Understanding the Relationships Between Inspection Results and Risk of Foodborne Illness in Restaurants

Introduction

In the United States, restaurants employ an estimated 14 million people across 1 million establishments. Annual restaurant industry sales of $700 billion account for 47% of food expenditures in the United States (National Restaurant Association (NRA), 2015). Restaurants are also important settings for foodborne disease outbreaks, accounting for 60% of outbreaks and 51% of outbreak-associated illnesses reported to the Centers for Disease Control and Prevention (CDC, 2015) in 2013. In addition, multiple case–control studies of sporadic Salmonella, Campylobacter, and Escherichia coli O157:H7 infections have found an increased risk of illness associated with foods eaten in restaurants (Jones and Angulo, 2006). Outbreak investigations and systematic studies have evaluated food handling practices and restaurant polices that may contribute to this increased risk of illness (Hedberg, 2013).

The U.S. Food and Drug Administration (FDA) developed a model Food Code to provide uniform standards for food handling practices in restaurants and a framework for regulatory agencies to inspect against those standards (Food and Drug Administration, 2013). The model Food Code, inspection practices, and report forms have been adopted by many states, frequently amended to meet unique local food safety and licensing requirements. Methods for scoring inspections and reporting results vary widely. In some jurisdictions, results are posted at the establishment in the form of scores or grades and many jurisdictions now make inspection results available online. This has led to the development of smart phone “apps” that allow consumers to view inspection results when considering where to dine. The usefulness of this information to reduce the risk of foodborne illness is unknown, in part, because of public misconceptions and unrealistic expectations of restaurant inspections (Jones and Grimm, 2008).

Several studies suggest that posting restaurant inspection results as grades in the restaurant window or as scores on the internet leads to improved sanitary practices in the restaurant (Fielding et al., 2001; Waters et al., 2013; Wong et al., 2015). Such improvements could account for observed reductions in the risk of foodborne illness following their implementation (New York City Health Department, 2012; Simon et al., 2005). However, studies that have attempted to evaluate the relationship between inspection results and the risk of outbreaks have produced mixed results (Irwin et al., 1989; Cruz et al., 2001; Jones et al, 2004; Patel et al., 2010). The risk of illness associated with Food Code violations is pathogen specific. For example, Food Code provisions regarding holding temperatures may control the growth of bacteria such as Clostridium perfringens, but would have no effect on norovirus. Because of this, the mixed results of previous studies may be due, in part, to the mix of outbreaks included in the studies (Petran et al., 2012). To further examine the relationship between inspection results and the risk of foodborne disease outbreaks, we evaluated results of routine inspections conducted in multiple restaurants in a chain that was associated with a large Salmonella outbreak.

The Illinois Department of Health investigated an outbreak of Salmonella enterica ser. Hvittingfoss (Salmonella Hvittingfoss) associated with multiple restaurants in a chain (Chain A) during May and June, 2010. A detailed summary of the investigation was released on October 14, 2010 (Illinois Department of Public Health, 2010). Ninety-seven patron cases and 12 food handlers with confirmed Salmonella Hvittingfoss infections were identified. Onsets of illness ranged from April 25 to June 30. The first indication of the outbreak was the recognition of an increased number of cases seen by a local health department on May 25, 2010. A cluster of Salmonella Hvittingfoss was identified by the Illinois Department of Public Health Laboratory on May 27, 2010, and an investigation was initiated. Six days later, on June 2, 2010, Chain A restaurants pulled four suspected produce items (lettuce, tomatoes, green peppers, and onions) from all restaurants in the outbreak area.

Materials and Methods

To evaluate the relationship between inspection results and risk of illness, a retrospective study was conducted on restaurant inspection reports from Chain A restaurants in counties involved in the outbreak. The objective of the study was to determine whether there were indicators present in the restaurant inspections that might have predicted which of the Chain A restaurants would be involved in the outbreak. Because the outbreak involved restaurants served by a common distributor in central IL, it was assumed that contaminated food items may have been served by any of the Chain A restaurants in the outbreak area. Thus, we hypothesized that reported cases from some restaurants may have resulted from amplification of contamination at these restaurants.

A list of restaurants linked to the outbreak by confirmed cases was obtained from the Illinois Department of Health. Cases reported exposure to Chain A restaurants in 28 counties. Multiple cases reported exposure to Chain A restaurants in 16 counties, accounting for 85% of outbreak associated cases. We attempted to request inspection records from each of these counties. From eight of these counties, we obtained all inspection reports from Chain A restaurants from January 1, 2008 through May 27, 2010. The collaborating counties included more case-associated restaurants and more outbreak-associated case exposures than the noncollaborating counties. Inspection reports were obtained for all Chain A restaurants in each county, regardless of whether the individual restaurant was linked to any illnesses. The cutoff date for inspections included in the study was May 27, 2010 since that was the date the Illinois Department of Health notified local health departments of the outbreak. Only routine inspections were included in the study. All preopen, seasonal, complaint, and follow-up inspections were excluded, because these inspections do not typically involve broad-based assessments of food-handling practices. Restaurants that were not linked to any confirmed cases were compiled into a control group.

Results of inspections were entered into Excel spreadsheets. The following information was extracted for analyses: county, address, type of inspection, date of inspection, violation category, county violation number, violation description, and demerit points. The data were analyzed using Excel Descriptive Statistics and Pivot Tables. The final descriptive analyses were done using CDC Epi Info 7. Descriptive Statistics, ANOVA (Parametric Test for Inequality of Population Means), Chi-square and the Mann–Whitney/Wilcoxon Two-Sample Tests were calculated for each variable. The parameters studied included total demerit scores over the period of January 1, 2008 through May 27, 2010 in the outbreak and nonoutbreak restaurants; the total demerit scores of the last inspections before the outbreak; hand washing and good hygienic practice violation scores; cross-contamination violation scores; and all violation categories found per establishment per inspection. We also applied grading criteria from Los Angeles and New York City to evaluate whether these grading systems would change prediction for a restaurant to be involved in the outbreak (NYC, 2012; County of Los Angeles, 2014).

In response to a freedom of information act (FOIA) request, we received de-identified outbreak records from the Illinois Department of Health, to conduct additional epidemiological analyses to identify potential contributing factors to outbreak transmission. De-identified illness and food histories were obtained in an Excel spreadsheet. Data were aggregated and analyzed to reproduce results summarized in the report. To test several hypotheses regarding transmission in the restaurants that were not directly addressed in the report, additional analyses were conducted.

The environmental health evaluations of the restaurants found that green peppers were sliced on the same equipment used to slice cucumbers and onions. To evaluate whether cross-contamination of the slicer may have contributed to transmission, a combined slicer variable was constructed. Twelve food handlers in seven different restaurant locations were found to be infected with Salmonella Hvittingfoss. To determine whether these food handlers may have contributed to transmission to patrons, a variable identifying restaurant with infected food handlers was introduced.

Results

Inspection results were collected from 106 Chain A establishments in the eight counties included in the study. Forty-six outbreak-associated cases were linked to 23 of the Chain A restaurants. Single cases were linked to 13 restaurants, with clusters of 2–6 cases associated with 10 restaurants (Table 1).

The data for the mean total demerit score for the overall period for the control restaurants and the outbreak restaurants were 13.5 and 10.9, respectively, and the mean total demerit for the last inspections was 13.9 and 10.0, respectively (Table 2). There were no significant differences between the outbreak and nonoutbreak restaurants for overall demerit points or for the number of demerit points attributed to hand washing or cross-contamination. Restaurants with multiple cases did not have higher demerit scores than restaurants with single cases. Of 10 inspections conducted during May 2010 when cases were actively exposed, 8 were conducted among control restaurants and 2 were conducted in case-associated restaurants. One of these had six cases associated with it. An inspection on May 4 and a repeat inspection on May 18 noted violations for unclean food contact surfaces and inadequate sanitary solutions. Only one of the eight control restaurants had either of these violations cited.

There were no significant differences observed between outbreak and nonoutbreak restaurant inspections with respect to grades imputed from either the LA County or the NY City restaurant inspection grading criteria, either for all inspections conducted during the study period or for the last inspection conducted before the outbreak was recognized (Table 3). Comparing all inspections, 68% had the same grades by both criteria, and 31% of inspections received a higher grade by the LA County criteria.

The results of the case-companion analysis as presented in the Illinois report were replicated in the current evaluation, suggesting that a fresh produce item was the likely source of the outbreak. There was no association between illness and eating vegetables sliced on the common slicer (odds ratio [OR] = 1.4; 95% confidence limits [CI] = 0.5, 3.9) or between illness and eating at a restaurant with an infected food worker (OR = 2.1; CI = 0.6, 8.3). Five of 12 infected food handlers reported illness. Thus, the clinical attack rate among all 504 food handlers was 1%.

Discussion

Routine restaurant inspections based on the FDA Food Code are an important tool of regulators to ensure that the principles of food protection are being adhered to by commercial food preparers (Food and Drug Administration, 2013). Consumers also rely on the results of routine restaurant inspections to assure them that foods are prepared under hygienic conditions and will not serve as sources for foodborne illnesses. Increasingly, inspection results are made available at food facilities or websites for easy and convenient access to be used by consumers when deciding on a safe place to eat (New York City Health Department, 2012; Simon et al., 2005; Waters et al., 2013; Wong et al., 2015). Scores, letter grades, or other icons are viewed as symbols guaranteeing compliance with the Food Code and that compliance translates into food safety. In other words, restaurant inspection results are held as predictors for the risk of foodborne illness.

In practice, results of inspections appear to have limited value for predicting the general risk of foodborne illness outbreaks in restaurants. Studies that have demonstrated positive association have focused on outbreaks caused by bacterial agents such as Staphylococcus aureus, C. perfringens, or Salmonella, where specific violations of temperature control or cross-contamination may lead to significant growth of the agent in food and high attack rates leading to detection of the outbreak (Patel et al., 2010; Irwin et al., 1989). In studies that included a high proportion of agents that cannot be propagated on food, such as norovirus, results of inspections have not been linked to outbreak occurrence (Jones et al., 2004). Because specific Food Code violations may increase the risk of illness caused by some agents, but not others, the overall inspection scores lack predictive value (Petran et al., 2012). This is especially true in an era where norovirus is the leading cause of foodborne outbreaks and foodborne illnesses (CDC, 2015).

The results of our study provide further insights to our understanding of the relationship between inspection results and the risk of foodborne illness in restaurants. In addition to the agent, the agent and food item pairing and route of transmission must also be taken into consideration. In our analysis of the Salmonella Hvittingfoss outbreak, total demerit scores for routine inspections and specific analyses for hand washing and cross-contamination revealed no contributing factors that would lead to the outbreak. The Illinois Department of Health listed the contributing factor as item C7 (ingestion of contaminated raw product) and suspected several produce items as a likely vehicle (Illinois Department of Public Health, 2010). Thus, Food Code provisions would not be expected to prevent transmission in this setting. In contrast, a similar study of a Salmonella Montevideo outbreak associated with chicken in a chain of restaurants in Maricopa County, AZ found that individual restaurants with multiple cases were more likely to have critical violations associated with potential for cross-contamination on routine inspections (Patel et al., 2010).

As observed in the Maricopa County outbreak, routine restaurant inspections are very useful in detecting Food Code violations associated with temperature control and cross-contamination. Because these may be linked to amplification of bacteria in food, results of inspections may identify establishments at risk for classical foodborne outbreaks caused by Salmonella or toxin-producing bacteria. The current low level of occurrence of outbreaks caused by toxin-producing bacteria in restaurants is evidence for the success of the restaurant inspection system. The Food Code also contains guidelines aimed at employee health and hygienic practices. However, these may be more difficult to observe during limited inspections. This is one reason for emergence of norovirus as the leading cause of foodborne illness.

In the Salmonella Hvittingfoss outbreak, the restaurants merely served to pass the contaminated produce on to customers. Inspections at these facilities would be unlikely to detect or predict the foodborne illness outbreak because there are no Food Code items in place to stop the introduction of contaminated food from an otherwise approved commercial food source. The original point of contamination occurred too far back in the chain for the restaurant inspection to be a useful indicator for potential foodborne illness. Restaurant purchasing specifications and Food Safety Modernization Act provisions that could address source contamination concerns are needed to enhance the ability of the Food Code to prevent foodborne illness outbreaks that occur in an otherwise hygienically sound environment. In addition, to provide farm to fork protection for food safety, employee health, proper hygienic, and food handling practices must be implemented in processing and manufacturing facilities. Companies should implement best management practices and/or a hazard analysis of critical control points or similar food safety management procedures accompanied with good documentation. This is especially important due to the weight that is given to postings on websites or posted on facility windows and doors. The public translates good scores as safe food. However, the occurrence of outbreaks caused by episodes of food handler illness and contamination of fresh produce items could undermine public confidence in the restaurant inspection system (Jones and Grimm, 2008).

In the absence of definitive measures to prevent outbreaks associated with contaminated fresh produce, the speed and effectiveness of outbreak investigations to identify and control the source of contamination assumes greater importance. Because this was a large outbreak caused by an uncommon serotype of Salmonella, a common restaurant chain source was rapidly identified. The geographic distribution of restaurants served by a common distributor further facilitated the identification and removal of suspected produce items within a week of the outbreak recognition (Illinois Department of Public Health, 2010). Removal of these produce items appeared to have effectively stopped transmission to patrons.

Improving outbreak investigations is important to identifying contaminated food items and developing better prevention strategies. It is also important for improving our understanding of the ability of restaurant inspections to inform the public about the risk of illness associated with restaurants.