Food Safety and Foodborne Disease Surveillance Systems: Proceedings of an Iranian-American Workshop (2006)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Surveillance for Foodborne and Diarrheal Diseases, Including Outbreak Investigations: An American Perspective William E. Keene, Ph.D., M.P.H. Senior Epidemiologist, Acute and Communicable Disease Program, Oregon Public Health Services CASE STUDIES Case studies of three important foodborne disease outbreaks are presented to illustrate how disease surveillance works in the United States and the evolution of outbreak investigative methods. These outbreaks can be reviewed in detail in the original literature (see also Barrett et al., 1994; Bell et al., 1994; CDC, 2004; Cody et al., 1999; Griffin et al., 1994). Outbreak 1 was the 1992-1993 Jack-in-the-Box outbreak of Escherichia coli O157:H7 infections that affected several western states. The outbreak was caused by widespread undercooking of contaminated frozen ground beef hamburger patties at many outlets of the Jack-in-the-Box fast food chain. Although infec- tions from this source began to appear in California and Nevada as early as November 1992, the outbreak was not detected until it reached Washington state in late December. At that time only Washington state was conducting routine surveillance for these infections. In California and Nevada, in contrast, the infec- tion was not reportable, and few laboratories ever used the special media needed to identify the pathogen. Physicians and the general public there were largely unaware of this pathogen; in the absence of recognized outbreaks, there was no publicity and no public education. This outbreak was a landmark event in mod- ern epidemiological history. It had an enormous effect on the public’s perception of the problem of foodborne illness. The political ripples from this event contin- ue to this day. We can quite reasonably talk about foodborne disease epidemiol- ogy in the United States before and after Jack-in-the-Box. The outbreak illus- trates how large outbreaks with many hospitalizations and even fatalities can 31 

32 FOOD SAFETY AND FOODBORNE DISEASE SURVEILLANCE SYSTEMS easily escape notice, investigation, and control in the absence of routine disease surveillance. Outbreak 2, which occurred in 1997, involved E. coli O157:H7-contaminat- ed, unpasteurized commercial Odwalla-brand apple juice. Advances in molecu- lar subtyping of this organism and integration of laboratory subtyping with rou- tine surveillance data made it easy (at least in Washington state) to identify this outbreak. Once the outbreak was identified, the source was quickly identified through traditional case interviews and a case-control study. The outbreak led to changes in the way fresh juices were labeled and processed. Outbreak 3, of Salmonella Enteriditis infections linked to consumption of raw almonds in the United States and Canada, was first identified and investigat- ed in May 2004. But in retrospect, cases may have occurred as early as 2002. The outbreak led to changes in almond processing by industry, which is moving toward a ban on the sale of untreated almonds (i.e., raw almonds or those not processed in a way that would kill pathogens). INTRODUCTION TO DISEASE SURVEILLANCE Public health epidemiology has a number of goals. One is to monitor the health status of the population (e.g., indexes of morbidity and mortality) as well as the contributors to that health status (e.g., access to medical care, level of personal hygiene, and food consumption histories). Another is to manage health crises as they occur, including the investigation and control of disease outbreaks and the emergence of new pathogens (e.g., SARS, avian influenza, and bovine spongiform encephalopathy). Public health epidemiologists are also charged with providing and interpreting available scientific information to help inform public policy decisions. In this context, disease surveillance plays a major role. Surveillance was famously defined by Langmuir (1963) as the “ongoing systematic collection, collation, analysis, and interpretation of data; and the dissemination of informa- tion to those who need to know in order that action be taken.” Surveillance activities involve the collection of raw data, the “cleaning” (correction and stan- dardization) of those data, and the organization and analysis of those data. It is axiomatic in modern public health practice that data are collected in order to be used, and in order to be used they must be disseminated to relevant parties, which may include the medical and academic communities, other public health agencies (domestic or international), policy makers in both public and private sectors, and the general public. In the United States, communicable disease epidemiology and disease sur- veillance practices are largely set by state government agencies, not by the feder- al government. The federal government is a significant source of funding for state programs, and it is a source of logistic and technical support for state agencies, but has surprisingly little authority to investigate disease outbreaks or 

SURVEILLANCE FOR FOODBORNE AND DIARRHEAL DISEASES 33 take control measures directly. Rather, individual states determine which diseas- es must be reported, and the procedures by which such reporting will be done. As a result, while the approach is broadly similar in most states, there are consid- erable variations in public health practice from jurisdiction to jurisdiction. In addition to legal differences, public health agencies vary considerably in fund- ing, staffing levels, and degrees of experience and expertise, and these differenc- es result in different capacities. Reporting practices in the state of Oregon (a relatively large but sparsely populated state on the Pacific coast, with 3.5 million people spread over 250,000 km2) are fairly typical of many states. By law the state public health agency specifies a list of diseases and conditions that must be reported by both physi- cians and laboratories.1 (Bear in mind that—in the United States—medical care is almost entirely in the private sector.) These lists are distributed to laboratories and clinicians on printed posters, in newsletter reminders, through website post- ings, and other media. Most notifiable conditions are specifically named infec- tions or defined conditions (e.g., salmonellosis, campylobacteriosis, listeriosis, hepatitis A, meningococcal infection, and lead poisoning). There are also several catch-all categories, including any suspected common-source outbreak (e.g., multiple people in a group with acute gastroenteritis) or any unusual disease of potential public health significance (e.g., imported exotic diseases such as SARS or avian influenza). Both laboratory-confirmed and suspect or presumptive diag- noses are reportable. As a practical matter, although reporting requirements ap- ply both to physicians and to laboratories,2 in practice most clinicians are not very compliant with reporting laws. The vast majority of reports originate in laboratories. Periodic audits of private laboratories in Oregon indicate that this reporting is reasonably complete (95-100 percent for specified conditions). Of course, the nonspecific conditions (e.g., outbreaks) are aimed at non-laboratory- confirmed cases, so they would have to originate from clinicians, as would pre- sumptively diagnosed cases. Physician reporting of these conditions is relatively incomplete. ABOUT FOODBORNE DISEASE Foodborne disease is surprisingly difficult to define, and some reports may have little if anything to do with food, or with transmission by food. The specif- ic route of transmission for most cases of reported enteric illness is unknown, and the proportion that is transmitted by contaminated food is difficult to esti- mate with precision. “Foodborne” is often used casually to cover almost any 1In the United States, the great majority of medical care and diagnostic laboratory work is done in the private sector. Reporting laws apply equally to the private and public sectors. 2Such that, at least in theory, cases diagnosed based on a specific lab test should be reported twice. 

34 FOOD SAFETY AND FOODBORNE DISEASE SURVEILLANCE SYSTEMS enteric disease, notwithstanding that many illnesses transmitted by food are not gastrointestinal (GI), and that many GI illnesses can have routes other than food, including waterborne, person-to-person, and direct animal-to-person trans- mission. Foodborne and diarrheal diseases can be caused by a wide range of bacteri- al, viral, and parasitic agents. The incidence of these illnesses varies consider- ably around the world, reflecting differing practices in agriculture and animal husbandry, food processing, consumer behaviors, diet, local and regional ecolo- gy, and many other factors. PROCESSING DISEASE REPORTS The legal requirements and mechanisms for reporting potential foodborne diseases are completely merged with reporting laws for all communicable dis- eases. All reports for all diseases funnel through the same public health agen- cies. Depending on the size of those agencies, however, different people may be responsible for tracking different diseases. In Oregon, for example, communi- cable disease epidemiology is divided into three main groups: HIV and other sexually transmitted diseases, tuberculosis control, and everything else (which includes foodborne and diarrheal disease, hepatitis, rabies, meningitis, zoonotic diseases, and hospital infections). Again referring to Oregon practices, individual case reports stream in pri- marily from private laboratories and occasionally from private physicians. By law, reports must be made within one day of a laboratory turning out the relevant result—and, while not perfect, most reporting is indeed quite rapid (within a day or two).3 For most common diseases, case reports result in some type of investi- gation, initially conducted by local health department nurses or environmental health specialists.4 Following disease-specific investigative guidelines developed by state epidemiologists, an attempt is made to interview patients (or their par- ents or other proxy) and collect a variety of demographic, clinical, and epidemi- ological data using standardized, disease-specific forms. These guidelines and forms, as well as other information about reporting practices in Oregon, are available on our department’s webpage (http://www.dhs.state.or.us/publichealth/ acd/disrpt.cfm#forms). Cases include questions about a variety of recognized risk factors for most infections. Persons with nontyphoidal salmonellosis, for example, are asked about 3Pilot projects are under way to allow “instant” electronic reporting directly from major private laboratories, but this is proving a complicated and very expensive proposition. 4Local (county) health departments are the frontline public health agencies in Oregon and most states. Oregon’s 35 counties range in population from < 2,000 to 675,000, and the local health departments vary correspondingly in size and sophistication. 

SURVEILLANCE FOR FOODBORNE AND DIARRHEAL DISEASES 35 consumption of meat and poultry, eggs, unpasteurized milk and cheese, and sprouts; meals away from home; contact with reptiles (unfortunately, rather pop- ular pets in the United States and a regular source of these infections); other pets, livestock, sick and incontinent people; and recent travel. The time period of interest for these questions corresponds to the likely exposure period, which is derived from the patient’s date of onset and the pathogen’s typical incubation period. Thus, cases with E. coli O157:H7-infected patients might be asked about the period from 1 to 10 days before onset (with emphasis on 2 to 6 days), while salmonellosis cases would consider the 1 to 5 days before onset. A “yes” answer to one or more question does not, of course, confirm the source of infection. Most individuals report multiple potential sources—and no doubt forget or ne- glect to mention other possibilities. In the short term these responses are used to help frame various health education messages (e.g., about the risks of pet reptiles or the importance of good hand-washing behavior). Exposure data are consid- ered more systematically if suspicions of common-source outbreaks are raised. These interviews are generally conducted by telephone. Local health department staff generally find telephone numbers by contacting the patient’s physician, who is named on the laboratory report. Most individuals are relatively easy to contact, but there are exceptions, of course. Only a very small number of indi- viduals refuse to cooperate with these interviews. Case reports are forwarded (usually by fax, or increasingly, electronically) from local health departments to the state public health department as soon as interviews are concluded. This is typically within 2 to 3 days of the initial labo- ratory report, and often within 24 hours. Even if the patient cannot be contacted, a report is filed with as much information as possible (e.g., demographic infor- mation obtained from the clinician). Surveillance data are entered into a customized database at the state level. The Oregon database currently includes case reports back through 1988—some 75,000 as of August 2005. State epidemiologists have immediate access to both individual case reports and the ability to generate on-demand generic or ad hoc summary reports (e.g., the number of reported cases of Salmonella serotype Newport affecting males between the ages of 15 and 35 in counties X, Y, and Z during each of the past 15 years). This information is critical to assessing what is normal and what might constitute an aberration (e.g., an outbreak). To work effectively, these data must flow quickly. For most enteric infections the lag from first laboratory identification to local health department notification to in- vestigation to report to state authorities is 2 to 4 days. THE ROLE OF THE PUBLIC HEALTH LABORATORY Communicable disease epidemiology requires partnerships with effective public health laboratories (PHLs). Every state has a PHL, which by law is the primary reference laboratory for public health-related microbiology, virology, 

36 FOOD SAFETY AND FOODBORNE DISEASE SURVEILLANCE SYSTEMS and other areas. Public health laboratory staff and epidemiologists work together closely. The capacities of these laboratories vary considerably, and they are sup- plemented by formal and informal collaborations with neighboring state labora- tories, as well as laboratories of the U.S. Centers for Disease Control and Pre- vention (CDC). In Oregon, diagnostic laboratories are required not only to report but also to forward (at their own expense) isolates of specified pathogens5 to the PHL. These specimens usually arrive within a few days of initial isolation. The identi- ty of these pathogens is confirmed at the PHL, and many of the isolates are now being subtyped by a combination of traditional and molecular methods. Although rarely of any clinical importance (and hence not attempted by for- profit private laboratories), subtyping is of critical importance for epidemiology. For example, all Salmonella isolates are serotyped by the Kaufmann-White scheme; all Shigella and Vibrio isolates are speciated. We currently use national- ly standardized methods to subtype all isolates of E. coli O157, Salmonella, and selected Shigella received by pulsed-field gel electrophoresis (PFGE). Subtyping proceeds as quickly as staff time allows, with most results being available within 2 to 8 days of receipt. Laboratory subtyping data are added to the epidemiologi- cal case database. Molecular subtyping data are also shared with other laborato- ries around the country through CDC’s national PulseNet network (http:// www.cdc.gov/pulsenet/). States that do not get this kind of laboratory data, or who cannot get it quickly, are rarely able to identify or solve the epidemiologic puzzles that come along. DATA LIMITATIONS Official statistics only reflect numbers of reported cases, and they are an incomplete and imperfect index of disease incidence. Many factors affect the likelihood that an infection will be reported. Asymptomatically infected individ- uals, for obvious reasons, are unlikely to be identified, as are symptomatic indi- viduals who, for whatever reason (e.g., mild symptoms, lack of health insurance, inconvenience), do not seek medical care. Even if they do see a physician, they must be given an appropriate test (e.g., a stool culture) and they must test posi- tive. Depending on the disease, probably only a minority—and for most enteric illnesses, probably only a small minority—of infected persons are reported. The numbers are not only reduced, but reported cases are a biased sample of infected persons—biased to include those most likely to be tested or diagnosed (e.g., hospitalized individuals, others with more severe illness, very young children, and those with better health insurance). Surveillance statistics must always be interpreted with caution. 5Including all Salmonella, Shigella, E. coli O157, Vibrio, Yersinia, and Listeria isolates, to name the potentially foodborne agents. 

SURVEILLANCE FOR FOODBORNE AND DIARRHEAL DISEASES 37 ABOUT OUTBREAKS Based on identified epidemiological connections to other cases, we infor- mally classify enteric disease reports into one of three categories: sporadic, household, or outbreak. Sporadic cases have no recognized epidemiological links to any other cases. Household cases are linked to other illnesses within the same household but not elsewhere. Outbreak cases are epidemiologically linked to cases in other households. Thus, as few as two cases may be sufficient to define an outbreak, given adequate epidemiological information.6 Cases are reclassi- fied as new information becomes available. While outbreak cases gather much of the attention and most of the publicity, the great majority of enteric disease case reports are sporadic. For example, although we have investigated more than 30 outbreaks of E. coli O157 infections since 1990, totaling 430 reported cases, most cases reported since then—1110 (63 percent) of 1743 as of October 2004—are considered sporadic. Almost by definition the causes of sporadic cases are unknown. Even in the aggregate (e.g., large and expensive FoodNet case-control studies), most analyses fail to explain many new exposures of great importance. We investigate outbreaks for a number of reasons, some of which are obvious: • To stop ongoing transmission (e.g., from a contaminated commercial product that is still available in stores or in consumers’ homes); • To facilitate diagnosis and proper treatment or prophylaxis in the setting of a community outbreak; • To identify risk factors for infection (e.g., consumption of unpasteurized milk or alfalfa sprouts, or recent antibiotic use); • To stimulate research (e.g., when outbreaks raise questions regarding food microbiology or consumer behavior); or • To provide the information necessary to develop sensible long-term pre- vention strategies. Outbreak investigations also provide important training opportunities for public health workers and a chance to develop and test new investigative ap- proaches. Even experienced staff need to keep in practice. Outbreak cases and outbreak investigations have proven to be dispropor- tionately important to public health practice (Keene, 1997). Given a successful investigation, the specific sources of transmission can be identified, and risk factors associated with pathogen amplification and transmission (e.g., food han- 6Larger outbreaks tend to get more attention than very small ones. Very small outbreaks (< four cases) may get only a cursory review, depending on circumstances. 

38 FOOD SAFETY AND FOODBORNE DISEASE SURVEILLANCE SYSTEMS dling practices or time or temperature abuse) can be confirmed. Given an identi- fied exposure time, we can calculate incubation periods. Some outbreaks provide useful natural experiments, which can provide insight into important data, such as infectious dose or host risk factors for illness (e.g., immunocompetence, age, or concurrent medication). Not least of all, outbreaks can provide an opportunity for health education that is amplified through the private media. (In the United States, for example, outbreaks often provide the only context for public health representatives to have access to television news, which is privately controlled.) Outbreak investigations typically begin with one of two scenarios. The most common, which we can call “Type 1,” begin when a private citizen (or, less often, a physician or infection control nurse) contacts the local health department to report that “a bunch of people became sick” after some event (e.g., a wedding, a shared restaurant meal) or that illness rates seem abnormally high at some institution (e.g., a prison, school, camp, or nursing home). These clusters are most commonly caused by Norwalk-like viruses (noroviruses), and less often by other infections or intoxications (e.g., Clostridium perfringens, Bacillus cereus, Staphylococcus aureus, or scombroid). Laboratory confirmation of an etiology generally depends on public health resources, both for specimen collection and for testing. Testing for Norwalk-like viruses—by far the most common cause of identified outbreaks, and perhaps sporadic gastrointestinal illness as well—is essentially unavailable in the private sector. Diagnosis (by polymerase chain reaction) only occurs in public health labs, and often only in the case of out- breaks. The second category of outbreaks, which we can call “Type 2,” begin as scattered routine surveillance reports of laboratory-confirmed cases (e.g., salmo- nellosis, E. coli O157). It is usually not apparent to the patient (or the physician) that these cases are part of a cluster. Instead, this becomes apparent only when epidemiological and laboratory data are pooled over an extended area (multiple counties, multiple states) and time period (days to months). For example, more than one case with an uncommon Salmonella serotype or E. coli O157 PFGE pattern might be reported within a short time period. These investigations may begin slowly, but quickly become high-priority investigations, often involving multiple epidemiologists and complicated liaisons with other public health agen- cies inside and outside the state (e.g., other state health departments, CDC, and state and federal regulatory agencies, such as the Food and Drug Administration [FDA] or the U.S. Department of Agriculture). Finding the source of outbreaks can be quite challenging, particularly if states do not have enough cases locally to develop specific hypotheses regarding possible sources. Many investigations are unsuccessful. Delayed reporting and a failure to obtain clinical specimens, such as stool samples, are common prob- lems. Coordination between agencies is sometimes excellent, and sometimes poor—often more a matter of personalities and chance than we would like to admit. The CDC often plays a useful role in coordinating efforts between states, 

SURVEILLANCE FOR FOODBORNE AND DIARRHEAL DISEASES 39 and can provide technical assistance if needed. Sometimes the CDC is bogged down in internal bureaucratic procedures and is more of a hindrance than a help. Type 2 outbreaks demand regular and timely review of surveillance data. In Oregon these lab data are monitored by epidemiology staff throughout the day. Reporting totals must be interpreted in the context of historical norms. Outbreak cases are sometimes first recognized as such following reports or inquiries from public health agencies outside our state. We have almost daily contacts by e-mail or telephone with our counterparts in neighboring states, and less often with other epidemiologists around the country and in Canada. E-mail networks and Listservs provide a fast, convenient way to query other public health agencies. For example, someone might post an e-mail notice saying: “Our state seems to be getting more than our usual number of Salmonella Braenderup’s over the past few weeks. Are you seeing any in your state?” This kind of inquiry would then prompt a comparative review of epidemiological and laboratory data, and poten- tially we would join with the other state(s) to look for a source using common questionnaires and methods. Once primarily done just by epidemiologists, nowa- days these kinds of inquiries go from laboratory to laboratory, from laboratory to epidemiologist, from epidemiologist to laboratory, as well as from epidemiolo- gist to epidemiologist. Molecular subtyping data have become critical to these efforts over the last five years in the United States. OUTBREAK INVESTIGATIONS While following certain general patterns, each outbreak investigations is unique. In Oregon, local health agencies take the lead in the investigation and control of most Type 1 outbreaks, which tend to be less complicated. Investiga- tions typically involve both public health nursing staff and environmental health specialists—the same people who conduct routine inspections of restaurants, swimming pools, and child care establishments (among other duties). Different places have a different array of licensing and regulatory agencies that cover wholesale and retail establishments, including restaurants, food processing plants, markets, bakeries, schools, hospitals, and nursing homes, and investigations de- mand coordination between these agencies. Epidemiologists take the lead in these investigations, which are usually collaborative efforts involving environ- mental health, nursing services, regulatory agencies, and others.7 (In the United States, epidemiologists almost never have any regulatory authority; we have little if any enforcement power, nor do we issue fines or penalties of any kind.) Type 1 outbreaks tend to reflect local problems—most commonly poor food handling practices or inadequate worker hygiene. Despite our best efforts, expe- rience suggests that such outbreaks are easier to explain than they are to prevent. 7At least we epidemiologists like to think we’re in charge…. 

40 FOOD SAFETY AND FOODBORNE DISEASE SURVEILLANCE SYSTEMS State epidemiologists provide technical assistance in questionnaire development and data analysis. If local staff are overwhelmed, state personnel help with inter- views and occasionally will go into the field. We try to encourage our staff to go into the field more often, as it is by far the best way to investigate outbreaks; but competing demands from other projects, family responsibilities, and budget con- straints can limit these opportunities. We have developed a system of templates that Oregon epidemiologists use to quickly develop outbreak-specific questionnaires, data entry databases, and data analysis programs.8 These templates save a great deal of time and help us focus more efforts on data collection and interpretation and less on raw data shuffling and preparation. Given enough cases, we usually attempt some type of cohort or case-control study to determine specific source(s) of infection. Oregon state epidemiologists are always available to consult with local health departments and offer technical advice, often designing questionnaires and performing data analysis after interviews conducted by local staff. One of a pool of staff epidemiologists is always available on call outside normal office hours and if necessary, staff are usually available to travel on short notice to support or direct investigations in the field. Disseminated, surveillance-anomaly- driven Type 2 outbreak investigations are almost always run by state epidemiol- ogists, as they require more sophisticated epidemiological training and experi- ence, as well as (usually) much more liaison between agencies inside and outside Oregon. Besides the official agencies, we sometimes enlist the assistance of persons in academic or other institutions who may have special expertise (e.g., to conduct experimental tests not available at PHLs). In some states with fewer resources, federal epidemiologists from the CDC (usually trainees with close supervision by more experienced staff in Atlanta) play these supporting roles more often. At the outset of these investigations we often have no idea what connects the cases, other than their illness. We scrutinize the routine reports that we typi- cally already have on these individuals for demographic or behavioral clues (e.g., similar ethnic backgrounds, unusual age or sex distributions, attendance at the same mass event or different outlets of the same restaurant chain). Failing any immediate success from such a review, we will re-interview cases with hypothesis-generating questionnaires about recently consumed food.9 Of course, in the meantime we compare notes with our counterparts in nearby states, as already mentioned. This type of case finding is often critical to uncovering enough cases to yield statistically meaningful sample sizes. As specific hypothe- ses emerge we may conduct case-control or other studies to test them. 8See http://www.healthoregon.org/acd/keene.cfm for a more extensive description of these and related tools. 9Again, refer to http://www.healthoregon.org/acd/keene.cfm and particularly our “shotgun” ques- tionnaire. 

SURVEILLANCE FOR FOODBORNE AND DIARRHEAL DISEASES 41 Outbreak investigations should be treated as matters of extreme urgency. Staff schedules are rearranged as needed, and staff may work evenings and week- ends until the source is identified. It is often easier to reach people by telephone in the evening, for example, so we have to be able to call when our targets are available, no matter how inconvenient it may be for us. Moreover, delayed re- porting is the rule, not the exception, and memories can quickly fade. Case-control methods vary, but we often target households with matching telephone prefix numbers as controls for cases.10 Once initiated, these kinds of studies can usually be completed quickly, and rarely take more than 24 to 48 hours. We emphasize speed, because our basic premise is that the problems may be ongoing. Delays could mean additional morbidity and mortality. Once a source is identified, we may advise the public through press releases if there is a public health purpose to be served by doing so (e.g., recall of con- taminated product). If dangerous products are identified that may still be avail- able to consumers (e.g., ground beef contaminated with E. coli O157:H7), press releases are issued without delay. Decisions to issue press releases are usually made by senior public health staff, but without political input. While we are aware of the adverse consequences of such publicity on business interests, this is usually given little if any consideration. We work closely with regulatory agen- cies on product recalls, but often we have to keep prodding them to move as quickly as we epidemiologists can. In part this reflects the more formal legal structure at the federal level of, say, an FDA-instigated product recall. It is important to note that confirmation of an outbreak source only rarely involves recovery of a pathogen from an implicated product. In fact, culture of food products typically plays a very small role in outbreak investigation and is often all but irrelevant to the conduct of the investigation. Contaminated foods are often long gone before outbreaks are even recognized, or contamination may occur at such a low and intermittent level as to be very difficult to detect. Confir- mation means establishment of strong epidemiological evidence linking con- sumption of a product to illness, corroborated by biological plausibility and usu- ally compatible information about product distribution, even if not (yet) having convincing product trace-back or trace-forward information. More detailed in- formation about product distribution is usually sought as soon as possible, but in general we do not delay public notification for such information unless it is necessary to confirm the source. Food testing is often attempted after the fact, and is sometimes successful (more often not), but we never wait for such results to notify the public if the epidemiological evidence is compelling. For many years regulatory agencies in the United States were very uncom- 10For example, if the case’s number is 503-731-4024, we might start with 503-731-4025, and continue -4026, -4027, . . . until we found willing respondents who were eligible (e.g., not recently ill, similar age bracket). 

42 FOOD SAFETY AND FOODBORNE DISEASE SURVEILLANCE SYSTEMS fortable with this—often to the infuriation of epidemiologists—but over the last 5 to 10 years regulators have become much better about proceeding based on epidemiological evidence alone. The credibility of epidemiologists is on the line every time we conduct these studies and draw conclusions from them, so this work must be done carefully and thoroughly to avoid either false negative or (usually worse) false positive results (i.e., implicating a product that had nothing to do with the illness). Even in our state of only 3.5 million people, we lead or participate in these types of studies (i.e., Type 2) about once or twice each month. In just the last year we have used these approaches to link salmonellosis outbreaks to produce (alfalfa sprouts and honeydew melons), almonds, and processed food (commer- cial egg salad being distributed at a chain of grocery stores in several western states). Type 1 investigations begin much more frequently—about two or three per week—in our state, with an investigation resulting in some kind of cohort or case-control study perhaps once a week. This means that our staff can obtain considerable experience developing questionnaires and carrying out these inves- tigations under realistic time pressures. While many investigations are relatively inconsequential, this experience is invaluable as training. In addition, it can be difficult to predict at the outset which investigations will be the most consequen- tial. We have made highly significant discoveries in the course of investigations of very small clusters (Keene, 1997). OUTBREAK REPORTING We ask our state or local health department epidemiologists to summarize outbreak investigations, using a standardized questionnaire.11 This practice is state-specific, and many states do not track investigations as completely as we do. In addition to paper records, including questionnaires, outbreak summaries are logged into a computer database. Foodborne and waterborne outbreaks are routinely reported to the CDC (nationally). Interesting or instructive outbreaks are sometimes presented at national meetings or written up for publication in peer-reviewed journals. We average about three to five such outbreaks per year. Fifteen to 20 outbreaks each year merit at least local or regional attention. With increased funding and staffing over the last 10 years, our department has become much better at investigating outbreaks and in particular at logging investigations when they occur. The number of outbreaks logged annually has increased from less than 10 in 1995 to over 180 in 2004. The great majority (85 percent) represent clusters of acute gastroenteritis. About one-half of these out- 11The Foodborne Outbreak Summary form is posted on our Web page (http://www.dhs.state.or.us/ publichealth/acd/foodrpt.cfm). There are similar forms for nonfoodborne outbreak investigations. 

SURVEILLANCE FOR FOODBORNE AND DIARRHEAL DISEASES 43 breaks are clearly foodborne, and a significant proportion of the rest had indeter- minate routes of transmission (i.e., may have been at least in part foodborne). As previously noted, the most commonly identified etiology by far for reported outbreaks in Oregon are Norwalk-like viruses (noroviruses). Since our PHL be- came able to assay stool specimens for Norwalk by reverse transcription-poly- merase chain reaction in 1999, the number of these outbreaks logged has steadily increased. Multicounty and multistate outbreaks (typically salmonellosis) are investigated with some regularity. We encourage reporting through whatever channels possible. We assure local health department staff that the more they report, the better we think they are doing. One of the biggest hurdles is convincing people in medical institutions and local health agencies that outbreaks do not reflect badly on them; indeed, it is just the opposite. We know that common-source clusters are occurring with great regularity everywhere and that we hear of only a minority that take place. So the better the surveillance is, the more outbreaks we will hear of. It’s that simple. To return to Langmuir’s original dictum, we must appreciate that surveil- lance data do not originate in a vacuum, and that we cannot long succeed without providing feedback to our data sources and other collaborators. We go to consid- erable lengths and expense to maintain contacts with the community of laborato- ry scientists, clinicians, infection control practitioners, and public health nurses who provide us with our raw data. We mail a biweekly newsletter (http:// www.dhs.state.or.us/publichealth/cdsummary/index.cfm) to all licensed physi- cians in the state; the newsletter provides information and news of epidemiolog- ical interest and in general reminds clinicians that we exist. We also regularly speak to medical, academic, and lay audiences around the state. We present outbreak reports and other topics at scientific meetings and prepare manuscripts for publication in peer-reviewed medical journals. We have regular contact with the news media concerning outbreaks and other developments. There is an obvi- ous feedback loop at work here, as news about outbreaks and other surveillance data generates interest and heightens awareness, which in turn improves report- ing, which in turn facilitates the identification of new outbreaks, which in turn generates more news. If outbreaks are not publicized, or the public is not made aware of the function, existence, and value of public health investigations, few outbreaks will be identified. Although not covered in this presentation, we also participate in collabora- tive research projects concerning foodborne and diarrheal disease under the Food- Net umbrella.12 Oregon has been a FoodNet site since the program’s inception in 1995. FoodNet sites have special grant funding from the federal government that enhances epidemiological capacity. FoodNet states conduct all the usual 12See http://www.cdc.gov/foodnet/ for more information on FoodNet. 

44 FOOD SAFETY AND FOODBORNE DISEASE SURVEILLANCE SYSTEMS functions of state health departments, but also participate in special multistate projects aimed at identifying causes of foodborne and diarrheal disease. These projects are often very resource intensive. These kinds of projects presuppose well-established, mature surveillance networks. CONCLUSIONS Surveillance for foodborne and diarrheal disease is a complex and collabo- rative effort that involves laboratory, environmental health, and epidemiological resources. Regulatory, industry, and academic agencies also contribute to the process. Surveillance protocols in the United States depend on clearly defined legal responsibilities and authorities, generally specified at the state level, which give selected public health agencies access to otherwise confidential medical information, and ensure the availability of specimens for specialized character- ization for epidemiological and other purposes. Mandatory reporting of selected diagnoses and laboratory test results is a pillar of our system, which in turn depends on at least a significant number of people seeking medical care for their illnesses and ending up being cultured or otherwise tested to determine a specific etiology. Meaningful follow-up to dis- ease reports depends on rapid communication between public health officials, physicians, and patients. Data are collected using standardized instruments and rapidly pooled for analysis at local, state, and national levels. Informal and un- structured contacts between state and national public health agencies are among the most important means of pursuing potential common source outbreaks that may involve multiple jurisdictions. Trained epidemiologists are available to re- spond to presumptive anomalies, including outbreaks of disease. Database de- velopment and maintenance are important considerations for both routine case reporting and outbreak investigations. Epidemiologists, laboratorians, regulators, and academics have attempted to integrate reports from outbreak investigations and other surveillance data, labo- ratory characterization of pathogens, and food sampling programs conducted by regulatory agencies and industry to achieve a “big picture” of the causes of foodborne disease. While progress has been made, these efforts have proven frustratingly inconclusive. Large and expensive population-based case-control studies of sporadic cases, for example, often end up explaining relatively few cases. Outbreak investigations, while often definitive, are difficult to extrapolate from. Routine epidemiological data are often biased—sometimes seemingly hopelessly so. Questions of attributable risk (e.g., How much campylobacteriosis comes from undercooked poultry? How important is poor hand washing by res- taurant workers to disease transmission? Is imported produce more of a problem than domestic produce?) remain largely unanswered. There is no consensus about how to solve these problems, or even agreement that they can be solved. Despite all these problems, public health agencies probably deserve much of 

SURVEILLANCE FOR FOODBORNE AND DIARRHEAL DISEASES 45 the credit for stimulating changes in food processing and handling practices at both commercial and consumer levels, not to mention improving the quality of related medical care. Good surveillance data can be used to assess temporal trends in the incidence of foodborne and diarrheal disease, and most indicators suggest that a number of them may be declining in the United States over the last few years. Of course, foodborne-illness-associated morbidity and mortality are greatly reduced from levels seen 100 years ago, reflecting improved hygiene and sanitation at all levels. The American system of disease surveillance is expensive, complicated, and not at all an idealized model. It can be maddeningly bureaucratic and inefficient. At the same time, there is a great tradition of hard and often productive work that often translates to rapid identification, investigation, and resolution of public health problems with consequent prevention of unnecessary morbidity and mor- tality. It is a tradition that we are happy to share with our counterparts in Iran in hopes that this will help stimulate the development of your own protocols and traditions, in turn benefiting the health of the Iranian people. REFERENCES Barrett, T. J., H. Lior, J. H. Green, R. Khakhria, J. G. Wells, B. P. Bell, K. D. Greene, J. Lewis, and P. M. Griffin. 1994. Laboratory investigation of a multi-state food-borne outbreak of Escheri- chia coli O157:H7 by using pulsed-field gel electrophoresis and phage typing. Journal of Clin- ical Microbiology 32:3013-3017. Bell, B. P., M. Goldoft, P. M. Griffin, M. A. Davis, D. C. Gordon, P. I. Tarr, C. A. Bartleson, J. H. Lewis, T. J. Barrett, J. G. Wells, R. Baron, and J. Kobayashi. 1994. A multi-state outbreak of Escherichia coli O157:H7-associated bloody diarrhea and hemolytic uremic syndrome from hamburgers: The Washington experience. Journal of the American Medical Association 272:1349-1353. CDC (Centers for Disease Control and Prevention). 2004. Outbreak of Salmonella serotype Enteriti- dis infections associated with raw almonds—United States and Canada, 2003-2004. Morbidity and Mortality Weekly Report 53:484-487. Cody, S. H., M. K. Glynn, J. A. Farrar, K. L. Cairns, P. M. Griffin, J. Kobayashi, M. Fyfe, R. Hoffman, A. S. King, J. H. Lewis, B. Swaminathan, R. G. Bryant, and D. J. Vugia. 1999. An outbreak of Escherichia coli O157:H7 infection from unpasteurized commercial apple juice. Annals of Internal Medicine 130:202-209. Griffin, P. M., B. P. Bell, P. R. Cieslak, J. Tuttle, T. J. Barrett, M. P. Doyle, A. M. McNamara, A. M. Shefer, and J. G. Wells. 1994. Large outbreak of Escherichia coli O157:H7 infections in the western United States: The big picture. In Recent Advances in Verocytotoxin-Producing Es- cherichia coli Infections, eds. M. A. Karmali and A. G. Goglio, pp. 7-12. Amsterdam: Elsevier. Keene, W. E., E. Sazie, J. Kok, D. H. Rice, D. D. Hancock, V. K. Balan, T. Zhao, and M. P. Doyle. 1997. An outbreak of Escherichia coli O157:H7 infections traced to jerky made from deer meat. Journal of the American Medical Association 277:1229-1231. Langmuir, A. D. 1963. The surveillance of communicable diseases of national importance. New England Journal of Medicine 268:182-192. Shefer A. M., D. Koo, S.B. Werner, E. D. Mintz, R. Baron, J. G. Wells, T. J. Barrett, M. Ginsberg, R. Bryant, S. Abbott, and P. M. Griffin. 1996. A cluster of Escherichia coli O157:H7 infections with the hemolytic-uremic syndrome and death in California: A mandate for improved surveil- lance. Western Journal of Medicine 165:15-19.