Estimating the Burden of Foodborne Disease,South Korea,2008–2012

Introduction

Estimating the actual occurrence of foodborne illnesses is important for establishing and monitoring food safety policies (Scallan et al., 2011a; Thomas et al., 2013). However, making these estimations is challenging because only a small proportion of foodborne illnesses are confirmed by laboratory testing and reported to public health agencies (Scallan et al., 2011a). Many cases of foodborne illness are not identified by public health surveillance systems because of underdiagnoses or underreporting (MacDougall et al., 2008). Therefore, several studies have attempted to accurately estimate the overall number of cases of foodborne illness (Mead et al., 1999; Hall et al., 2005; Scallan et al., 2011a, b; Tam et al., 2012; Thomas et al., 2013; Wilking et al., 2013).

In national-level initiatives, England, The Netherlands, and the United States were the first countries to study the burden of foodborne disease, followed by Australia, Canada, and Ireland (Flint et al., 2005). These efforts identified many data gaps and methodological limitations, and now, better-developed data and methods have become available (Scallan et al., 2011a). Although the estimation of the overall number of episodes of foodborne illness is helpful for allocating resources and prioritizing interventions (Thomas et al., 2013), no attempts have been made yet to accurately estimate the burden of common foodborne diseases in South Korea. In addition, eating habits and differences in local food cultures may affect foodborne disease occurrence (Havelaar et al., 2010). The unique food culture of South Korea varies significantly from other countries and therefore may highlight key differences in the role of food culture in the occurrence of foodborne diseases patterns. Understanding the characteristics and causes of foodborne disease in South Korea is a critical component of understanding foodborne disease internationally.

The current study used data from the Health Insurance Review and Assessment Service (HIRA, 2013), an institute that reviews medical fees, to calculate the number of cases and hospitalizations due to 18 specific pathogens and unspecified agents commonly transmitted by contaminated food from 2008 to 2012 in South Korea.

Methods

Determining diagnostic accuracy

One concern with using the HIRA data is the accuracy of diagnosis by hospital physicians because HIRA data are collected from the first medical examination only. According to a study of medical diagnoses in South Korea in 2006 (KCDC, 2007), the proportion of false-diagnosis (including nonjudgment) was approximately 2.3% for inpatient stays and 8.0% for outpatient visits; these values were considered the diagnostic specificities, which were defined as 1 minus the false-diagnosis rate (i.e., 97.7% for inpatient stays and 92.0% for outpatient visits, as shown in Fig. 1). In addition, the agreement rate between the first and confirmed diagnoses by disease code was about 75.9% for inpatient stays and only 55.8% for outpatient visits; these values were considered the true-positive rates (diagnostic sensitivities). Thus, to capture the uncertainty associated with diagnostic specificity and sensitivity, 90% CI were calculated for the proportion of diagnostic accuracy (i.e., specificity and sensitivity) (Bluman, 2009). As shown in Figure 1, PERT (Project Evaluation and Review Techniques) distributions were used to capture the ranges (5^th and 95^th percentiles) and most likely values for the proportions of diagnostic specificities and sensitivities.

OPEN IN VIEWER

FIG. 1.

Scheme of data sources and methods used to estimate the number of inpatient stays (hospitalizations), outpatient visits (foodborne disease infections), and patients' experiences (without visiting physicians) of foodborne diseases in South Korea. HIRA, Health Insurance Review and Assessment Service; ICD-10, International Classification of Diseases, 10th Revision; ICD-10 codes are listed in Table 1; PERT, Project Evaluation and Review Techniques.

Results

The 5-year annual average number of cases reported in HIRA from foodborne illness and infectious enteritis (i.e., acute gastroenteritis) in South Korea during 2008–2012 was 192,611±6132, with inpatient hospital stays and outpatient hospital visits estimated at 25,079±1495 and 167,532±8149, respectively (Table 1). From these data and using the method presented in Figure 1, foodborne disease cases caused by contaminated food were estimated at 336,138 (90% CrI: 258,379–430,740), with 2.3% (n=7,809 [90% CrI: 7,016–8,616]), 14.4% (n=48,267 [90% CrI: 45,883–50,695]), and 83.3% (n=280,062 [90% CrI: 201,795–374,091]) accounting for inpatient stays (hospitalizations), outpatient visits (foodborne disease infections), and patients' experiences (without visiting physicians), respectively. An estimated 314,104 cases (90% CrI: 241,456–402,649) were caused by bacterial infections (93.4%), 20,746 cases (90% CrI: 15,111–27,783) by viral infections (6.2%), and 1288 cases (90% CrI: 902–1,762) by parasites (0.4%). An estimated 92,444 (27.5%) cases were caused by 18 known foodborne pathogenic agents, whereas 243,694 (72.5%) cases were caused by miscellaneous (unspecified) (A04.8, A05.8, A08.3, A07.8) (17.4%) and unknown (A04.9, A05.9, A08.4, A07.9) (55.1%) agents (Table 1).

Pathogenic Escherichia coli (ICD-10 code A04.0–4) (87.4 cases per 100,000) caused most of the total foodborne illnesses, followed by nontyphoidal Salmonella spp. (A02.0, A02.8–9), Staphylococcus aureus, hepatitis A virus, and norovirus. However, the leading causes of hospitalizations were norovirus (60.3%), Clostridium botulinum (49.6%), enterohemorrhagic E. coli (46.3%) of pathogenic E. coli, Listeria spp. (46.3%), and hepatitis A virus (44.0%) (Table 2).

Discussion

In this study, the estimated annual number of foodborne disease cases caused by contaminated food in South Korea (attributed to 18 known pathogens and unspecified and unknown agents) was 336,138, including 7809 inpatient stays (hospitalizations), 48,257 outpatient visits (foodborne disease infections), and 280,062 patients' experiences (without visiting physicians), between 2008 and 2012. During the same period, 769 foodborne disease outbreaks affecting a mean of 6773 individuals per year were reported (KMFDS, 2013). This reported figure only accounts for cases identified and reported in South Korea. The estimated total number of cases caused by contaminated food was approximately 49.6 times the reported numbers.

In the current study, the methods used were suggested by Flint et al. (2005) and Hall et al. (2005) to estimate the extent of foodborne illness and included three components: (1) the total number of gastroenteritis cases in the country, (2) the proportion of gastroenteritis caused by foodborne pathogens, and (3) the proportion of experiences of foodborne diseases using survey results. The sum of these three estimates gives the actual burden of foodborne disease. The number of gastroenteritis cases (i.e., foodborne illness and infectious enteritis) in South Korea was determined based on the 37 ICD-10 codes from the HIRA data set. The HIRA data represent the total patient population in South Korea (HIRA, 2014). These HIRA data include the occurrence of foodborne illness and many other infectious diseases and can be used for public health surveillance. Therefore, the number of inpatient stays and outpatient visits is a more accurate indicator of the prevalence of foodborne illness. However, some weaknesses are associated with the HIRA data. The first is the limited accuracy of hospital physician diagnosis, even if sensitivity and specificity are considered. Although Hall et al. (2005) noted that any missed or faulty diagnoses for specific pathogens can be included in an unknown or a miscellaneous category; the current study used probability distributions to account for diagnostic uncertainty and performed Monte Carlo simulations to generate a mean estimate and 90% CrI. A second weakness is that it is unclear from the HIRA data what proportion represents cases in which a patient's illness was actually caused by contaminated food. In addition to contaminated food, transmission of foodborne pathogens to humans may occur from infected persons with secondary transmission, contaminated environments, or animals (Chin, 2000). As shown in Supplementary Table S1, some pathogens, such as Bacillus cereus and Clostridium perfringens, are believed to be transmitted by contaminated food, whereas others such as rotavirus and adenovirus are largely or probably transmitted by a fecal–oral route, with possible contact or respiratory spread through nonfoodborne routes (Koopmans and Duizer, 2004). Most foodborne pathogens have various paths of transmission to humans (Hall et al., 2005); therefore, it is difficult to accurately estimate the proportion of pathogen-specific illnesses resulting from the consumption of contaminated food (Scallan et al., 2011a). However, Hall et al. (2005) suggested that the overall proportion of infectious gastroenteritis in the community from foodborne transmission is estimated by dividing the sum of all cases from known pathogens by the sum of foodborne cases from known pathogens. Assuming that the proportion of gastroenteritis caused by foodborne pathogen transmission is the same internationally, the current study used values estimated from Australian (Hall et al., 2005) and U.S. (Scallan et al., 2011a) data because Korean data were unavailable. However, there are limits applied to these data and in particular for underdeveloped countries. Therefore, in order to take advantage of these data, the environmental and sanitary or food safety conditions in each country should be compared with the reference state (i.e., Australian and U.S. data). In this study, adjustments were made for the proportion estimated as acquired overseas for individuals and unknown pathogens. To reflect this uncertainty, probability distributions and Monte Carlo simulations were used to describe a range of plausible values for foodborne pathogen transmission. Finally, to estimate the number of individuals who experienced foodborne diseases but did not visit the hospital or a physician, we used survey results (Kim et al., 2015) because HIRA only collects hospital (inpatient and outpatient) diagnostic data. However, the proportion of experiences of foodborne illnesses was applied equally regardless of the type of causal pathogens. These survey results were used to adjust for underreporting, but the appropriate values for each pathogen are difficult to estimate. Therefore, further studies are needed to give more robust estimates of the true level of underreporting of the types of pathogens in the community.

In the United States, Scallan et al. (2011a) estimated that 31 known pathogens in contaminated food caused 55,961 hospitalizations annually. An Australian study estimated that 3640 hospitalizations occur annually because of food contaminated with 14 known pathogens (Hall et al., 2005). Our study shows that 18 known as well as unspecified and unknown pathogenic agents in contaminated food caused an estimated 7809 hospitalizations in South Korea. Comparisons among various country estimates are difficult to make, given the inherent differences in methodological approaches and data sources (Thomas et al., 2013). Furthermore, in this study in particular, the number of deaths from foodborne diseases was not included because HIRA data do not include mortality information. Thus, a direct comparison of mortality due to foodborne illnesses with other similar studies is not possible.

Compared with other countries, the main pathogens associated with foodborne diseases in South Korea are different, with E. coli, including enterohemorrhagic E. coli, and nontyphoidal Salmonella spp. representing about 61% of the total pathogen-specific diseases. However, many overseas national foodborne illness estimates, including those from the United States, Australia, The Netherlands, New Zealand, and France, indicate norovirus as a main contributor and Campylobacter spp. and nontyphoidal Salmonella spp. (particularly in the United States, Australia, New Zealand, France, the United Kingdom, and Greece) as the leading bacterial pathogens (Thomas et al., 2013). This difference may vary depending on the specific food culture (Havelaar et al., 2010), eating habits including the consumption of raw or lightly cooked food, and food consumption patterns and amounts (e.g., high intake of kimchi in South Korea, which is a unique traditional Korean food not typically consumed in other countries). In addition, environmental and sanitary conditions including food handling or public health systems for managing food safety policies in each country may be an important cause. The surveillance systems (e.g., lack of testing) and literacy of diagnostic systems operating in each country may influence this difference. On the other hand, the occurrence of foodborne pathogens depends on different time periods. Before 1960, the major causes of gastrointestinal disease were recognized as Salmonella spp., Shigella spp., and C. botulinum, and after the 1990s, they were norovirus, Campylobacter spp., and E. coli O157:H7 (Newell et al., 2010). However, norovirus and Campylobacter spp. infections have been increasingly recognized recently in South Korea (Korean Ministry of Food and Drug Safety, 2013), as in other countries. Therefore, future findings may be altered because of changing patterns in the pathogens responsible for foodborne diseases in South Korea.

Often, many foodborne illnesses are caused by unspecified or unknown pathogenic agents (Tauxe, 2002; Scallan et al., 2011b). Comprehensive diagnostic studies of intestinal infectious diseases indicate that between 50% and 60% of all causative agents are unidentified (Newell et al., 2010). The current study also shows that an estimated 55.1% and 17.4% of cases are caused by unknown and unspecified agents, respectively. In Australia, 30% of illnesses are caused by known pathogens (Hall et al., 2005), which is similar to the current finding of 27.5% in South Korea. Agents of foodborne illness continue to be discovered, such as Campylobacter spp. and E. coli O157, which were unknown only a few decades ago (Tauxe, 2002; Newell et al., 2010; Scallan et al., 2011b). More unknown pathogens are likely to be major causes of illness, some of which will become apparent with time and investigation (Hall et al., 2005).

Conclusions

Some of the methods and data used in the current study differ from those used in other countries (Hall et al., 2005; Scallan et al., 2011a; Thomas et al., 2013). There are not enough laboratory data from routine surveillance such as FoodNet (Jones et al., 2007) in the United States, except for data officially reported on foodborne diseases. Thus, we used the HIRA data as an alternative. Although this study provided an estimation of the annual number of foodborne disease cases caused by contaminated food in South Korea, future studies with high-quality data are needed to further refine foodborne illness estimates. Nevertheless, the current study is a first step toward improving data quality and methods to enhance the robustness of foodborne disease estimations in South Korea. This estimation will be useful for food safety policymakers to more precisely design prevention activities and to control and prioritize the effectiveness of risk management options for foodborne pathogen transmission in South Korea.