Using machine learning improves predictions of herd-level bovine tuberculosis breakdowns in Great Britain.

In the United Kingdom, despite decades of control efforts, bovine tuberculosis (bTB) has not been controlled and currently costs ~ £100 m annually. Critical in the failure of control efforts has been the lack of a sufficiently sensitive diagnostic test. Here we use machine learning (ML) to predict herd-level bTB breakdowns in Great Britain (GB) with the aim of improving herd-level diagnostic sensitivity. The results of routinely-collected herd-level tests were correlated with risk factor data. Four ML methods were independently trained with data from 2012-2014 including ~ 4700 positive herd-level test results annually. The best model's performance was compared to the observed sensitivity and specificity of the herd-level test calculated on the 2015 data resulting in an increased herd-level sensitivity from 61.3 to 67.6% (95% confidence interval (CI): 66.4-68.8%) and herd-level specificity from 90.5 to 92.3% (95% CI: 91.6-93.1%). This approach can improve predictive capability for herd-level bTB and support disease control.

Endemic foot and mouth disease: pastoral in-herd disease dynamics in sub-Saharan Africa.

Foot and mouth disease (FMD) burden disproportionally affects Africa where it is considered endemic. Smallholder livestock keepers experience significant losses due to disease, but the dynamics and mechanisms underlying persistence at the herd-level and beyond remain poorly understood. We address this knowledge gap using stochastic, compartmental modelling to explore FMD virus (FMDV) persistence, outbreak dynamics and disease burden in individual cattle herds within an endemic setting. Our analysis suggests repeated introduction of virus from outside the herd is required for long-term viral persistence, irrespective of carrier presence. Risk of new disease exposures resulting in significant secondary outbreaks is reduced by the presence of immune individuals giving rise to a period of reduced risk, the predicted duration of which suggests that multiple strains of FMDV are responsible for observed yearly herd-level outbreaks. Our analysis suggests management of population turnover could potentially reduce disease burden and deliberate infection of cattle, practiced by local livestock keepers in parts of Africa, has little effect on the duration of the reduced risk period but increases disease burden. This work suggests that FMD control should be implemented beyond individual herds but, in the interim, herd management may be used to reduced FMD impact to livestock keepers.

Inferring within-herd transmission parameters for African swine fever virus using mortality data from outbreaks in the Russian Federation.

Mortality data are routinely collected for many livestock and poultry species, and they are often used for epidemiological purposes, including estimating transmission parameters. In this study, we infer transmission rates for African swine fever virus (ASFV), an important transboundary disease of swine, using mortality data collected from nine pig herds in the Russian Federation with confirmed outbreaks of ASFV. Parameters in a stochastic model for the transmission of ASFV within a herd were estimated using approximate Bayesian computation. Estimates for the basic reproduction number varied amongst herds, ranging from 4.4 to 17.3. This was primarily a consequence of differences in transmission rate (range: 0.7-2.2), but also differences in the mean infectious period (range: 4.5-8.3 days). We also found differences amongst herds in the mean latent period (range: 5.8-9.7 days). Furthermore, our results suggest that ASFV could be circulating in a herd for several weeks before a substantial increase in mortality is observed in a herd, limiting the usefulness of mortality data as a means of early detection of an outbreak. However, our results also show that mortality data are a potential source of data from which to infer transmission parameters, at least for diseases which cause high mortality.

Contact patterns as a risk factor for bovine tuberculosis infection in a free-living adult brushtail possum Trichosurus vulpecula population.

The aim of this study was to identify risk factors for bovine tuberculosis (TB) in a free-roaming, capture-mark-recapture monitored possum Trichosurus vulpecula population in a 22-ha study site at Castlepoint, New Zealand from 1 April 1989 to 31 March 1994. A matched case-control design was used to evaluate the influence of sex, habitat and contact opportunities on TB risk. Cases comprised possums identified as TB-positive throughout the study period. Controls were selected from the group of possums that were captured and showed no clinical signs of TB throughout the study period. Measures derived from a social network analysis of possum capture locations such as degree, clustering coefficient (CC) and betweenness were used to represent potential contact opportunities among possums. Network analysis measures recorded for individual possums in the 12-month period before a diagnosis of TB were evaluated in a conditional logistic regression model. We found no evidence of an association between case status and the total number of possums with which there was potential contact (degree) (P=0.5). The odds of cases being exposed to unit increases in the number of TB-positive contacts was 2.50 (95% CI 1.24-5.05; P<0.01) times that of controls. This effect was conditional on the total number of potential contacts made, with a negative interaction with increasing degree. These findings indicate that potential contact with TB-positive possums increases the odds of disease whereas potential contact with large numbers of possums does not. This suggests that multiple contacts with TB-positive possum(s) are necessary for transmission of TB and this is more likely to occur in networks that are smaller. We challenge the hypothesis that contact with large numbers of individuals increases the probability of becoming TB infected and argue that individual contact behaviour is a determinant of the creation of TB foci within free-living possum populations.

Analysis of Amblyomma surveillance data in the Caribbean: lessons for future control programmes.

Amblyomma variegatum, the Tropical Bont Tick (TBT), is the principal vector of heartwater and is associated with dermatophilosis, major causes of losses in animal production and mortality in Caribbean livestock. From 1995 to 2007, the Caribbean Amblyomma Programme (CAP) supported treatment and surveillance activities in 11 islands of the Eastern Caribbean with an initial objective of eradicating TBT. In addition to control activities, surveillance data were collected between 1997 and 2006 in a unique regional database. We report the analysis of the surveillance data from four islands (Nevis, St Kitts, St Lucia, and Barbados) where control and surveillance followed the initial protocol and where enough data were collected. We describe the evolution of TBT infestation levels and the efforts carried out throughout the surveillance period. Logistic regression identified factors associated with herds found infested with TBT. Overall, treatment programmes were associated with a decrease in proportion of TBT-infested farms. High surveillance efforts were carried out throughout the 1997-2007 period for all island of interest, but inadequate level of surveillance was observed in several quarters especially for St Kitts. Third quarter of the year, as indication of adult TBT seasonality on livestock, was significantly associated with the risk of detecting TBT in Nevis and St Kitts livestock farms. Also, presence of cattle in Nevis farms was shown associated with an increasing probability of farms being declared TBT-infested. Outcomes of these analyses provide basis for recommendations to improve future national and regional control and surveillance activities. This analysis demonstrates the usefulness of long term and adequate surveillance data for control programmes and identification of factors associated with risk of having infested herds.

Use of a mathematical model to study the dynamics of Ctenocephalides felis populations in the home environment and the impact of various control measures.

The biology of fleas has been studied by a number of authors, as has the impact of various types of control measures. However, there are no mathematical models simulating the dynamics of a population of Ctenocephalides felis felis fleas on their host (the cat) and in their close environment (apartment). The model presented in this paper allows for integration of the numerous biological and behavioural parameters of the parasites and their hosts and for the variation of these same parameters. The various types of control measures can be programmed so that their impact over time can be studied. The model confirms the key role played by adult fleas, or emerged fleas contained in the cocoon. Only regular applications of persistent insecticides to the host animal will enable control of the parasite population. A combination of these insecticides with an IGR (Insect Growth Regulator) will accelerate decontamination of the home environment and see the disappearance of the parasites altogether if they are not reintroduced. The association of additional measures such as vacuum cleaning will accelerate the process of decontamination but will have no impact if carried out in isolation. One-off treatment with insecticide will not enable a reduction in the parasite population, even if carried out frequently. Use of insecticides on the home environment premises alone does not appear to be an adequate means of control. The present model can be used to test various integrated control measures which take into account different factors such as the number of host animals, the frequency of movement outdoors, the impact of the seasons.