From February 1998 to January 1999, 268 incidents were recorded, compared with 31 to 44 incidents reported yearly between 1993 and 1997. Of the 268 incidents, 183 (68%) were single cases and 85 (32%) were outbreaks (Table 1). Collectively, 515 cases of foodborne illness were documented; interviews were conducted for 354 of these; 61% were women. Each week during the study period, 1 to 21 incidents were reported, but no obvious trend over the year was apparent (data not shown).

For 101 (38%) of the 268 incidents, 123 fecal samples were collected. A microbiologic agent was detected in 47 (38%) of the 123 fecal samples and in 45 (45%) of the 101 sampled incidents (Table 2).

In most incidents, no relevant food samples were available for analysis. In about one third of the 66 incidents in which food was sampled, potentially pathogenic microorganisms such as Bacillus cereus, E. coli, and Staphylococcus aureus were detected at levels between 1.3-5.9 log CFU g^-1 (Table 2).

In 213 (79%) of the 268 incidents, the etiologic agent was unknown because of insufficient information (Table 2). These incidents involved 334 (65%) of the 515 documented, single and outbreak-associated, cases. Bacteria caused 26 (10%) of the 268 incidents and 128 (25%) of the 515 documented cases; viruses caused 24 (9%) of the 268 incidents and 45 (9%) of the 515 documented cases (Table 2). The most common etiologic agent was calicivirus (20 incidents, 41 cases), followed by Campylobacter spp. (12 incidents, 16 cases) and S. aureus (5 incidents, 99 cases, Table 2).

Of the 268 incidents, 76 (8 single cases and 68 outbreaks) were classified as verified or probable (Table 3). These incidents resulted from ingestion of food prepared in a) restaurants or similar establishments (46%); b) homes (8%); c) grocery stores (4%); d) other places (2%); and e) unknown places (40%). The source of contamination was unknown in 27 incidents involving 58 ill persons. In 10 incidents (25 ill persons), only the meal, not the specific food item, was implicated. Meat and meat products (13 incidents, 34 ill persons) and mixed dishes (12 incidents, 128 ill persons) were the two most implicated food categories; poultry (5 incidents, 10 ill persons), beef (4 incidents, 7 ill persons), and sandwiches (4 incidents, 112 ill persons) were the most implicated subcategories.

One or more contributing factors could be identified in 18 (24%) of the 76 verified and probable incidents. These factors were a) lack of hygiene in processing, preparing, storing, and handling food (11 incidents); b) temperature errors, i.e., inadequate refrigeration, cooking, or cooling (8 incidents); c) contamination by an infected person or equipment (5 incidents); d) cross-contamination from other products, ingredients, or the environment (4 incidents); e) contaminated raw food (2); and f) other factors (6).

Most (85%) of the 268 incidents were detected by a telephone call from an affected person, 13% (36/268) were detected through medical authorities, and 2% (5/268) were detected through other sources. However, in several incidents the caller had been in previous contact with the medical authorities.

The illness forced 122 (79%) of the 154 employed or self-employed patients to miss work (Table 4). Most patients never contacted medical authorities. In total, 45 (14%) of the 312 respondents included in this analysis had visited a doctor and 19 (6%) were hospitalized (Table 4). The average cost for a case of foodborne illness was estimated to be 2,164 Swedish Krona (SEK) ($246), of which 1,027 SEK ($116) were direct costs including doctor visits, hospitalization, and medicine; the rest were indirect costs (i.e., loss of production) (Table 4). The average expenses to a patient were estimated at 500 SEK ($57), mostly from loss of income. For a hospitalized patient, the average cost was 18,652 SEK ($2,117), time spent in the hospital was 3.1 days, and loss of production was 5.6 days.

Of the 266 respondents in the retrospective interview study, 10 persons (3.8%) suspected they had had a foodborne illness during the study period. This translates to an incidence of 38 illnesses per 1,000 inhabitants per year. Only 1 of the 10 affected respondents (10%) had called the municipal authority. Based on this degree of reporting, the actual number of foodborne incidents was estimated to be 2,700 (268/0.1) When the average number of illnesses per incident (1.9) was used, the number of illnesses per year was calculated to be 5,100 (1.9 x 2,700). This translates to an incidence of 28 illnesses per 1,000 inhabitants per year (5,100 illnesses / 186,000 inhabitants), which is in the same range as the first estimate.

Enhanced surveillance in combination with a telephone interview study was used to improve our understanding of foodborne illnesses and to address three limitations commonly named in foodborne research (16): 1) underreporting; 2) lack of data on the incidence and severity of foodborne illnesses; and 3) lack of medical cost data on foodborne illness episodes, including those for which no medical care is sought. Surveillance was enhanced by improving some of its preconceived weaknesses, i.e., the awareness of foodborne illnesses and motivation to report them on the part of consumers and physicians and surveillance activities of the health authorities (8). This approach has both strengths and weaknesses. The advantages include the theoretical size of the study population (186,000 people), which makes the detection of rare disease agents possible, and the detection and investigation of single (17) and milder cases (3), not only outbreak-associated cases or cases occurring in persons who seek medical attention. Study limitations include the difficulty of defining the actual size and composition of the study population and establishing a case definition, which is partly based on suspicion. These limitations were addressed by conducting the retrospective telephone interview study and by classifying incidents on the basis of available evidence (Table 3).

Public attention raised by the study could have led to a shift from the normal underreporting to overreporting. However, the telephone interview study indicated that a bias towards underreporting still existed. Further, the use of this degree of reporting and the number of illnesses per incident yielded an annual incidence estimate that was in reasonable agreement with the first estimate based on the telephone interview study (28 and 38 illnesses, respectively, per 1000 inhabitants). The uncertainty of the second estimate is probably greater than the first since it is based on only 10 suspected cases, and the average number of illnesses per incident is probably an underestimate because of the large proportion of single cases.

The annual incidence estimates obtained from our telephone interview study and from a national interview study (11) were also in reasonable agreement, 38 compared to 79 illnesses per 1,000 inhabitants. In comparison, 6.5 to 33 million cases of food-related illness per year have been estimated to occur each year in the United States (18). This translates to an incidence (25 to 130 cases per 1,000 inhabitants) similar to that in our study. A more recent report estimated the annual U.S. incidence of foodborne illnesses to be 278 cases per 1,000 inhabitants (4). However, caution should be exercised when comparing incidence estimates from different studies since they may partly reflect differences in the surveillance systems used and the assumptions behind the estimations.

Based on the number of foodborne illnesses reported after this study (130 and 100 incidents reported in 1999 and 2000, respectively), improved detection appears to be persisting. The average annual number of incidents in 1993-97 was 40, which (by using the estimated actual number of incidents, 2,700) indicates underreporting by a factor of 67 (2,700/40).

Based on data from 1987 (19), the cost per case of salmonellosis in Sweden can be estimated at $1,322 (converted from United Kingdom Pounds), which is much higher than our estimate for the cost per case of foodborne illness (Table 4). Our lower estimate is not unexpected since it is based on illnesses caused by a variety of agents and a spectrum of symptoms, from mild to more severe. Comparing costs between countries is difficult since the methods, types of illnesses, and health-care systems may differ. Razem and Katusin-Razem (20) estimated the cost per case of salmonellosis in Croatia to be $284 by adjusting estimates from different countries based on the ratio of their gross national products. In New Zealand (21), the estimated cost per case of foodborne infectious disease, $200, was in the same range as our estimate, whereas a considerably higher cost, $1,250, was estimated for a case of foodborne illness in the United States (22). The New Zealand estimate, however, was based on infectious diseases only, and the second estimate included costs for business losses, deaths, legal settlements, and investigation (22). Our estimate did not include the latter costs nor costs resulting from potential medical sequelae (5,6) and personal consequences not usually estimated in monetary terms (3). By combining the present cost per illness with the previously estimated 500,000 cases of foodborne illness per year in Sweden (11), the costs to society can be estimated at 1,082 million SEK ($123 million).

Both the present data and those from voluntary reports from the local authorities to the National Food Administration (10) indicate that a substantial proportion of foodborne illnesses occur because of mistakes in or a lack of knowledge of food-handling procedures at commercial food establishments. Another similarity is the relatively large proportion of incidents with unknown causes. These comparisons indicate that the surveillance system gives useful information but also has several limitations (2). A lower proportion of incidents in which Salmonella spp. was implicated was found in this study (Table 2), compared to other reports of foodborne illnesses both in North America (8) and in Europe (23). It is not likely that this result is due to a sampling bias since fecal samples were analyzed for salmonellae in the same frequency as for caliciviruses and Campylobacter spp. Instead, it may reflect the low prevalence in Sweden of salmonella in food, cattle, pigs, and poultry (<1%) because of an extensive control program (10).

The study failed to establish a rapid link between the physicians and the municipal public health inspectors as indicated by a review of communicable diseases reported to the CMO during the study period. Indeed, at least 27 incidents, which according to the public health inspectors were possible foodborne illnesses, were not reported. These incidents were not part of the study. This means that serious incidents may go undetected and that valuable time is lost since the CMO generally receives notification more than a week after the incident occurs. Also, the number of incidents forwarded from medical staff in the different areas in the municipality varied. This may reflect a true difference in incidence but more likely is a reflection of their motivation to participate in the study. Guzewich et al. (9) stressed the importance of the motivation of those involved in the surveillance and suggested improved feedback to stimulate motivation.

On the basis of the results and experiences obtained during our study (12), several suggestions to improve detection of foodborne illness incidents can be proposed. First, information should be directed to the general public and the medical staff to motivate them to report suspected incidents to the local public health authorities. To meet the increased number of reported incidents, probably mostly involving persons not in need of medical care, and to optimize use of available resources, the municipal public health authorities should develop criteria for when and how to investigate incidents. The minimum requirement should be conducting an interview of the affected person according to a standardized questionnaire. Detection and investigation of incidents can be improved by arranging opportunities for medical and public health staff to meet on a regular basis and by establishing channels for rapid communication. This will facilitate both cooperation and coordination. These groups also need more training in food safety and in modern techniques to investigate incidents. Reporting of investigated incidents to the National Food Administration should be facilitated through a web-based system, and the feedback to those contributing should be improved. Foodhandlers in restaurants and similar establishments should be educated regarding food hygiene. Finally, to improve our knowledge, the existing passive surveillance should be supplemented with additional studies and approaches.