Data collection

In total, 665 farms were included in the case-control study. Nine control farms, including three backyard farms and six commercial farms, were diagnosed with ASF less than 28 days after being interviewed as controls. These farms were therefore interviewed (again) as outbreak farms, and excluded as controls, as we could not exclude the possibility that the farms might have already been infected at the time of the first interview. Furthermore, one farm did not have pigs present at the day of the visit and was therefore excluded from the analyses.

All farms included in the study were contacted by an official veterinarian (OV) and invited to participate in the study. The OVs visited all study farms and filled out a questionnaire regarding potential risk factors related to management and biosecurity measures implemented on the farm. The questionnaire consisted of 42 questions with the following themes: number and age groups of animals (6 questions), practice of slaughter, outdoor access and other species (5 questions), whether there were wild boar around the farm area (6 questions), feed and water (8 questions), indirect contacts, i. e. vehicles and visitors (3 questions), bedding, manure and fencing (4 questions), contacts to other farms in the high-risk period (6 questions) and observations of ticks, mosquitoes and midges (4 questions). The full questionnaire is available in the Supplementary Materials. Each interview lasted approximately 30 minutes for backyard and 60–90 minutes for commercial holdings. Some questions were related to the high-risk period, interpreted as the time period in which ASF introduction presumably occurred. From outbreaks in other countries, very short time periods between ASF introduction and detection have been estimated, e.g. 2–4 weeks in Latvia⁶, 5–14 days in Russia in 2008–2012²⁹, and 1 week from first clinical sign to diagnosis in Estonia⁴. For backyard farms, the high-risk period was therefore defined as 2 weeks, while for commercial farms, this period was defined as 6 weeks, based on mortality data from a very large commercial farm (data not published). Given their intermediate herd size, the high-risk period for type A farms was defined as 4 weeks. For outbreak farms this period, was calculated as 2, 4 or 6 weeks prior to the date of initial ASF detection. For control farms, the same time period as in the matching outbreak farm was used, i.e. starting from the detection date of the outbreak farm and counting backwards for 2, 4 or 6 weeks.

Based on data from the EU Animal Disease Notification System (ADNS), geographic coordinates of all outbreaks in domestic pigs and cases of ASF in wild boar in Romania and neighbouring countries were extracted. For both outbreak and control farms, the distances to other outbreaks during the high-risk period were calculated. With these calculated distances, some covariates were defined to be used in the final logistic regression model namely i) the logarithm of the distance to the nearest wild boar (WB) case, ii) the logarithm of the distance to the nearest outbreak in domestic pigs (DP) and iii) the total numbers of cases in wild boar or outbreaks in domestic pigs, respectively, within 1, 2, 5 and 10 km. Additionally, data on the density of domestic pigs, based on data provided by the National Sanitary Veterinary and Food Safety Authority of Romania, were calculated. The number of pigs per hunting ground and the relative wild boar abundance were calculated based on the numbers of hunted wild boar divided by the size (in km₂) of the hunting ground in the hunting season 2018–19. Hunting data were provided by the Romanian Ministry of Waters and Forests to calculate the relative wild boar abundance as the number of hunted wild boar per hunting ground. The percentage of forest cover in each hunting ground unit, in which outbreak and control farms were located, was calculated. The presence of water bodies within 1 km of the outbreak or control farms was determined as a binomial variable (yes/no). The percentage of forest cover and presence of water bodies were each extracted from the 2018 raster version of CORINE Land Cover data.