1 Introduction

Approximately 351,000 deaths occur annually on a global scale due to food poisoning (Sifferlin 2015). Most cases of food poisoning can be traced back to the presence of physical, biological, or chemical contaminants introduced at some point in the food supply chain. According to the Centers for Disease Control (CDC), there are over 250 foodborne diseases caused by the presence of contaminants introduced accidentally or intentionally. This has led to food safety concerns on a national and global scale.

In the United States, several key stakeholders play a vital role in protecting the food supply against intentional and accidental contamination incidents. The primary organizations within the federal government are the United States Department of Agriculture (USDA), The Food and Drug Administration (FDA), and the Centers for Disease Control and Prevention (CDC). The USDA Food Safety and Inspection Services (FSIS) is responsible for ensuring meat, poultry, and processed egg products are safe. FSIS creates regulations for food safety, inspects and enforces the regulations at food processing facilities, and works collaboratively with other agencies to respond to contamination incidents (USDA 2014). The FDA is also responsible for ensuring the safety of the food supply. However, their specific focus is on products not addressed by the USDA (American Bar Association 2014). Some of their responsibilities include creating regulations that food producing companies must follow, ensuring firms are compliant with food safety regulations, performing inspections and product sampling, conducting research into new technologies for inspection, and creating food defense tools that can be used by companies to prevent and respond to food safety incidents (FDA 2017). They also work with the CDC to track cases of foodborne illness. The CDC tracks information about foodborne illness to assist with risk mitigation and response efforts.

Developing optimal risk mitigation strategies requires an understanding of the prevalence of certain contaminants and food type combinations as well as any trends in consumer behavior. In addition, Batz et al. (2011) define three components to use a risk-based approach for food safety: (1) gather information on food that causes significant risk from a public health perspective, (2) prioritize resources and information to reduce these risks based on their effectiveness and cost associated with particular interventions, and (3) develop the optimal intervention and allocate the necessary resources. Batz et al. (2011) address the first component as part of their research and note that there are four high-risk food groups: produce, baked goods, dairy, and meat (Batz et al. 2011). These food groups are associated with the nutrients needed for daily intake. Besides these high-risk foods there are other trends associated with consumer behavior that impact food safety and the increase in foodborne illness. Consumers are dining at restaurants that grow their own produce or herbs outside of the establishment (Kramer and Fasone 2016). This causes regulatory concern with the food being grown in an environment susceptible to contaminants (Kramer and Fasone 2016). More consumers are seeking locally grown produce foods which when used in other products causes challenges for tracing the product back to the origin (Kramer and Fasone 2016). Consumers are also more focused on healthy eating. In 2016, there was an increase in the amount of recalls for frozen fruits and vegetables used in other products and smoothies (Kramer and Fasone 2016). New food services are available that deliver both cooked and uncooked meals to homes based on consumer demand for convenience foods (see for example companies likes Blue Apron and Hello Fresh). However, there is no way to identify if the food has been stored at the proper temperature for consumption. Lastly, consumer sensitivity to certain food products coupled with mass processing of convenience foods has contributed to several recalls due to cross-contamination or mislabeling. For example, food companies are mandated to label foods that are considered gluten free (GF) according to the standards given by the FDA (Kramer and Fasone 2016). Many restaurants have stated that they are not asked to validate GF during food safety inspections nor given the proper tools to determine the concentration of gluten (Kramer and Fasone 2016).

In this chapter, we will explicitly examine this relationship between consumer behavior and food safety. We use purchasing and consumption behavior as a proxy for estimating consumer preferences for certain convenience foods. We also incorporate a consumer compliance component that represents the absence or delay associated with receiving and responding to recall information.

1.1 Chapter Outline

This chapter discusses the operational challenges and latest research in food supply chain risk mitigation (FSCRM). We discuss some novel approaches to investigate food supply chain risk management bringing special attention to those models within the operations research/management science discipline. We also discuss new approaches to inform decision-making in FSCRM based on epidemic models and agent-based simulation. The epidemic model is based on the well-known SEIR (susceptible-exposed-infective-recovered) model. We adapt this model to determine the number of people that can become ill based on their purchasing and consumption behavior. We also present an agent-based modeling methodology that incorporates three interacting components within the food supply chain: the consumer, the manufacturer, and public health. The first component considers the human population as agents that can progress through various stages of health. Their health status is updated based on their behavior (product purchasing/consumption). The second component incorporates the recall process issued by the manufacturer. We specifically consider a recall intervention because it is the most common strategy used by companies to reduce the spread of the foodborne illness caused by a contaminated food product. The third component consists of assessing different messaging techniques during these food contamination events. A numerical study is conducted to understand the impact of consumer behavior and messaging on the spread of foodborne illness.

The remainder of this chapter is organized as follows. The second section presents the related challenges and opportunities associated within the food supply chain. The third section discusses the literature for food supply chain contamination, consumer behavior, and messaging. The fourth section presents two new approaches and corresponding results for modeling food supply chain contamination incidents. The conclusions and suggestions for future work are presented in Sect. 17.5.

2 Consumer Behavior and Food Supply Chain Safety

This section discusses some of the challenges associated with food safety and defense and the role of consumer behavior along three main themes: (1) communicating with the consumer, (2) measuring the consumer’s level of compliance, and (3) applying various intervention strategies. Tracking contamination events is one of the first lines of defense for reducing the risk of illness to the consumer. However, communicating with the consumer, under-diagnosis and under-reporting of illness, and new food and toxin combinations make tracking contamination events a challenging process.

2.1 Communication with the Consumer

Many consumers do not receive information regarding contamination events until a recall is issued or they are contacted by a distribution channel with the information (Hallman and Cutie 2009). Currently, the easiest way for a retail distribution channel to inform the consumers of this event is by telephone or email since the store can track consumer food purchases through their rewards program. However, all consumers do not necessarily use this program. Furthermore, certain segments of the population (e.g., minorities and underserved populations) might not receive or have access to food contamination and recall information. According to the survey done by Steelfisher et al. (2010), African Americans (86%) and Hispanics (81%) were less likely to remember a recall during the time of their survey compared to whites (94%). While government websites contain information about recalls, approximately 15% of American consumers that used those sites indicated the sites were not organized in a way to make it easy to find the food products involved in a recall.

The National Research Service reported that 49% of managers from the FDA felt that communication and coordination would help them to improve their ability to carry out the mission. Once a recall ends, most consumers are unaware unless a food agency shares the information on their website (Government Accountability Office 2012).

According to a Rutgers survey on improving communication during recall events (Hallman and Cutie 2009), less than 60% of American consumers have checked their home for a recalled food product. Even with information from the media about contaminated food, 12% of consumers stated that they ate a food product that was recalled. Out of this 12%, 9 respondents from the study thought they became ill from consuming the food product, with only 4 respondents taking action to go the doctor for the symptoms. Under-diagnosis and reporting also contributes to the lack of timely and accurate tracking of foodborne outbreaks. The Public Health Department in each state is responsible for reporting instances of foodborne outbreaks to the CDC for tracking and reporting purposes. Since, most consumers do not go to their healthcare provider to report symptoms, this leads to under-diagnosis and reporting. As a result, there is uncertainty in the amount of people affected by the contamination event.

2.2 Consumer Compliance

Although food agencies give notice to the public through various avenues, consumers do not heed the warnings. Many of these same consumers still eat food that is unsafe because they do not deem it to be a risk or they have not received any information on a recall being issued.

The FDA has experienced challenges with coordinating with other agencies to issue recalls appropriately to give consumers accurate information (Government Accountability Office 2012). In 2006, the FDA issued a warning about spinach that resulted in companies losing a large amount of money. In addition, many agencies do not give out information until they receive it which may leave consumers confused on what steps to take to discard food (Government Accountability Office 2012).

Many consumers, after purchasing and consuming food, are often unaware that some of their food puts them at risk from contaminants. For this reason, recalls occur. During the recall process, some consumers may choose to follow the guidelines to discard food or to not comply.

Steelfisher et al. (2010) did telephone surveys with regard to two major food recalls. This survey evaluated the consumers’ thoughts on food recalls and how it impacts their actions for the future.

The findings from this survey shows (1) better communication is needed on the actual products that are a part of a recall; (2) it is helpful to use media outlets such as television or radio to send out information to consumers, and (3) the need to reach out to minority populations about recalls and foodborne illness. Saulo and Moskowitz (2011) develop a set of food safety messages, which were shown to a population of respondents through the Internet. These respondents numerically answered how they would react to the various messages. From this data, conclusions were drawn on how specific messages can influence various demographics and behavior of consumers. Freberg (2012) uses a consumer panel survey to understand the effects of compliance, based on messages received from an organization versus those generated by a user. The results show that more people comply with messages received from organizations.

2.3 Intervention Strategies

The food protection plan lists interventions as one of its core elements (FDA 2007). An intervention is defined as targeted inspections and testing that verify that preventive controls are working and that resources are applied to high priority areas (FDA 2007). Interventions can be proactive, prior to food distribution, or reactive in response to a foodborne outbreak. Some of the common proactive intervention strategies are inspections, sampling, and surveillance within the supply chain. The most well-known reactive strategy is a product recall. Recalls can be initiated by the product manufacturer or by the FDA. The Federal Drug Administration (FDA) has developed standards to help determine if a food contamination event should result in a recall; however, the FDA is not given sufficient data in order to issue a recall in a timely manner (Government Accountability Office 2012). For example, their internal procedures do not help them effectively conclude that there is enough evidence to proceed with a recall.

Many recalls are issued by food companies that are possibly at fault. However, this results in false alarms, loss of revenue and customer support. There are mechanisms in place to help companies recover from the negative effects associated with a recall or advisory that might have been false; however, they are not always beneficial.

If they are falsely accused, companies are able to move through the judicial process to take the FDA to court. Companies may also be eligible to participate in government purchasing to improve their image with the public. For example, the Federal government provided one-time funding for tomato growers in South Carolina after the 2008 outbreak was incorrectly linked to their farms (Government Accountability Office 2012). Companies may also take out loans or insurance for these cases. This gives cause to consider the guidelines and protocol associated with timing of implementing a recall or other type of intervention.

Since there are a diverse set of contaminants that can enter the food supply chain, it is challenging to know the best time to initiate the appropriate intervention strategy.

To address these challenges, food companies need to test a variety of intervention strategies on contamination events with different levels of severity to improve food safety and defense. By utilizing information from past events, it is possible to develop baseline rates that will help food companies and organizations to respond to these types of events. It is also important to incorporate inspection and detection technology to help with removal of tainted food from the production phase of the food supply chain.

The third challenge focuses on the emerging pool of food and toxin combinations. The CDC maintains databases that track and use surveillance data through the National Outbreak Reporting Systems in order to link contamination information over states, agents, or food responsible for illness to the FOOD Tool. Discovery of these new agents can lead the way to create new technologies, techniques, and studies. They can provide more insight into how toxins evolve and track their growth. Public health officials will also need to develop new guidelines and courses of action to handle new cases.

2.4 Opportunities for Food Supply Chain Modeling

The prior sections illustrate the role of the consumer, beyond just someone who purchases the product. Reducing foodborne illness requires a coordinated effort to communicate with the consumer effectively about recalls, get the consumer to recognize and report foodborne illness to improve tracking of these events, and also getting the consumer to comply. Each of these challenges will be addressed through the agent-based simulation model.

3 Literature Review

Food supply chain contamination is a growing concern due to the rise in consumer illnesses and recalls. Within the literature there are various models to assess consequences, recall strategies, as well as individual information about a consumer and risk communication. However, there is limited research that utilizes consumer behavior data in these models.

3.1 Food Supply Chain Models

The food supply chain models focus on three main areas: production, distribution, and intervention strategies. Vlijac et al. (2012) characterize a set of supply chain disturbances and vulnerabilities which are used to develop redesign principles for supply chains. This framework is tested on a meat supply chain. Rong and Grunow (2010) develop a model that sends food from a distribution facility in batches to the retailer. They consider the effects of dispersion by reducing the amount of batches that are sent to specific retailers in order to prevent a recall. Akkerman et al. (2010) gives a summary of the literature related to models that solve diverse problems with food quality, distribution, and planning. Chen et al. (2013) create a model to evaluate quality control methods in the Chinese dairy industry under two supply chain structures. For each structure, they calculated the cost of sampling strategies from the retailer and supplier perspective. Buchanan and Appel (2010) discuss the integration of analysis and mathematical models to enhance information used by risk managers to meet regulatory requirements of food safety and defense. Manizini and Accorsi (2013) introduce a framework for an integrated food supply chain which considers the following characteristics such as quality, safety, efficiency, food products, and food processes. Chebolu-Subramanian and Gaukler (2015) propose an analytical model which is validated through simulation to study the origin and mode of detection for contamination on various food supply chain designs where illness has already occurred.

There have been some studies based on agent-based simulation modeling. Knowles-McPhee (2015) presents an agent-based model that tests three inspection strategies at a retailer in order to understand the interaction of consumers, retailers, and inspectors. Chaturvedi et al. (2014) develop an agent-based model that can be used for food defense training. The model incorporates the total food supply chain to help companies understand and prepare for a food contamination event. Crooks and Hailegiorgis (2014) create an agent-based model to show the interactions between humans and their environment by using a Susceptible-Exposed-Infected-Recovered (SEIR) model as the underlying structure. This model considers the spread of cholera from unclean water resources and is used to understand the humanitarian relief perspective. Zechmann (2011) uses a multiagent-based approach to determine the optimal mitigation strategies after a water contamination event.

3.2 Consequence Assessment Models

There is a limited number of food consequence assessment models, but they focus on various areas to capture information to determine which mitigation strategies to implement during some event. One of the main characteristics of these models is to determine the number of people that will be affected by an event. Many models capture this by counting the total number of casualties of a food contamination event (Liu and Wein 2005, 2008; Hartnett et al. 2009). Another characteristic is the effect of contaminants that enter into the supply chain through various modes such as production facilities, distribution, and retailers (Jaine 2005; Liu and Wein 2005, 2008). After food contamination takes place, some models consider the impact of public health response directed at consumers (Hartnett et al. 2009). Lastly, none of these models consider the effects of consumer behavior on their risk of illness and compliance.

3.3 Consumer Behavior

Various studies and surveys are used to capture information on consumers’ behavior in relation to food choices, purchases, consumption, food handling, and compliance. Erongul (2013) reports on a survey to evaluate the relationship between consumer food safety and consumption patterns. The results from this study show that more consumers need education in best practices for handling food in order to prevent illness. Grunert (2002) presents a Total Food Quality Model framework. This framework is used to understand the consumers’ perception of food quality, food technology, and changes in behavior due to a contamination event. Caraballo-Martinez and Burt (2011) use a survey to collect data on the changes in a consumer’s choice in grocery or household products. This behavior can assist retailers in determining which products to sell and whether to open new stores.

3.4 Interventions

Fendyur (2011) provides an overview of the operations research techniques used to handle outbreak management of infectious diseases. Dasakalis et al. (2012) provide a review of models that focus on epidemic control and logistics with many applications. Many researchers have developed models that focus on control measures related to public health measures for diseases and food.

Liu and Wein (2005) develop a differential equation model that implements two antibiotic residue testing of trucks after an attack occurs. With testing, the number of people affected reduced to half for one strategy. The other strategy when used in isolation prevents even more people from becoming poisoned. Liu and Wein (2008) consider detection after a certain amount of consumers develop symptoms from eating a contaminated food product. After this threshold is reached, all food consumption is stopped. Hartnett et al. (2009) create a discrete event simulation model. One component of the simulation considers the number of reported cases. Once this number reaches a threshold, then an advisory is sent out to the public. They issue a delay to determine the point of contamination. After this is confirmed, exposures can still occur based on the way consumers comply to the advisory and all products are moved out of the supply chain. Jaine (2005) develops a simulation that has an interface for food safety officials to use to pick the type of intervention based on the outbreak. The model uses information from literature, various websites, and data sets that were provided by certain companies. Chang et al. (2015b) develops a partially observable Markov game which allows two agents to interact based on little information to make decisions on how to affect the overall system. A food supply chain is used to illustrate this model. Chang et al. (2015a) uses a partially observable multi-objective game, to help a production manager determine the best way to maximize productivity of a supply chain and minimize the number of products that leave the production facility contaminated. Chebolu-Subramanian and Gaukler (2015) use sampling during the production process to detect contamination in food. Chen et al. (2013) uses an analytical model to do an acceptance sampling test of food products to make sure they conform to specifications required by the retailer. Based on the results from this test, they can choose to accept or reject the production lot.

3.5 Research Contribution

While agent-based models have been used to study some contamination problems, there are very few for the food supply chain. The papers discussed in the literature review, related to food consequence assessment models, do not focus on the consumers at the individual level. In this research, we show the progression of illness by using the following parameters: the rate of purchase, the rate of consumption, and the time until illness and recovery. The time it takes to purchase and consume the contaminated food products is considered stochastic. The literature presented does not consider illness from the perspective of the consumer. This research merges the characteristics of the consequence assessment models and food supply chain models with consumer behavior, specifically purchasing, consumption, and compliance (Table 17.1).

Table 17.1 Paper by characteristics
Full size table

4 New Methods to Model Food Supply Chains

Two new methods are presented to model food supply chains. The first method is based on epidemic and compartmental models. The second method is an agent-based simulation model.

4.1 Consequence Assessment Compartment Model

4.1.1 Methodology

4.1.1.1 Problem Overview

This research considers a food supply chain with the following structure: (1) The producer is a farm or company that produces an original product; (2) Processors are farms or companies that change the original product into a different form; (3) Distributors supply food products to different companies and industries; (4) Food reaches a consumer from one of three different distribution channels: (1) Food Retail, (2) Food Manufacturing, and (3) Food Service (Fig. 17.1).

Fig. 17.1
figure 1

The food supply chain

Full size image

A chemical or biological agent can be released at different stages of the food supply chain.

Distribution channels receive impure food, which consumers purchase and ingest. From this attack the amount of casualties are determined based on symptoms that affect consumers from the agent. A deterministic SEIR model is presented to show the purchase and consumption behavior of consumers assuming a non-constant population.

4.1.1.2 Model Description and Assumptions

A deterministic ordinary differential equation progression model is developed to illustrate food contamination within consumer populations. The model shows the progression of a consumer from a healthy to unhealthy state because of eating contaminated food. The phase of progression is as follows: (1) Susceptible (S) population represents the people who purchase contaminated food products from a distribution channel but have not yet consumed the product; (2) Exposed (E) population represents the people that consume a contaminated product but show no signs of illness; (3) Infected (I) population represents the people that show signs of illness after consuming a contaminated product; (4) Recovered (R) population represents the people that are no longer ill after a single outbreak. This model is different from the original SEIR model because sickness is a result of contact with a particular food product and not a person. Therefore, there is no interaction term. Based on the flow of each stage we construct a system of differential equations (Fig.~17.2).

Fig. 17.2
figure 2

SEIR model flow

Full size image

Table 17.2 displays the notation used for the variables and parameters in this model.

Table 17.2 Model notation
Full size table

Using the notation from (Table 17.2) the deterministic SEIR model for a food contamination event is formulated as follows:

$$ \frac{\mathrm{d}S}{dt}=\mu N- mS(T)-\nu S(t)\vspace*{-16pt} $$
(17.1)
$$ \frac{dE}{dt}=\nu S(t)- mE(t)- aE(t)\vspace*{-12pt} $$
(17.2)
$$ \frac{dI}{dt}= aE(t)- mI(t)- gI(t)\vspace*{-12pt} $$
(17.3)
$$ \frac{dR}{dt}= gI(t)- mR(t) $$
(17.4)

The total population N becomes susceptible by purchasing tainted food products with rate μ. They can exit this compartment (S) due to natural death with rate m or from exposure to a contaminant in food with rate n (Eq. 17.1). A consumer can move to the exposed compartment with rate ν only after they purchase a contaminated product. They exit the compartment with rate m because of natural causes or if they become ill with rate a (Eq. 17.2). The consumers that become ill with rate a can eventually move to recovery with rate g. They can also exit compartment (I) due to natural causes with rate m (Eq. 17.3). Consumers receive some medical care in order to go to the recovered class. They can exit with a natural death with rate m (Eq.~17.4). Consumers can progress through the five stages consecutively to denote their place during a contamination event (i.e., susceptible, exposed, infective, recovered). This model considers a non-constant homogeneous population of consumers that can purchase food items that are always in stock. Contaminated food products are sent to a distribution channel (DC) (i.e., food retail, food service) for purchase and consumption. The consumer illness is based on the interval of time between the consumer purchasing and ingesting a contaminated food product. This model represents a worst case where food continuously enters the market undetected by the producer.

4.1.1.3 Experimental Design

The purpose of this research is to understand the effects of various population parameters on the number of people affected by food contamination. The parameters considered in this study are population size, consumer consumption behavior, consumer purchasing behavior, and discarding policy. A numerical example is developed to answer the research questions in Table 17.3.

Table 17.3 Research questions
Full size table

Each model case uses data from a food and consumer behavior survey which includes time until symptoms occur from contaminants and the food shelf life (Watkins 2015). There were 83 responses to the survey, which capture various demographic, consumer purchasing and consumption behavior information. The limitations of this study were the lack of diversity and size of the sample population. However, two questions were used from this survey to determine purchase and consumption rates.

Question 1. In the last 30 days, how many times did you buy the following food from the grocery store?

Question 2. How many days after purchase do you store the food (in the refrigerator or pantry) before you first eat it?

Stocking policies were created for each food type. Dairy, vegetables, bread, and eggs have a short shelf life, which is five days or fewer. This could result in people stocking these products more often because they are consumed faster. Baked goods have a medium shelf life which is two to three weeks with meat having the longest shelf life of over one month. Using the data from Question 2, a range was created for each food type to represent the time until consumption. The data from Question 1 was used to develop the purchasing rates.

Table 17.4 summarizes the notation used to develop these rates. Equation 17.5 calculates the consumption rate by using the average time until consumption in days for all food types and responses. Equation 17.6 calculates the purchasing rate using the average frequencies that a respondent shops for each food type. Equation 17.7 calculates the shelf life rate using the shelf life of the various food types. The shelf life data is obtained from publicly available sources (Tasty 2015).

$$ {\nu}^{-1}=\frac{1}{M}\sum \limits_{r=1}^M{c}_{\mathrm{fr}}\vspace*{-12pt} $$
(17.5)
$$ {\mu}_f=\frac{1}{M}\sum \limits_{r=1}^M\frac{j_{\mathrm{fr}}}{30}\vspace*{-12pt} $$
(17.6)
$$ {m}_{\mathrm{f}}=\frac{1}{h_{\mathrm{f}}} $$
(17.7)
Table 17.4 Notation for parameter computation
Full size table

4.1.2 Results

4.1.2.1 Single parameter sensitivity analysis

The SEIR model captures the amount of consumers that purchase, ingest, become ill and recover due to contaminated food products. Figure 17.3 shows the general behavior of the SEIR Model for the case where there are no interventions introduced into the model. The total population of 100 consumers all purchase food products that are contaminated. The long-term behavior shows that the amount of people purchasing contaminated food products decreases. This is a result of consumption and consumer illness. The number of consumers exposed to the contaminants peaks at around 3 days. The long-term behavior shows that eventually the amount of people exposed reaches a steady state. The number of ill consumers increases based on the rate of consumption. Over time all of those consumers receive some treatment that allows them to recover from their illness.

Fig. 17.3
figure 3

Baseline results

Full size image
4.1.2.2 Purchasing Behavior

Figure 17.4 shows the amount of people exposed to contaminated food based on consumption and purchasing behavior. In general, the exposure percentage increases as the purchasing rate increases. In addition, a 50% increase in the purchasing rate results in a 50% increase in the exposure percentage. The relationship between the purchasing and consumption rate is intuitive and serves to validate the model. Also, an increase of population size increases the number of people that become ill.

Fig. 17.4
figure 4

The percentage of consumers that are exposed to contaminated food products based on purchasing rate

Full size image

Vegetables are a short shelf life product which consumers purchase at a rate of 4.55 every 30 days; this results in 29.87% of consumers being exposed to salmonella. Baked goods have a medium shelf life which consumers purchase at a slower rate of 1.24 every 30 days; this results in 8.12% of the total population being exposed to contaminants.

4.2 Agent-Based Simulation Model

4.2.1 Methodology

This section describes the agent-based simulation model based on the Overview, Design Concepts, and Details ODD framework (Grimm et al. 2006).

4.2.1.1 Purpose

The purpose of this study is to model the effects of three types of consumer behavior: purchasing, consumption, and compliance, on (1) the food supply chain, (2) risk of illness, and (3) recall.

4.2.1.2 State Variables and Scales

The intervention simulation will consist of two agents: the consumers and food (Figs. 17.5 and 17.6). The consumer population moves into the purchase state where they can buy food products from two different distribution channels (i.e., food service and food retail). After purchasing products, consumption takes place and people can become ill. Based on the threshold of consumers that can become ill, a recall or some other intervention can occur. If the recall occurs, all stakeholders are notified and a warning given so that contaminated food products are removed from the shelf. Also, at this time a message is sent to the consumer to give warning that the food they purchased has been contaminated. Lastly, a consumer can recover from illness. The food agent models the recall process. Table 17.5 shows the notation used for this model.

Fig. 17.5
figure 5

Consumer statechart

Full size image
Fig. 17.6
figure 6

Food statechart

Full size image
Table 17.5 Food simulation model notation
Full size table
4.2.1.3 Process Overview and Scheduling

A consumer progresses through different compartments based on their health status. Based on the rates associated with the agent, some may progress faster than others. Once the number of ill consumers is greater than the threshold, the recall occurs. Based on this recall, the consumer compliance behavior is tested to see if they discard or keep products.

4.2.1.4 Design Concepts

For this simulation, the general food supply chain is considered; however, all parts are not modeled explicitly: (1) producer, (2) distributor, (3) retailer, and (4) consumer. For the recall process, we use a threshold value to signify that a food contamination outbreak has occurred; at the same time as testing of products is initiated. Lastly, we utilize information of consumer behavior to model purchasing, consumption, and compliance.

4.2.1.5 Emergence

The results from this model will show (1) the number of people that are affected in relation to compliance measures and (2) the number of products that were discarded as a result of the recall process.

4.2.1.6 Adaptation/Learning

Consumers update their compliance behavior based on information regarding a recall. The recall process is updated based on consumer purchasing and consumption behavior.

4.2.1.7 Objectives

The objectives of this research are (1) to understand how consumer compliance reduces illness and (2) the effects of uncertainty in consumer purchasing and consumption behavior.

4.2.2 Results

Table 17.6 displays the number of people that are affected from a food contamination event when compliance is introduced. For each compliance rate there are fewer people that are exposed and ill at the end of the time period. However, most consumers eventually progress to the recovered class at the end of the time period. Overall the results show that with a lower discard rate, less consumers (748.08) comply.

Table 17.6 Number of consumers affected based on compliance
Full size table

Figures 17.7 and 17.8 display graphs of the distribution of the recovered and discard class, which has the most impact on the number of people affected by contamination based on compliance.

Fig. 17.7
figure 7

Recovered class with compliance rate = 0.25

Full size image
Fig. 17.8
figure 8

Recovered class with compliance rate = 0.75

Full size image

Table 17.7 and 17.8 provide descriptive statistics for each class. At a lower discard rate, the recovered class shows more variability (17.2) in the number of consumers that are affected compared to a higher discard rate (12.9) (Figs. 17.7 and 17.8, Table 17.7). This represents people that receive information about a contaminated food product but ignore the warnings. It also represents people that may have already consumed the product before they were aware of any information about contamination.

Table 17.7 Statistics for recovered class
Full size table
Table 17.8 Statistics for discard class
Full size table

Similarly, with the discard class, the lower discard rate also shows more variability (17.23) than the higher discard rate (12.82) (Figs. 17.9 and 17.10, Table 17.8). However, a higher discard rate results in more consumers heeding the information of food safety officials, especially for major food outbreaks (Fig. 17.10). The information presented to consumers from food safety organizations can promote better prevention of illness.

Fig. 17.9
figure 9

Discard class with compliance rate = 0.25

Full size image
Fig. 17.10
figure 10

Discard class with compliance rate = 0.75

Full size image

Tables 17.9 and 17.10 display the descriptive statistics for the main compartments associated with a person coming into contact with a contaminated food product with using messaging and without messaging. Table 17.9 shows that most consumers become exposed to a contaminated food product however they are not remaining sick for long. Also, we can see the number of products has decreased based on introducing messaging to the consumers. This is due to the low compliance policy of 0.25 or consumers behaving contrary to the guidelines (Table 17.9). Overall, without messaging we have more consumers coming through the system without any warning which can keep contaminated food products in homes. Although all consumers do not comply, it is important to observe this behavior to develop the best policies to prevent further spread of contaminated food. Table 17.11 shows us the relative increase or decrease in the recovered and discard class given the messaging and non-messaging case. During the messaging case between 400 and 600 messages were sent to consumers to warn them the food they purchased was contaminated. However, comparing the two cases there was a decrease in the amount of people that recover which shows that more people are receiving the information about the contaminated food products even if they do not necessarily discard the food. The consumers might choose not to eat the food product anymore which shows that over time they may develop new behavior based on warnings.

Table 17.9 Statistics for 0.25 discard rate with messaging
Full size table
Table 17.10 Statistics for 0.25 discard rate without messaging
Full size table
Table 17.11 Relative increase and decrease of compliance of 0.25
Full size table

5 Future Research Opportunities for Food Supply Chain Contamination, Mitigation, and Risk Communication

This chapter focuses on challenges and opportunities that arise from intentional contamination in the food supply chain such as (1) tracking contamination events, (2) intervention strategies, and (3) consumer compliance. It addresses each of the challenges by using various approaches. Two models were developed to understand the number of consumers that become ill, the number of products linked to illness, and the effects of a simple intervention strategy. An agent-based simulation model was presented to understand how a simple intervention can affect the number of illnesses and consumer compliance.

The first set of models showed that compartment models can be used to understand population data and progression of symptoms in consumers. It allowed for flexibility in the number of characteristics to show for the population. Results for these models show that even by changing all the population sizes the percentage of people affected from contaminated food remains the same. Risk increases as the consumer purchasing frequency increases for all food types this is due to exposure to contaminants. Depending on the purchasing and consumption behavior some food products result in higher risk. The third model presents an agent-based model to consider consumer compliance. The results show that as we increase information about compliance, fewer illnesses occur. This modeling approach can be used to analyze consumer behavior at the individual level.

As food contamination events continue to rise, more research is needed to address how to prevent illness from spreading to help food safety officials. The future work will consider the following areas: (1) Consumer characteristics, (2) Data collection, (3) Cost analysis, (4) Intervention and mitigation strategies, (5) Compliance, (6) Public health, and (7) Traceability. The current models only account for a limited number of characteristics however, introducing more into the model could add extra insight especially if the consumer may be in a vulnerable population. More data is needed on consumer behavior as well an information about the recall process and various contamination events to use in the models. This can be used to develop real-world scenarios to test and validate the models. Currently there are a limited amount of studies done to show the impact of cost related to intervention strategies and cost of illness in a food contamination event. This can help manufacturers to develop more robust food defense plans. More research is needed to evaluate various timelines of food contamination and the implementation of intervention strategies. This will allow the FDA and other food organizations to create a baseline for responding to different levels of severity for food contamination events. Although messaging is being used to warn consumers about contaminated food, there is still low compliance. All messaging needs to be able to reach all populations. There needs to be better guidelines in place to instruct consumers on how to dispose or return contaminated food so it can be tested for the agent considered. Another aspect of this problem that needs more development is public health. Many consumers do not report their symptoms or may not think they are symptomatic from eating contaminated food. Better ways are needed to track these consumers so they can receive the proper care and understand the effects of the spread of illness in various regions especially for multi-state outbreaks. This can help identify food outbreaks earlier. Lastly, traceability is becoming harder as the global food supply chain expands. Some areas for development are gaining understanding of how producers and manufacturers work together when using certain food in other products and developing policies to continue traceability throughout the whole supply chain process.

Based on these developments in our food systems it is clear that transparency and safety is key to keep our food safe and consumers supportive because they have healthy food choices.