Use of spatiotemporal analysis of laboratory submission data to identify potential outbreaks of new or emerging diseases in cattle in Great Britain

Abstract

BackgroundNew and emerging diseases of livestock may impact animal welfare, trade and public health. Early detection of outbreaks can reduce the impact of these diseases by triggering control measures that limit the number of cases that occur. The aim of this study was to investigate whether prospective spatiotemporal methods could be used to identify outbreaks of new and emerging diseases in scanning surveillance data. SaTScan was used to identify clusters of unusually high levels of submissions where a diagnosis could not be reached (DNR) using different probability models and baselines. The clusters detected were subjected to a further selection process to reduce the number of false positives and a more detailed epidemiological analysis to ascertain whether they were likely to represent real outbreaks.Results187,925 submissions of clinical material from cattle were made to the Regional Laboratory of the Veterinary Laboratories Agency (VLA) between 2002 and 2007, and the results were stored on the VLA FarmFile database. 16,925 of these were classified as DNRs and included in the analyses. Variation in the number and proportion of DNRs was found between syndromes and regions, so a spatiotemporal analysis for each DNR syndrome was done. Six clusters were identified using the Bernoulli model after applying selection criteria (e.g. size of cluster). The further epidemiological analysis revealed that one of the systemic clusters could plausibly have been due to Johne's disease. The remainder were either due to misclassification or not consistent with a single diagnosis.ConclusionsOur analyses have demonstrated that spatiotemporal methods can be used to detect clusters of new or emerging diseases, identify clusters of known diseases that may not have been diagnosed and identify misclassification in the data, and highlighted the impact of data quality on the ability to detect outbreaks. Spatiotemporal methods should be used alongside current temporal methods for analysis of scanning surveillance data. These statistical analyses should be followed by further investigation of possible outbreaks to determine whether cases have common features suggesting that these are likely to represent real outbreaks, or whether issues with the collection or processing of information have resulted in false positives.

European outbreaks of atypical myopathy in grazing equids (2006–2009): Spatiotemporal distribution, history and clinical features

“…pasture slope, should also be recognised and the fact that data were obtained from a large number of veterinary surgeons and horse owners/carers with resultant potential biases. Alternative methods that could be used in future studies to investigate the spatiotemporal clustering of AM cases include k function analysis [18,19], the Biased Sample Hospital‐based Area Disease Estimation (also called the B‐SHADE technique) [20] and the use of SaTScan software (SaTScan software for the spatial and space‐time scan statistics) b [21]. Despite these limitations, this study contains important information for the future management of cases.…”

Prospective study of avian influenza H9 infection in commercial poultry farms of Punjab Province and Islamabad Capital Territory, Pakistan

“…All laboratory-confirmed cases of H9N2 between December 2013 to February 2014 were geocoded to street addresses. SatScan software version 9.1.1 developed by Martin Kulldorff, Havard Medical School (Boston, USA) (Hyder et al 2011). This method consists of thousands of cylinders that move across space and/or time.…”

Retrospective space-time cluster analysis of whooping cough re-emergence in Barcelona, Spain, 2000-2011

“…These methods may also have the potential to detect clusters of new or emerging disease. However, not all clusters detected using statistical analysis will represent real outbreaks and further epidemiological investigation is required to determine whether these statistical clusters represent real outbreaks (Hyder et al, 2011).…”