi: 10.3389/fvets.2022.897763. eCollection 2022.
Risk Prediction of Three Different Subtypes of Highly Pathogenic Avian Influenza Outbreaks in Poultry Farms: Based on Spatial Characteristics of Infected Premises in South Korea
Dae-Sung Yoo[SUP] 1 2 [/SUP], Byung Chul Chun[SUP] 1 3 4 [/SUP], Kwan Hong[SUP] 3 [/SUP], Jeehyun Kim[SUP] 3 4 [/SUP]
Affiliations
Abstract
From 2003 to 2017, highly pathogenic avian influenza (HPAI) epidemics, particularly H5N1, H5N8, and H5N6 infections in poultry farms, increased in South Korea. More recently, these subtypes of HPAI virus resurged and spread nationwide, heavily impacting the entire poultry production and supply system. Most outbreaks in poultry holdings were concentrated in the southwestern part of the country, accounting for 58.3% of the total occurrences. This geographically persistent occurrence demanded the investigation of spatial risk factors related to the HPAI outbreak and the prediction of the risk of emerging HPAI outbreaks. Therefore, we investigated 12 spatial variables for the three subtypes of HPAI virus-infected premises [(IPs), 88 H5N1, 339 H5N8, and 335 H5N6 IPs]. Then, two prediction models using statistical and machine learning algorithm approaches were built from a case-control study on HPAI H5N8 epidemic, the most prolonged outbreak, in 339 IPs and 626 non-IPs. Finally, we predicted the risk of HPAI H5N1 and H5N6 occurrence at poultry farms using a Bayesian logistic regression and machine learning algorithm model [extreme gradient boosting (XGBoost) model] built on the case-control study. Several spatial variables showed similar distribution between two subtypes of IPs, although there were distinct heterogeneous distributions of spatial variables among the three IP subtypes. The case-control study indicated that the density of domestic duck farms and the minimum distance to live bird markets were leading risk factors for HPAI outbreaks. The two prediction models showed high predictive performance for H5N1 and H5N6 occurrences [an area under the curve (AUC) of receiver operating characteristic of Bayesian model > 0.82 and XGBoost model > 0.97]. This finding emphasizes that spatial characteristics of the poultry farm play a vital role in the occurrence and forecast of HPAI outbreaks. Therefore, this finding is expected to contributing to developing prevention and control strategies.
Keywords: HPAI; avian influenza; highly pathogenic avian influenza; machine learning; risk assessment; spatial analyses.
. 2022 May 31;9:897763.
doi: 10.3389/fvets.2022.897763. eCollection 2022.
Risk Prediction of Three Different Subtypes of Highly Pathogenic Avian Influenza Outbreaks in Poultry Farms: Based on Spatial Characteristics of Infected Premises in South Korea
Dae-Sung Yoo[SUP] 1 2 [/SUP], Byung Chul Chun[SUP] 1 3 4 [/SUP], Kwan Hong[SUP] 3 [/SUP], Jeehyun Kim[SUP] 3 4 [/SUP]
Affiliations
- PMID: 35711796
- PMCID: PMC9194674
- DOI: 10.3389/fvets.2022.897763
Abstract
From 2003 to 2017, highly pathogenic avian influenza (HPAI) epidemics, particularly H5N1, H5N8, and H5N6 infections in poultry farms, increased in South Korea. More recently, these subtypes of HPAI virus resurged and spread nationwide, heavily impacting the entire poultry production and supply system. Most outbreaks in poultry holdings were concentrated in the southwestern part of the country, accounting for 58.3% of the total occurrences. This geographically persistent occurrence demanded the investigation of spatial risk factors related to the HPAI outbreak and the prediction of the risk of emerging HPAI outbreaks. Therefore, we investigated 12 spatial variables for the three subtypes of HPAI virus-infected premises [(IPs), 88 H5N1, 339 H5N8, and 335 H5N6 IPs]. Then, two prediction models using statistical and machine learning algorithm approaches were built from a case-control study on HPAI H5N8 epidemic, the most prolonged outbreak, in 339 IPs and 626 non-IPs. Finally, we predicted the risk of HPAI H5N1 and H5N6 occurrence at poultry farms using a Bayesian logistic regression and machine learning algorithm model [extreme gradient boosting (XGBoost) model] built on the case-control study. Several spatial variables showed similar distribution between two subtypes of IPs, although there were distinct heterogeneous distributions of spatial variables among the three IP subtypes. The case-control study indicated that the density of domestic duck farms and the minimum distance to live bird markets were leading risk factors for HPAI outbreaks. The two prediction models showed high predictive performance for H5N1 and H5N6 occurrences [an area under the curve (AUC) of receiver operating characteristic of Bayesian model > 0.82 and XGBoost model > 0.97]. This finding emphasizes that spatial characteristics of the poultry farm play a vital role in the occurrence and forecast of HPAI outbreaks. Therefore, this finding is expected to contributing to developing prevention and control strategies.
Keywords: HPAI; avian influenza; highly pathogenic avian influenza; machine learning; risk assessment; spatial analyses.