Andy is a lecturer on the Business Analytics MSc. His main research interests lie in the application of simulation and data analytics to assess risk and optimise production. His specialism lies in quantitative operational research methods, including Monte Carlo simulation, Machine Learning and Data Analysis.
Andy started out as a Risk Analyst at the Animal and Plant Health Agency, helping deliver the scientific evidence behind various UK policies and EU regulations in food safety and exotic disease control, working with Defra, the UK Food Standards Agency, the European Food Safety Authority and commercial companies such as Novartis and Unilever. Andy has also spent time as a Principal Modelling and Analysis Consultant at BAE Systems, delivering solutions to various problems including inventory forecasting and optimising processes for new digital technologies in forensics and warship energy management.
- Systems Analysis and Simulation
- Data Analytics
- Operational Research
Courses I teach on
Hill, A., Muñoz, V., Downes, J., Schuppers, M., Buncic, S., O'Brien, S. and Stärk, K.D.C. (2020), To Sample or Not to Sample? An Analysis of the Need for Salmonella Sampling of Smaller Poultry Processors. Risk Analysis. doi:10.1111/risa.13545
C Gavin, R Simons, A Berriman, D Moorhouse, E Snary, R Smith and A Hill (2018). A cost-benefit assessment of Salmonella-control strategies in pigs reared in the United Kingdom. Preventive Veterinary Medicine, 160:54-62.
Within the European Union (EU), microbiological criteria (MC) sampling for Salmonella in poultry was introduced in 2005. In particular, processors had to meet a target of fewer than seven positive samples out of 50. However, processors producing small amounts of poultry meat did not have to sample if national authorities determined this was an acceptable risk. The U.K. Food Standards Agency (FSA) thus has a sampling regime based on throughput that allows smaller processors not to sample. In 2011, the limit of 7/50 was reduced to 5/50. Given the current uncertainty regarding U.K. trade relations with the EU, the U.K. FSA decided to conduct a new risk assessment of the risks of Salmonella produced by smaller processors, to determine whether sampling was now necessary. Current evidence suggests that an MC sampling regime in smaller slaughterhouses is not warranted from a national public health perspective. Because of the insensitivities of the MC sampling scheme, the introduction of MC sampling into smaller slaughterhouses would only be necessary if the suspected carcass prevalence was 15% or more. While our analysis is prone to uncertainty, we estimated that the carcass prevalence in smaller processors is below this. Thus, we recommended that the current sampling framework, allowing smaller processors not to sample, was still applicable.
Foodborne infection is a result of exposure to complex, dynamic food systems. The efficiency of foodborne infection is driven by ongoing shifts in genetic machinery. Next-generation sequencing technologies can provide high-fidelity data about the genetics of a pathogen. However, food safety surveillance systems do not currently provide similar high-fidelity epidemiological metadata to associate with genetic data. As a consequence, it is rarely possible to transform genetic data into actionable knowledge that can be used to genuinely inform risk assessment or prevent outbreaks. Big data approaches are touted as a revolution in decision support, and pose a potentially attractive method for closing the gap between the fidelity of genetic and epidemiological metadata for food safety surveillance. We therefore developed a simple food chain model to investigate the potential benefits of combining 'big' data sources, including both genetic and high-fidelity epidemiological metadata. Our results suggest that, as for any surveillance system, the collected data must be relevant and characterize the important dynamics of a system if we are to properly understand risk: this suggests the need to carefully consider data curation, rather than the more ambitious claims of big data proponents that unstructured and unrelated data sources can be combined to generate consistent insight. Of interest is that the biggest influencers of foodborne infection risk were contamination load and processing temperature, not genotype. This suggests that understanding food chain dynamics would probably more effectively generate insight into foodborne risk than prescribing the hazard in ever more detail in terms of genotype.
An expert opinion workshop was held on the subject of the cause, identification and control of new and emerging Salmonella strains. Experts were invited to complete questionnaires, contribute to structured discussions and take part in cluster group tasks. Outputs of the workshop included that, with current surveillance methods, it might take up to 2.5 years from the first introduction of a new strain into the UK livestock population to its identification as a human epidemic strain. In order to reduce the time to detection and provide more effective control options, several recommendations were made, including better back-tracing of human cases to their source, which would require more effective communication between those responsible for human and veterinary surveillance.
The scientific understanding of the driving factors behind zoonotic and pandemic influenzas is hampered by complex interactions between viruses, animal hosts and humans. This complexity makes identifying influenza viruses of high zoonotic or pandemic risk, before they emerge from animal populations, extremely difficult and uncertain. As a first step towards assessing zoonotic risk of influenza, we demonstrate a risk assessment framework to assess the relative likelihood of influenza A viruses, circulating in animal populations, making the species jump into humans. The intention is that such a risk assessment framework could assist decision-makers to compare multiple influenza viruses for zoonotic potential and hence to develop appropriate strain-specific control measures. It also provides a first step towards showing proof of principle for an eventual pandemic risk model. We show that the spatial and temporal epidemiology is as important in assessing the risk of an influenza A species jump as understanding the innate molecular capability of the virus. We also demonstrate data deficiencies that need to be addressed in order to consistently combine both epidemiological and molecular virology data into a risk assessment framework. Â© 2015 The Authors.
As part of the evidence base for the development of national control plans for Salmonella spp. in pigs for E.U. Member States, a quantitative microbiological risk assessment was funded to support the scientific opinion required by the EC from the European Food Safety Authority. The main aim of the risk assessment was to assess the effectiveness of interventions implemented on-farm and at the abattoir in reducing human cases of pig meat–borne salmonellosis, and how the effects of these interventions may vary across E.U. Member States. Two case study Member States have been chosen to assess the effect of the interventions investigated. Reducing both breeding herd and slaughter pig prevalence were effective in achieving reductions in the number of expected human illnesses in both case study Member States. However, there is scarce evidence to suggest which specific on-farm interventions could achieve consistent reductions in either breeding herd or slaughter pig prevalence. Hypothetical reductions in feed contamination rates were important in reducing slaughter pig prevalence for the case study Member State where prevalence of infection was already low, but not for the high-prevalence case study. The most significant reductions were achieved by a 1- or 2-log decrease of Salmonella contamination of the carcass post-evisceration; a 1-log decrease in average contamination produced a 90% reduction in human illness. The intervention analyses suggest that abattoir intervention may be the most effective way to reduce human exposure to Salmonella spp. However, a combined farm/abattoir approach would likely have cumulative benefits. On-farm intervention is probably most effective at the breeding-herd level for highprevalence Member States; once infection in the breeding herd has been reduced to a low enough level, then feed and biosecurity measures would become increasingly more effective.
Previous modelling studies have estimated that between 1% and 10% of human salmonella infections are attributable to pig meat consumption. In response to this food safety threat the British pig industry have initiated a salmonella monitoring programme. It is anticipated that this programme will contribute to achieving a UK Food Standards Agency target for reducing salmonella levels in pigs at slaughter by 50% within 5 years. In order to better inform the monitoring programme, we have developed a stochastic transmission model for salmonella in a specialist grower-finisher pig herd, where data from a Danish longitudinal study have been used to estimate some of the key model parameters. The model estimates that about 17% of slaughter-age pigs will be infected with salmonella, and that of these infected pigs about 4% will be excreting the organism. In addition, the model shows that the most effective control strategies will be those that reduce between-pen transmission.
A farm‐to‐consumption quantitative microbiological risk assessment (QMRA) for Salmonella in pigs in the European Union has been developed for the European Food Safety Authority. The primary aim of the QMRA was to assess the impact of hypothetical reductions of slaughter‐pig prevalence and the impact of control measures on the risk of human Salmonella infection. A key consideration during the QMRA development was the characterization of variability between E.U. Member States (MSs), and therefore a generic MS model was developed that accounts for differences in pig production, slaughterhouse practices, and consumption patterns. To demonstrate the parameterization of the model, four case study MSs were selected that illustrate the variability in production of pork meat and products across MSs. For the case study MSs the average probability of illness was estimated to be between 1 in 100,000 and 1 in 10 million servings given consumption of one of the three product types considered (pork cuts, minced meat, and fermented ready‐to‐eat sausages). Further analyses of the farm‐to‐consumption QMRA suggest that the vast majority of human risk derives from infected pigs with a high concentration of Salmonella in their feces (≥104 CFU/g). Therefore, it is concluded that interventions should be focused on either decreasing the level of Salmonella in the feces of infected pigs, the introduction of a control step at the abattoir to reduce the transfer of feces to the exterior of the pig, or a control step to reduce the level of Salmonella on the carcass post‐evisceration.
Pork and pork products are a major source of human salmonellosis in the United Kingdom (UK). Despite a number of surveillance programmes, the prevalence of Salmonella in the UK slaughter pig population remains over 20%. Here, we present the results of a Cost-Benefit Analysis comparing five on-farm control strategies (where the cost is the cost of implementation and the benefits are the financial savings for both the human health and pig industries). The interventions considered were: wet feed, organic acids in feed, vaccination, enhanced cleaning and disinfection and movement of outdoor breeding units. The data originate from published papers and recent UK studies. The effectiveness was assessed by adapting a previous risk assessment, originally developed for the European Food Safety Authority. Using this method, none of the intervention strategies produced a net cost-benefit. Our results suggest that the cost of implementation outweighed the savings for all interventions, even if the effectiveness could be improved. Therefore, to achieve a net cost-benefit it is essential to reduce the cost of interventions. Analyses concluded that large cost reductions (up to 96%) would be required. Use of organic acids required the smallest reduction in cost (22.7%) to achieve a net cost benefit. Uncertainty analysis suggested that a small net gain might be possible, for some of the intervention measures. But this would imply that the model greatly underestimated some key parameters, which was considered unlikely. Areas of key uncertainty were identified as the under-reporting factor (i.e. the proportion of community cases of Salmonella) and the source attribution factor (i.e. the proportion of human Salmonella cases attributable to pork products).
The burden of Salmonella entering pig slaughterhouses across the European Union is considered a primary food safety concern. To assist E.U. member states with the development of national control plans, we have developed a farm transmission model applicable to all member states. It is an individual-based stochastic susceptible-infected model that takes into account four different sources of infection of pigs (sows, feed, external contaminants such as rodents, and new stock) and various management practices linked to Salmonella transmission/protection (housing, flooring, feed, all-in-all-out production). A novel development within the model is the assessment of dynamic shedding rates. The results of the model, parameterized for two case study member states (one high and one low prevalence) suggest that breeding herd prevalence is a strong indicator of slaughter pig prevalence. Until a member state's' breeding herd prevalence is brought below 10%, the sow will be the dominant source of infection to pigs raised for meat production; below this level of breeding herd prevalence, feed becomes the dominant force of infection.
In response to the European Food Safety Authority's wish to assess the reduction of human cases of salmonellosis by implementing control measures at different points in the farm-to-consumption chain for pork products, a quantitative microbiological risk assessment (QMRA) was developed. The model simulated the occurrence of Salmonella from the farm to consumption of pork cuts, minced meat, and fermented ready-to-eat sausage, respectively, and a dose-response model was used to estimate the probability of illness at consumption. The QMRA has a generic structure with a defined set of variables, whose values are changed according to the E.U. member state (MS) of interest. In this article we demonstrate the use of the QMRA in four MSs, representing different types of countries. The predicted probability of illness from the QMRA was between 1 in 100,000 and 1 in 10 million per serving across all three product types. Fermented ready-to-eat sausage imposed the highest probability of illness per serving in all countries, whereas the risks per serving of minced meat and pork chops were similar within each MS. For each of the products, the risk varied by a factor of 100 between the four MSs. The influence of lack of information for different variables was assessed by rerunning the model with alternative, more extreme, values. Out of the large number of uncertain variables, only a few of them have a strong influence on the probability of illness, in particular those describing the preparation at home and consumption.
The UK Food Standards Agency is currently funding research to build the evidence base for the modernisation of meat inspection. This includes an assessment of the risks to public health and animal health/welfare of moving to a visual-only post-mortem meat inspection (PMMI), where routine mandatory palpation and incision procedures are omitted. In this paper we present the results of a risk assessment for a change from current to visual-only PMMI for cattle, sheep/goats and farmed/wild deer.A large list of hazard/species pairings were assessed and prioritised by a process of hazard identification. Twelve hazard/species pairings were selected for full consideration within the final risk assessment. The results of the public health risk assessment indicated that all hazard/species pairings were Negligible with the exception of Cysticercus bovis in cattle, which was judged to be of low-medium increased risk for systems not conforming to criteria as laid down by EC Regulation 1244/2007, compared to systems that do conform to Regulations for visual-only PMMI.Most hazard/species pairings were concluded to pose a potential increased risk to animal health/welfare, including Mycobacterium bovis (very low - low increase in risk, but with considerable uncertainty), Fasciola hepatica (negligible - very low) and Cysticercus bovis (very low - low). Due to low feedback rates to farmers, the real risk to animal health/welfare for F.hepatica and C.bovis, including animals in non-conforming systems under visual-only PMMI, is probably negligible. That then leaves M.bovis as the only confirmed non-negligible animal health and welfare risk.
Summary: T e in the amount of 'omics' data available and in our ability to interpret those data. The aim of this paper was to consider how omics techniques can be used to improve and refine microbiological risk assessment, using dose-response models for RNA viruses, with particular reference to norovirus through the oral route as the case study. The dose-response model for initial infection in the gastrointestinal tract is broken down into the component steps at the molecular level and the feasibility of assigning probabilities to each step assessed. The molecular mechanisms are not sufficiently well understood at present to enable quantitative estimation of probabilities on the basis of omics data. At present, the great strength of gene sequence data appears to be in giving information on the distribution and proportion of susceptible genotypes (for example due to the presence of the appropriate pathogen-binding receptor) in the host population rather than in predicting specificities from the amino acid sequences concurrently obtained. The nature of the mutant spectrum in RNA viruses greatly complicates the application of omics approaches to the development of mechanistic dose-response models and prevents prediction of risks of disease progression (given infection has occurred) at the level of the individual host. However, molecular markers in the host and virus may enable more broad predictions to be made about the consequences of exposure in a population. In an alternative approach, comparing the results of deep sequencing of RNA viruses in the faeces/vomitus from donor humans with those from their infected recipients may enable direct estimates of the average probability of infection per virion to be made. Â© 2014 The Society for Applied Microbiology.
The transmission of pathogens across the interface between wildlife and livestock presents a challenge to the development of effective surveillance and control measures. Wild birds, especially waterbirds such as the Anseriformes and Charadriiformes are considered to be the natural hosts of Avian Influenza (AI), and are presumed to pose one of the most likely vectors for incursion of AI into European poultry flocks. We have developed a generic quantitative risk map, derived from the classical epidemiological risk equation, to describe the relative, spatial risk of disease incursion into poultry flocks via wild birds. We then assessed the risk for AI incursion into British flocks. The risk map suggests that the majority of AI incursion risk is highly clustered within certain areas of Britain, including in the east, the south west and the coastal north-west of England. The clustering of high risk areas concentrates total risk in a relatively small land area; the top 33% of cells contribute over 80% of total incursion risk. This suggests that targeted risk-based sampling in a relatively small geographical area could be a much more effective and cost-efficient approach than representative sampling. The generic nature of the risk map method, allows rapid updating and application to other diseases transmissible between wild birds and poultry.
Cryptosporidiosis caused by Cryptosporidium parvum infection is a major cause of enteric illness in man and there is a significant reservoir in animals, particularly young ruminant species. To preliminary assess the magnitude of the risk posed by contact with faeces produced by infected livestock, two microbiological risk assessments have been developed: one for the risk of human infection with C. parvum while camping on contaminated land recently grazed by infected suckler cattle and a comparable risk assessment for camping on land recently spread with contaminated cattle slurry. Using a worst-case scenario approach, the upper level of risk was estimated to be one infection in every 6211 person-visits for a camping event on land recently grazed by infected cattle. Translated into camping events of 100 persons, this risk estimate would most likely lead to zero (98% likelihood) or one infection (1% likelihood). The results for cattle slurry model are similar despite different pathways. Sensitivity analysis was conducted for the grazing cattle model only. This suggested that the time between grazing and camping was the most important control strategy, but increasing hand-washing frequency and the removal of cattle faeces before camping would also be beneficial. If the upper level of risk were to be judged unacceptable then further data would be required to more accurately estimate the risk of infection through these scenarios. Further research would also be required to assess the fraction of cases attributable to camping and/or environmental contact with Cryptosporidium oocysts.
The current system of post-mortem meat inspection, using typical macroscopic inspection techniques, is ineffective in identifying the most common foodborne illness hazards, e.g. Salmonella and Campylobacter. Therefore, there is a need to adopt a more appropriate, risk-based approach to meat inspection. One specific example of modifying traditional inspection techniques to represent a more cost-effective approach to meat inspection is the allowance in EC Regulation 854/2004 for pigs that have been reared under controlled housing conditions since weaning to only undergo a visual inspection. However, the definition of controlled housing excludes outdoor pig production, and hence the United Kingdom (UK) has yet to introduce this method of meat inspection into abattoirs because of the associated complications of having a large outdoor UK pig herd. Therefore, in the context of the UK Food Standards Agency's programme of work to modernise meat hygiene inspection, we have conducted a qualitative risk assessment to assess the comparative risks to public and animal health from allowing visual-only inspection of both indoor and outdoor pigs. In order for visual-only inspection to be of higher risk than traditional meat inspection, the sensitivity of detection of a hazard must significantly decrease under visual-only inspection. In addition, for outdoor pigs to pose a greater risk than indoor pigs, the hazard must be more prevalent in the former than the latter. From a large number of hazards originally identified as worthy of investigation, only one (porcine tuberculosis) was considered to be of significant public or animal health risk and would be less likely to be detected through visual-only inspection. Despite higher rates of porcine tuberculosis in outdoor pigs than indoor pigs, the relatively small number of additional heads/carcasses that are infected and would be missed by including outdoor pigs in visual-only meat inspection (compared to traditional meat inspection) would pose a negligible risk to public health and a negligible/very low risk to animal health/welfare. Hence, we concluded that the overall risk from all hazards to public health by transferring to a visual-only inspection method was assessed as negligible for all pigs.
Hill A, Crotta M, Wall B, Good L, O'Brien SJ, Guitian J. Towards an integrated food safety surveillance system: a simulation study to explore the potential of combining genomic and epidemiological metadata. R Soc Open Sci. 2017;4(3):160721. doi:10.1098/rsos.160721
R Kosmider, J Smith, S Gillings, L Snow, A Breed, R Irvine and A Hill (2016). Updated risk of H5N1 HPAI incursion to poultry in Great Britain via wild birds. Veterinary Record, doi:10.1136/vr.103700.
A Hill, R Simons, L Kelly, E Snary (2016). A farm transmission model for Salmonella in pigs applicable to EU Member States. Risk Analysis, 36(3): 461-481.
A Hill, A Swart, R Simons, H Vigre, A Calado Domingues, L Kelly, E Snary. (2016). Intervention analyses for Salmonella in pigs in the EU. Risk Analysis 36(3):546-560.
E Snary, R Simons, A Swart, H Vigre, A Coutinho Calado Domingues, E Evers and T Hald, A Hill (2016). Quantitative Microbiological Risk Assessment on Salmonella in Slaughter and Breeder pigs. Risk Analysis 36(3):437-449.
H Vigre, K Barfoed, A Swart, R Simons, A Hill, E Snary and T Hald (2016). Characterization of the human risk of Salmonellosis related to consumption of pork products in different E.U. countries based on a QMRA. Risk Analysis 36(3): 531-545.
A Hill, T Dewé, R Kosmider, S Von Dobschuetz, S Munoz, A Hanna, A Fusaro, M De Nardi, W Howard, K Stevens, L Kelly, A Havelaar and K Stärk (2015). Modelling the species jump: towards assessing the risk of human infection from novel avian influenzas. Royal Society Open Science 2: 150173.
P Gale, A Hill, L Kelly, J Bassett, P McClure, Y Le Mark and I Soumpasis (2014). Applications of omics approaches to the development of Microbiological Risk Assessment using RNA virus dose-response models as a case study. Journal of Applied Microbiology 117(6): 1537-1548.
A Hill, V Horigan, K Stärk, K Clarke, S O’Brien and S Buncic (2014). A qualitative risk assessment for visual-only post-mortem meat inspection of cattle, sheep, goats and farmed/wild deer. Food Control 38(4): 96-108.
A Hill, A Brouwer, N Donaldson, S Lambton, S Buncic and S Griffiths (2013). A risk and benefit assessment for visual-only meat inspection of indoor and outdoor pigs in the United Kingdom. Food Control 30(1): 255-264.
A Hill, P Nally, R Chalmers, G Pritchard and M Giles (2011). A quantitative risk assessment for zoonotic transmission of Cryptosporidium parvum infection attributable to recreational use of farmland. Zoonoses and Public Health 58(5): 323-333.
A Hill, E Snary, M Arnold, L Alban and A Cook (2008). Dynamics of Salmonella Transmission on a British Pig Grower-Finisher Farm: A Stochastic Model. Epidemiology and Infection 136(3): 320-333.