Time series modelling of swine lung lesion prevalence to predict the temporal dynamics of Mycoplasma hyopneumoniae and Actinobacillus pleuropneumoniae infections in Spain

Abstract

Respiratory diseases are considered one of the most important problems in swine production worldwide due to the significant economic losses associated. Lung lesion evaluation at slaughterhouses by different scoring systems is commonly used to monitor respiratory diseases in swine. Concretely, cranioventral pulmonary consolidation lesions are associated with enzootic pneumonia (EP) caused by Mycoplasma hyopneumoniae (Mhyo); whereas haemorrhagic necrotizing pneumonia, mainly in the dorso-caudal lung lobes, and chronic pleuritis (CP) are associated with Actinobacillus pleuropneumoniae (App). Despite the recent consideration of several statistical methods for modelling the temporal dynamics of diseases and the construction of monitoring systems, none have been applied to lung lesions data collected at slaughterhouses. Thus, this work aimed (1) to describe the temporal patterns of EP and CP-like lesions in Spain using time series methods to model the collected data on lung lesions at slaughterhouses; and (2) to construct and evaluate in quasi-real time a surveillance system for early detection of outbreaks and abnormal trends potentially related to both pathogens. In total, 16 time series were analysed including 3947 audits from 474 Spanish farms associated with 302 companies between 2016 and 2019. The monthly time series of the EP index between 2016 and 2019 in Spain (point estimate for Spain was −0.088 with an associated p = 0.073) and different Spanish subregions showed decreasing trend patterns (point estimates for Aragon was −0.028 with an associated p = 0.000 and for Catalonia was −0.064 with an associated p = 0.092), whereas the monthly time series of the CP index increased (point estimate for Spain was 0.004 with an associated p = 0.045 and for Aragon was 0.007 with an associated p = 0.000) over the same period. Additionally, the predictive performance of the estimated models was evaluated at quasi-real time using the data between 2020 and 2021. Results from this evaluation showed that overall, the selected models predicted the evolution of both the EP and CP indices in a reasonable manner being between 90 % prediction intervals. Therefore, time series models constructed in this work could be used to prevent and shorten the response time in implementing of control strategies against these respiratory pathogens minimizing their economic impact associated.