Hostname: page-component-7b7567c8c-7pwfn Total loading time: 0 Render date: 2026-11-12T09:04:28.115Z Has data issue: false hasContentIssue false
  • English
  • Français

Developing a risk management framework to improve public health outcomes by enumerating Salmonella in ground turkey – RETRACTION

Published online by Cambridge University Press:  06 December 2018

F. SampedroOpen the ORCID record for F. Sampedro[Opens in a new window],
S. J. Wells,
J. B. Bender and
C. W. Hedberg
F. Sampedro
Affiliation:
Center for Animal Health and Food Safety, College of Veterinary Medicine, University of Minnesota, Saint Paul, USA
S. J. Wells
Affiliation:
Veterinary Population Medicine, College of Veterinary Medicine, University of Minnesota, Saint Paul, USA
J. B. Bender
Affiliation:
Center for Animal Health and Food Safety, College of Veterinary Medicine, University of Minnesota, Saint Paul, USAEnvironmental Health Sciences, School of Public Health, University of Minnesota, Saint Paul, USA
C. W. Hedberg*
Affiliation:
Environmental Health Sciences, School of Public Health, University of Minnesota, Saint Paul, USA
*
Author for correspondence: C. W. Hedberg, E-mail: hedbe005@umn.edu
Rights & Permissions [Opens in a new window]

Abstract

Salmonella spp. continue to be a leading cause of foodborne morbidity worldwide. To assess the risk of foodborne disease, current national regulatory schemes focus on prevalence estimates of Salmonella and other pathogens. The role of pathogen quantification as a risk management measure and its impact on public health is not well understood. To address this information gap, a quantitative risk assessment model was developed to evaluate the impact of pathogen enumeration strategies on public health after consumption of contaminated ground turkey in the USA. Public health impact was evaluated by using several dose–response models for high- and low-virulent strains to account for potential under- or overestimation of human health impacts. The model predicted 2705–21 099 illnesses that would result in 93–727 reported cases of salmonellosis. Sensitivity analysis predicted cooking an unthawed product at home as the riskiest consumption scenario and microbial concentration the most influential input on the incidence of human illnesses. Model results indicated that removing ground turkey lots exceeding contamination levels of 1 MPN/g and 1 MPN in 25 g would decrease the median number of illnesses by 86–94% and 99%, respectively. For a single production lot, contamination levels higher than 1 MPN/g would be needed to result in a reported case to public health officials. At contamination levels of 10 MPN/g, there would be a 13% chance of detecting an outbreak, and at 100 MPN/g, the likelihood of detecting an outbreak increases to 41%. Based on these model prediction results, risk management strategies should incorporate pathogen enumeration. This would have a direct impact on illness incidence linking public health outcomes with measurable food safety objectives.

Keywords

Ground turkeypathogen enumerationrisk assessmentsalmonella

Information

Type
Original Paper
Information
Epidemiology & Infection , Volume 147 , 2019, e69
Check for updates
Creative Commons
Creative Common License - CCCreative Common License - BY
This is an Open Access article, distributed under the terms of the Creative Commons Attribution licence (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted re-use, distribution, and reproduction in any medium, provided the original work is properly cited.
Copyright
Copyright © The Author(s) 2018

Introduction

Non-typhoidal Salmonella species are responsible for an estimated 1.2 million illnesses, 23 000 hospitalisations, 450 deaths and approximately $365 million in direct medical costs annually in the USA [Reference Scallan1]. Rates of Salmonella cases in the USA (15.9 culture-confirmed cases per 100 000 population in 2015) have not appreciably declined over the past 15 years [Reference Marder2].

Poultry and poultry meat products are considered some of the main carriers of the organism and represent a significant share of the attributed sources of salmonellosis in humans. Approximately 33% of all food-related salmonellosis cases were associated with meat products regulated by the US Department of Agriculture's Food Safety and Inspection Service (FSIS) [Reference Painter3]. Out of those, poultry represented about 58% of the cases, with 85% being associated specifically with chicken [4]. Although not specifically estimated, one can assume that most of the other human cases related to poultry could be attributed to turkey consumption.

The characterisation of the occurrence of Salmonella spp. along the poultry production chain has been an area of research and policy focus for many years. Joint efforts between policymakers, poultry producers and industry have reduced the overall Salmonella prevalence in poultry products and, as a result, reduced the public health burden associated to their consumption. These efforts have reduced the prevalence of Salmonella spp. in ground turkey from 36.6% in 1998 to 15.2% in 2013 [5]. FSIS has recently modified the Salmonella spp. performance standards for ground turkey allowing contamination rates no greater than 13.5% over a 52-week moving window test period [5]. This new standard is in alignment with the Healthy People 2020 initiative by the US Department of Health and Human Services to achieve a 25% reduction in human illnesses attributed to Salmonella spp. in poultry products [6]. Despite these efforts, five Salmonella outbreaks attributed to ground poultry products have been reported in the USA since 1998. Recent multistate outbreaks of human salmonellosis have led to hundreds of reported illnesses and the recall of millions of pounds of raw ground meat and poultry products [Reference Routh7].

While most of the regulatory efforts have been focused on reducing the overall prevalence of Salmonella spp. (presence or absence), very little has been done to estimate the impact on public health by reducing the actual concentration of Salmonella in positive lots. As suggested by other authors, there is already evidence from microbiological risk assessment studies that levels of contamination can be even more important to public health than prevalence as they are directly related to the likelihood that the ingested dose exceeds the minimum infectious dose needed for disease development [Reference Mead8, Reference McEntire9]. There is a need to test new performance standards that are based on prevalence and enumeration levels rather than just on absence or presence alone.

Dose–response models have been developed to estimate the relationship between the probability of illness and the ingested Salmonella spp. dose in a food product. Models, based on the β-Poisson distribution, have been developed through use of volunteer and mouse feeding trials (a repository of dose–response models has been created by the Center for Advancing Microbial Risk Assessment (CAMRA), http://camra.msu.edu/) and outbreak data [10]. Recently, a two-tiered approach has been proposed to model separately the probability of infection and probability of illness using a hypergeometric β-Poisson model using data from Salmonella outbreaks [Reference Teunis11]. As pointed out by the authors, dose–response models from feeding trials seem to underestimate the risk of illness by using laboratory-adapted or low-virulence strains not involved in outbreaks. In contrast, dose–response models from outbreak data may overestimate the risk of illness by only accounting for high-virulence serotypes capable of producing outbreaks and ignoring the fact that a high proportion of Salmonella serotypes found in food are rarely involved in outbreaks, indicating a low-virulence profile. A combined approach using two different dose–response relationships for high- and low-virulence serotypes could better estimate the true incidence of Salmonella human cases due to the consumption of contaminated ground turkey.

The main objective of this study was to develop a risk management framework for Salmonella spp. in ground turkey based on the evaluation of public health risk associated with varying concentrations of Salmonella contamination.

Methodology

Input data

Prevalence and concentration levels of Salmonella spp. in ground turkey (2011–2016) were obtained from FSIS samples collected at processor and retail locations throughout the USA through a FOIA request (FOIA 2017-00288) (Table 1). Salmonella serotypes found in ground turkey were classified as low- and high-virulence depending on their involvement in human outbreaks [12] (Table 2). Data available from the scientific literature and provided by industry through personal communication were gathered to characterise the model inputs (Tables 2–4). Figure 1 shows the two main consumption scenarios evaluated separately in the risk assessment model (home and restaurant setting). For each one, consumer handling and cooking practices of ground turkey were used (in some cases, ground turkey data were not available and ground beef data were used instead). Fresh and frozen states were taken into account to estimate the impact of thawing methods (microwave, room temperature and refrigerator) on the final Salmonella concentration in the home consumption scenario. Current undercooking practices among US consumers at home and restaurant settings [Reference Phang and Bruhn13, Reference Bogard14] were modelled to estimate the ingested dose of Salmonella spp. at the time of consumption [Reference Murphy15, Reference Manios and Skandamis16]. No cross-contamination was modelled in the current scenarios to account solely for the direct effect of proposed risk mitigations strategies.

Fig. 1.

Risk assessment model framework. (a) Thawing and cooking scenarios. (b) Dose–response approach.

Table 1.

Summary of FSIS data on prevalence and concentration of Salmonella spp. in ground turkey

a Excluding mechanically separated meat.

b Mean and standard deviation.

Table 2.

Model inputs on Salmonella spp. prevalence and concentration levels in ground turkey in the USA

a High-virulent Salmonella serotypes as implicated in 2002–2012 outbreaks: Heidelberg, III_18:z4, z23 (Enterititidis), Saintpaul, I4,[5],12:I:-, Muenchen, Newport, Typhimurium, Montevideo, Infantis, Javiana, Anatum, Agona, Berta. Low-virulent Salmonella serotypes as not implicated in outbreaks: Schwarzengrund, Reading, Kentucky, Worthington, Hadar. CDC, 2014.

Table 3.

Model inputs on population and Salmonella dose–response data

Table 4.

Model input data on consumption patterns at home

a Same equations were used to estimate the log reduction (log CFU/g) after cooking for the rest of scenarios.

Dose–response relationship

Dose–response models have been published using volunteer and mouse feeding trials and outbreak data to characterise the relationship between the dose ingested and probability of illness. The first approach has been to use data from feeding trials to characterise the dose–response relationship. The CAMRA has created a repository of dose–response models by fitting the exponential and β-Poisson models to feeding trials data to find the optimal model. They have estimated dose–response parameters for several Salmonella serotypes, namely Salmonella Anatum, Salmonella Meleagridis, Salmonella Newport and generic non-typhoidal Salmonella.

The second approach has been to use outbreak data that relate the dose ingested with the attack rate. A recent risk assessment used a β-Poisson model to estimate the number of human cases using Salmonella outbreak data in chicken meat and egg products (Table 3) [10]. More recently, outbreak data with all products combined were used to characterise the dose–response relationship by using a Poisson γ mixture distribution [Reference Teunis11]. The authors used a two-tiered approach by estimating first the probability of infection (P inf) and then the conditional probability of the illness given infection (P ill/inf) and finally the probability of illness (P ill) as the product of both probabilities (P inf × P (ill/inf)) (Table 3).

In this study, a combined approach selected separate dose–response models for high- and low-virulence strains to avoid over- and underestimation. For high-virulence strains, three different dose–response models were used: (i) WHO and FAO β-Poisson single hit model; (ii) two-tiered approach suggested by [Reference Teunis11]; and (iii) β-Poisson model for Salmonella Typhimurium estimated by the CAMRA using data published by [Reference Meynell and Meynell17] in mice. Low-virulence strains were characterised by a β-Poisson model corresponding to Salmonella Anatum estimated by CAMRA using data from volunteer feeding trials [Reference McCullough and Elsele18].

Risk assessment model framework

A risk assessment model was developed to estimate key public health metrics (mean risk of illness, predicted number of cases and reported cases) in the US population after consumption of contaminated ground turkey at home and restaurant settings (Fig. 1 and Tables 2–5). Data on cooking practices for ground turkey were not available, thus data on ground beef were used instead assuming a similar thermal profile. Cooking practices observed at the restaurant level were safer (reaching or surpassing FDA and USDA recommended cooking temperatures more frequently) than those reported at home [Reference Phang and Bruhn13, Reference Bogard14]. Several dose–response approaches were evaluated using different models for high-virulence serotypes (Fig. 1 and Table 3). The stochastic risk assessment model was developed using Excel and @Risk 6.3 (Palisade Corp., NY, USA). Model outputs were obtained by Monte Carlo simulation techniques for 100 000 iterations. During each iteration, a Latin Hypercube sampling technique was used to select one random value of each variable or parameter from its respective distribution. Output simulation curves showed a positive skewness and high kurtosis values indicating a long tail to the right (higher values) due to the inherent variability and uncertainty of some of the input variables. This fact indicated the suitability of expressing the results as the median and 90% confidence intervals. A sensitivity analysis was carried out by a tornado graph showing the effects of changing the input variable values over the output (predicted number of illnesses).

Table 5.

Model input data on consumption patterns at the restaurant

Public health impact of different risk management strategies

The baseline scenario was compared with different risk management strategies by assuming the enumeration of Salmonella spp. on every positive lot of ground turkey (2000 pounds/907 kg) and removing high-contaminated lots (>1 MPN/g) or following FSIS guidelines of absence of Salmonella spp. in 25 g and removing lots with ⩾1 MPN per 25 g from the production chain. The baseline model and the different scenarios were run and compared by the impact on the public health metrics (mean risk of illness, number of illnesses and reported cases) (Table 6).

Table 6.

Model outputs by using different dose–response models

a Median and 90% CI.

Effect of level of contamination on outbreak detection

Salmonella is a nationally reportable disease, and isolates are routinely submitted to Public Health Laboratories for molecular characterisation by pulsed-field gel electrophoresis (PFGE) or whole genome sequencing (WGS). Individual cases are routinely interviewed to identify risk factors, and when multiple cases are linked by PFGE or WGS, the case cluster is investigated as a possible outbreak. The likelihood of identifying a common source increases with the number of cases in the cluster and the cluster investigation methods. In a FoodNET cluster investigation study, an outbreak source was identified for 8% of clusters involving two cases, 13% of clusters involving 3–5 cases, 19% of clusters involving 6–10 cases and 41% of outbreaks involving more than 10 cases [Reference Jones19].

The public health impact (predicted number of illnesses, reported cases and likelihood of outbreak detection) of a single contaminated lot of ground turkey was estimated assuming different contamination levels (−1 to 2 log MPN/g) to identify the level that would produce at least one reported case by using the feeding trial (Fig. 3a) and outbreak dose–response models (Fig. 3b).

Results and discussion

Overall prevalence, microbial load and serotype distribution in ground turkey

The mean prevalence of Salmonella spp. in ground turkey was slightly over 10% during the 2010–2016 period (Table 1). The mean Salmonella concentration was 1.8 (95% CI 0.016–137) MPN/g. Half of the samples had a concentration lower than 1 MPN/g, whereas 37% were between 1–10 MPN/g, 16% >10 MPN/g and 2% of the samples >100 MPN/g. This is consistent with other observations that higher concentrations of contamination may be associated with a relatively few farm production sources [Reference Cui20].

The median ingested dose at home and restaurant settings was 5.4 (95% CI 0–333) and 0.5 (95% CI 0–253) MPN, respectively. The median dose from both consumption scenarios was lower than that reported from outbreak data as [10] estimated an ID50 of 39.6–55.5 CFU from all Salmonella outbreaks combined and the ingested dose estimated by [Reference Cui20] from Salmonella Typhimurium 4,5,12:i:− in beef burgers was 315 MPN (142–685, 95% CI).

Serotype distribution in the FSIS 2010–2016 database included Heidelberg (12.6%), Saintpaul (9.9%), III_18 (5.2%), Muenchen (2.5%), Berta (1.4%), Newport (1.3%), Agona (1.3%), Typhimurium (0.7%), Anatum (0.7%), Montevideo (0.5%) and Enteritidis (0.4%). Sixty-three per cent of the positive samples contained serotypes rarely involved in human outbreaks as outlined in the CDC National Salmonella Surveillance Annual Report [12]. FSIS data showed highly contaminated samples corresponding to serotypes of Agona, Berta, Hadar, Heidelberg and Saintpaul. The high prevalence and concentration observed in S. Heidelberg may explain the previous outbreaks occurred in ground turkey in the USA involving that serotype [Reference Routh7].

Predicted illnesses and reported cases under different dose–response approaches

The median number of Salmonella illnesses ranged from 2705 to 21 099 by using different dose–response models (Table 6). The estimated median number of Salmonella cases reported to public health officials was 93–727. There were no observed differences in the number of predicted illnesses between the two models based on outbreak data (chicken and egg products vs. all products combined) for the high-virulence strains (Table 6). This can be explained by the fact that most of the outbreaks are caused by S. Enteritidis and S. Typhimurium which are also the most predominant serotypes found in egg and chicken meat products [10, Reference Teunis11]. This dose–response relationship may change in the future as more serotypes are involved in outbreaks that have different attack rates. Median number of cases using the model based on mouse feeding trial data for high-virulence strains was 10 ×  lower indicating a potential lower range prediction in the incidence of Salmonella-related cases.

Assuming that 15% of all poultry-attributed non-typhoidal Salmonella cases in the USA could be attributable to turkey corresponds to 34 500 estimated illnesses [Reference Painter3, 4]. Our estimates for ground turkey-related illnesses range from 8% of these illnesses based on the feeding trial model to 57% based on the outbreak-based models. The dispersion seen in the attributed cases predicted by the three dose–response modelling approaches can be explained by the limitations of each approach in under- or overestimating the risk of illness. A combined approach where outbreak-based and feeding trial-based models are combined for high- and low-virulence serotypes seems to better represent the true incidence of Salmonella human cases.

The high dispersion and wide confidence intervals were due to the inherent variability and uncertainty shown in the model inputs. The model proposed by [Reference Teunis11] showed narrower confidence intervals and lower dispersion and thus was selected among the two outbreak-based models to further compare the effect of different risk mitigation strategies.

Public health impact of risk management strategies

The sensitivity analysis highlighted distribution of cooking temperatures and microbial concentration as the most influential input variables on the number of illnesses predicted, which may be expected as concentration and cooking are intimately related to the ingested dose (Fig. 2). A reduction in the mean concentration of the pathogen in the food portions consumed would have a direct impact on public health. This is in agreement with previous risk assessment studies of Salmonella in poultry and pork products [10, Reference Straver21, Reference Snary22], indicating that concentration of contamination and consumer education on adequate handling and cooking practices have the greatest impact on public health, and efforts at any stage of production or processing that reduces the level of Salmonella on the end product will reduce risk to a greater extent. A better understanding of cooking practices for ground turkey (or chicken) will improve the estimate of the influence of this variable on public health.

Fig. 2.

Sensitivity analysis of the influence of the model inputs on the predicted number of illnesses.

The proportion of high-virulence serotypes and Salmonella spp. prevalence in ground turkey had a negligible effect on the predicted number of cases. The distribution of cooking temperatures at home and cooking time in the frozen scenario were negatively correlated (an increase in the cooking temperatures or time will decrease the number of illnesses) (Fig. 2). The mean risk of illness (probability of illness after a single exposure) after eating in a restaurant was 10 ×  lower than the one predicted at home due to these differences in cooking. A very small fraction of consumers (2.2%) reported cooking an unthawed food portion [Reference Phang and Bruhn13], yet, this scenario represented the highest mean risk of illness accounting for 38–52% of the total number of illnesses at home (Table 6). Better consumer education on thawing practices and the use of digital thermometers would decrease the overall risk of illness.

FSIS data showed 53% of positive samples enumerated contained a concentration higher than 1 MPN/g. Removing lots with a microbial load >1 MPN/g reduced the overall prevalence to 5.6% and mean concentration to 0.2 (95% CI 0.018–0.50) MPN/g (Table 7). Through removal of these samples above the 1 MPN/g threshold, the model predicted 159–2572 illnesses that would lead to 6–89 reported cases, which is a reduction of 86–94% and 84–96.5% in the median and upper limit of the confidence interval number of illnesses, respectively. This far exceeds the proportionate share of the Healthy People 2020 goal for a 25% reduction in Salmonella associated with poultry products. Absence of Salmonella spp. in 25 g criteria would remove 90% of the FSIS-positive samples enumerated in the 2011–2016 period. Through removal of samples above the 1 MPN/25 g threshold, the model predicts that the prevalence of contamination would be reduced to 1.19% and the mean concentration to 0.035 (95% CI 0.032–0.038) MPN/g, resulting in a net reduction of 99% of illnesses (6–110) and reported cases (0–4). Although the absence of Salmonella in 25 g guideline would reduce the number of illnesses, it would also reduce the availability of fresh ground turkey. An estimated 10.7% of production lots would need to be diverted to a thermal process, or potentially wasted, whereas only 6.3% would be removed or diverted if 1 MPN/g guidelines were used (Table 7).

Table 7.

Public health impact by using different risk management enumeration strategies

a Median values and 90% confidence intervals.

Effect of level of contamination on outbreak detection

Industry usually uses 2000 lb (907 kg) lots of ground turkey in their daily production schemes in the USA (Industry personal communication). Figure 3 shows the number of illnesses, reported cases and likelihood of outbreak detection at various contamination levels of a single contaminated lot using the two dose–response approaches (feeding trial 2A and outbreak data 2B). The mean Salmonella concentration in a lot needs to be higher than 1 MPN/g to produce at least one reported Salmonella case to public officials (Fig. 3b). This single case would not stimulate a cluster or outbreak investigation. At contamination levels of 10 MPN/g, there would be 13% chance of detecting an outbreak, and at 100 MPN/g, the likelihood of detecting an outbreak increases to 41%, based on the outbreak dose–response model (Fig. 3b). These data suggest that a high proportion of outbreaks and poultry-associated sporadic infections are attributable to products with relatively high levels of Salmonella spp. contamination.

Fig. 3.

Number of illnesses and reported cases predicted by the feeding trial dose–response model (a) and the outbreak data dose–response model (b) at various contamination levels of a single 2000 lb lot of contaminated ground turkey. Shaded bars correspond to predicted illnesses. Black bars correspond to reported cases. Dotted lines correspond to the mean Salmonella spp. concentration in a positive lot (log MPN/g). Percentages correspond to the probability of a later investigation identifying an outbreak source.

Conclusions

Different dose–response approaches based on feeding trial and outbreak data models for high- and low-virulent serotypes were used to estimate the Salmonella annual number of illnesses and reported cases from consumption of contaminated ground turkey. This combined approach allowed balancing under- and over-reporting. Removing highly contaminated lots reduced the occurrence of illnesses and the notifiable number of outbreaks. Risk management strategies focused on interventions that can reduce Salmonella spp. load to low levels of contamination will have great public health benefits while avoiding costs associated with the destruction of products with detectable but low levels. Currently, proposed regulatory practices do not consider dose–response, and are instead focused on the presence compared with the absence of pathogen. Ideally, regulatory efforts in food safety should link public health metrics with quantifiable food safety metrics based on the results of a risk assessment [Reference Guillier23].

Supplementary material

The supplementary material for this article can be found at https://doi.org/10.1017/S095026881800328X.

Acknowledgements

This project was funded by the MnDRIVE Global Food Ventures grant ‘Developing a Risk Management Framework to Improve Public Health Outcomes by Enumerating Salmonella in Ground Meat and Poultry Products’ from the University of Minnesota. Dr Hedberg has received consulting income from Cargill (not related to this research), which provided in-kind support for this research. This relationship was reviewed and managed by the University of Minnesota in accordance with its conflict of interest polices.

Author ORCIDs

F. Sampedro1 0000-0003-1155-2751

References

1.Scallan, E et al. (2011) Foodborne illness acquired in the United States-major pathogens. Emerging Infectious Diseases 17, 7–15.Google Scholar
2.Marder, EP et al. (2017) Incidence and trends of infection with pathogens transmitted commonly through food and the effect of increasing use of culture-independent diagnostic tests on surveillance-foodborne diseases active surveillance network, 10 U.S. sites, 2013–2016. Centers for disease control and prevention. Morbidity and Mortality Weekly Report (MMWR) 66, 397–403.Google Scholar
3.Painter, JA et al. (2013) Attribution of foodborne illnesses, hospitalizations, and deaths to food commodities by using outbreak data, United States, 1998–2008. Emerging Infectious Disease 19, 407–415.Google Scholar
4.Food Safety Inspection Service (FSIS) database. Available at http://www.fsis.usda.gov/wps/wcm/connect/180fc804-0311-4b4d-ae42-d735e8232e1c/Salmonella-Serotype-Annual-2012.pdf?MOD=AJPERES (Accessed April 2017).Google Scholar
5.Food Safety and Inspection Service (FSIS) (2015) Changes to the Salmonella and Campylobacter verification testing program: proposed performance standards for Salmonella and Campylobacter in not-ready-to-eat comminuted chicken and turkey products and raw chicken parts and related agency verification procedures and other changes to agency sampling. Federal Register 80, 3940–3950.Google Scholar
6.U.S. Department of Health and Human Services (DHHS) database. Available at https://www.healthypeople.gov/2020/default (Accessed January 2017).Google Scholar
7.Routh, JA et al. (2015) Nationwide outbreak of multidrug-resistant Salmonella Heidelberg infections associated with ground turkey: United States, 2011. Epidemiology and Infection 143, 3227–3234.Google Scholar
8.Mead, G et al. (2010) Scientific and technical factors affecting the setting of Salmonella criteria for raw poultry: a global perspective. Journal of Food Protection 73, 1566–1590.Google Scholar
9.McEntire, J et al. (2014) The public health value of reducing Salmonella levels in raw meat and poultry. Food Protection Trends 34, 386–392.Google Scholar
10.WHO/FAO (2012) Risk Assessments of Salmonella in Eggs and Broiler Chickens. Microbial Risk Assessment Series 2. ISSN: 1726-5274. Geneva, Switzerland.Google Scholar
11.Teunis, PFM et al. (2010) Dose-response modeling of Salmonella using outbreak data. International Journal of Food Microbiology 144, 243–249.Google Scholar
12.Centers for Disease Control and Prevention (CDC) database. Available at http://www.cdc.gov/ncezid/dfwed/pdfs/salmonella-annual-report-2012-508c.pdf (Accessed December 2016).Google Scholar
13.Phang, H and Bruhn, C (2011) Burger preparation: what consumers say and do in the home. Journal of Food Protection 74, 1708–1716.Google Scholar
14.Bogard, A et al. (2013) Ground beef handling and cooking practices in restaurants in eight states. Journal of Food Protection 76, 2132–2140.Google Scholar
15.Murphy, RY et al. (2004) Thermal process validation for Escherichia Coli O157:H7, Salmonella, and Listeria monocytogenes in ground turkey and beef products. Journal of Food Protection 67, 1394–1402.Google Scholar
16.Manios, S and Skandamis, P (2015) Effect of frozen storage, different thawing methods and cooking processes on the survival of Salmonella spp. and Escherichia coli O157:H7 in commercially shaped beef patties. Meat Science 101, 25–32.Google Scholar
17.Meynell, GG and Meynell, EW (1958) The growth of micro-organisms in vivo with particular reference to the relation between dose and latent period. Journal of Hygiene 56, 323–346.Google Scholar
18.McCullough, NB and Elsele, CW (1951) Experimental human salmonellosis: I. Pathogenicity of strains of Salmonella meleagridis and Salmonella anatum obtained from spray-dried whole egg. The Journal of Infectious Diseases 88, 278–289.Google Scholar
19.Jones, TF et al. (2016) Characteristics of clusters of Salmonella and Escherichia coli O157 detected by pulsed-field gel electrophoresis that predict identification of outbreaks. Foodborne Pathogen and Disease 13, 674–678.Google Scholar
20.Cui, Y et al. (2015) Salmonella levels in turkey neck skins, drumstick bones, and spleens in relation to ground turkey. Journal of Food Protection 78, 1945–1953.Google Scholar
21.Straver, JM et al. (2007) Number of Salmonella on chicken breast filet at retail level and its implications for public health risk. Journal of Food Protection 70, 2045–2055.Google Scholar
22.Snary, EL et al. (2016) Quantitative microbiological risk assessment for Salmonella in pigs for the European Union. Risk Analysis 36, 437–448.Google Scholar
23.Guillier, L et al. (2013) Use of quantitative microbial risk assessment when investigating foodborne illness outbreaks: the example of a monophasic Salmonella Typhimurium 4,5,12:i:- outbreak implicating beef burgers. International Journal of Food Microbiology 166, 471–478.Google Scholar
Figure 0

Fig. 1.Risk assessment model framework. (a) Thawing and cooking scenarios. (b) Dose–response approach.

Figure 1

Table 1.Summary of FSIS data on prevalence and concentration of Salmonella spp. in ground turkey

Figure 2

Table 2.Model inputs on Salmonella spp. prevalence and concentration levels in ground turkey in the USA

Figure 3

Table 3.Model inputs on population and Salmonella dose–response data

Figure 4

Table 4.Model input data on consumption patterns at home

Figure 5

Table 5.Model input data on consumption patterns at the restaurant

Figure 6

Table 6.Model outputs by using different dose–response models

Figure 7

Fig. 2.Sensitivity analysis of the influence of the model inputs on the predicted number of illnesses.

Figure 8

Table 7.Public health impact by using different risk management enumeration strategies

Figure 9

Fig. 3.Number of illnesses and reported cases predicted by the feeding trial dose–response model (a) and the outbreak data dose–response model (b) at various contamination levels of a single 2000 lb lot of contaminated ground turkey. Shaded bars correspond to predicted illnesses. Black bars correspond to reported cases. Dotted lines correspond to the mean Salmonella spp. concentration in a positive lot (log MPN/g). Percentages correspond to the probability of a later investigation identifying an outbreak source.

Supplementary material: File

Sampedro et al. supplementary material

Sampedro et al. supplementary material 1

Download Sampedro et al. supplementary material(File)
File 95.7 KB