Statistical Analyses

Crude incidence was calculated from the reporting study, corrected for incomplete years. Corrections for incompleteness of participation of GPs and cases and for list inflation (persons no longer belonging to the practice population) were used to calculate an adjusted incidence. List inflation was calculated, on the basis of findings in an ongoing, population-based study, performed in the same general practices (22).

The response rate in cases and controls was estimated by using the registration forms. Reported cases and cases in the case-control study were compared by week, practice, age group, and gender to estimate the number of cases in both and each study part. The total cases reported, recruited for the case-control study, or both were used to estimate an incidence corrected for nonresponse. A logistic regression model was used to identify factors independently associated with total response. A variable was excluded if exclusion did not significantly decrease the log likelihood of the model. Because information on the response of recruited cases was too limited to study effects of patients and practice characteristics, only the total of known cases to participating cases was studied in this model.

The effects on incidence of age and gender of the patient, year of study, degree of urbanization, region of the Netherlands, and participation of the sentinel practice in the case-control study were estimated univariately; independent effects were estimated by using Poisson regression. A variable was excluded from the model if exclusion did not increase the deviance significantly. Cases that did not meet the case definition and controls that did, according to self-reported symptoms in the questionnaire, were excluded from analyses.

Case-Control Study

Cases and Controls

From May 1996 until April 1999, 985 cases and 717 controls returned a questionnaire to the National Institute of Public Health and the Environment. Of these, 878 cases and 581 controls could be included in the analyses. Stool samples were examined from 857 (98%) patients and 574 (99%) controls. The median age of patients (29 years) was significantly lower than that of controls (37 years). Gastrointestinal complaints of longer than 1 month were more frequent in patients (31.5%) than in controls (7.3%). The percentage of patients (18%) under treatment of a specialist was similar to that of controls, as was the number of GP consultations in the last 3 months (1 consultation). Patients were slightly more often born outside the Netherlands than controls, and the educational level of patients was higher than that of controls (both not statistically significantly).

Clinical Symptoms of Patients

Almost all patients reported loose stools (Table 3). Other commonly reported symptoms were frequent stools, abdominal pain, abdominal cramps, and nausea. Only 12 (1.4%) patients reported vomiting and no diarrhea. In cases with a higher frequency of stools than normal, the median of the maximum frequency was 6 times in 24 hours (first quartile-third quartile: 4-8 times in 24 hours). The median duration of symptoms before a GP was consulted was 6 days (first quartile to third quartile: 3-20 days); 20% of patients had been symptomatic for more than 4 weeks before consulting a GP.

Bedrest was required for 47% of patients for a median of 2 days (mean 3.1 days); 41% of the children who regularly visited a day-care center had to stay home for a median of 3 days (mean 3.2 days), 58% of schoolchildren were absent from school for a median duration of 3 days (mean 3.8 days), and 60% of working patients were absent from work for a median duration of 2 days (mean 3.1 days). In 8% of cases, someone had to miss school or work for a median of 1.5 days (average 2.0 days) to care for the patient; 56% of patients used medication for gastroenteritis: 4% used antibiotics, 18% analgesics, 27% antidiarrheic medication, 10% oral rehydration solution, and 21% additional medications.

Symptoms and Diagnoses of Controls

A small proportion of controls (6.7%) reported consulting a GP for gastrointestinal symptoms when they were recruited for the study but did not meet the case definition; 14% consulted for other reasons (e.g., to pick up a prescription, routine physical examination, accompanying a relative); most controls gave no information (43%) or had consultations for other symptoms (36%).

Pathogens

The most frequently detected pathogens in cases were Campylobacter, Giardia lamblia, rotavirus, Norwalk-like viruses, and Salmonella (Table 4). Bacterial pathogens were found almost solely in cases, except for Yersinia and verocytotoxin-producing E. coli. All isolated Yersinia spp. were nonpathogenic serotypes. The verocytotoxin-producing E. coli serotypes found in controls were different from those found in patients but included pathogenic types, such as O26. E. coli O157 K-H- was isolated from one case. The percentages of Campylobacter, Salmonella, and Salmonella Enteritidis did not differ significantly over the study years. Viral pathogens were found in 1% to 5% of patients and in a small percentage of controls. The possibly nonpathogenic parasite Dientamoeba fragilis, and the nonpathogenic parasite Blastocystis hominis were common and were found more frequently in controls than in patients.

In 37.5% of patients and 9.8% of controls, a pathogen could be detected (excluding possibly nonpathogenic microorganisms, Table 4). From May 1998 until May 1999, these percentages were increased by Sapporo-like viruses with 1.7% in patients and 0.6% in controls. When D. fragilis was considered a pathogen, these percentages were 44.1% and 22.9%, respectively. The percentage in which a pathogen was found was similar for the 50 controls who reported gastrointestinal symptoms (but did not meet the case definition) and controls without gastrointestinal symptoms.

Most stool samples of patients were received at the National Institute within 1 day of collection (57%), 21% after 2 days, 9% after 3 days, 7% after 4 days, 3% after 5 days, and 2% after >6 days. No significant differences were noted in the percentage of samples in which a pathogen was detected for different postal delays. For Campylobacter, however, a decreasing trend in the proportion of positive cases was observed with an increasing postal delay: 1 day: 12% positive, 2 days: 10%, 3 days: 9%, 4 days: 6%, 5 days 5%, 6 days or more: no Campylobacter.

Incidence and Participation Rate

This is the first GP-based national study in the Netherlands covering the role of a wide range of microorganisms in a representative population of gastroenteritis patients and controls. Only England has conducted a similar study.

The incidence of gastroenteritis in general practices was estimated at 79.7 per 10,000 person years, suggesting that in the Netherlands, 1 in every 125 persons, or 128,000 persons, will seek physician care for gastroenteritis every year. This estimate was adjusted for list inflation and partially for nonparticipation (15%) of GPs but not for the number of underascertained cases absent from both study components. In a similar study in England, underascertainment was estimated to be 36% (23). The correction for list inflation (2.5%) in our study should be considered a minimum estimate because only patients who were actively reported as no longer belonging to the general practice population were counted; by contrast, the more active approach in England (searching medical records of nonrespondents) yielded an estimate of 10% (23). Consequently, the incidence from our study must be considered a conservative, but the best available, estimate.

A previous estimate of the incidence of gastroenteritis in a GP-based study in 1992 to 1993 in the Netherlands was somewhat higher at 90 cases per 10,000 person years (not corrected for list inflation) (8). Since the percentages of Salmonella and Campylobacter have also decreased since 1992-1993 (from 5% to 4% and from 14% to 10%, respectively), the decrease in incidence could partially be due to preventive measures in the poultry industry that focused on Salmonella Enteritidis, but as expected, also caused a slight decrease in Campylobacter infections in poultry. Nevertheless, these measures were not fully implemented until April 1997, and therefore a decreasing trend in the incidence within the study period would be expected, which was not observed for gastroenteritis as a whole, nor for the percentages positive for Salmonella and Campylobacter. Several other factors might have contributed to this decrease, such as a more widespread use of Hazard Analysis and Critical Control Points in the food production industry, greater awareness and knowledge in the population about foodborne infections, and a change in consultation behavior because of GPs' increasing deferral policy for gastroenteritis.

The low incidence in the first study year was presumably due to the absence of a winter peak in 1996 to 1997. Such peaks coincide with an increase in infections with enteric viruses, such as astrovirus, rotavirus, and calicivirus (24–26). Therefore, the variation in incidence in different years likely is due to the annual fluctuation of the peaks in viral pathogens.

Our data indicate an incidence almost fourfold lower than the incidence in England. The GP-based incidence of gastroenteritis estimated for England in a study from 1993 to 1996 was 330 per 10,000 person years after correction for list inflation and underreporting (27). In Wales in 1992, an incidence of 244 per 10,000 person years was found (28). Most likely, gastroenteritis patients' higher consultation rate in England accounts for the difference: approximately 1 in 6 gastroenteritis case-patients in England consult a GP, whereas an estimated 1 patient in 10 to 1 patient in 50 does in the Netherlands (5,6,8). A lower consultation rate could be explained by Dutch GPs' policy of deferring gastroenteritis cases. In the Dutch Guidelines for Acute Diarrhea, a consultation by telephone is considered adequate for uncomplicated acute diarrhea and can be dealt with by the GP's assistant (29). Of the 19 participating GPs who completed a questionnaire at the end of our study, 12 (63%) reported fully or partially discouraging patients with gastroenteritis from visiting their practice. In England, a deferral policy for consultations for gastroenteritis is not common.

The incidence of gastroenteritis was independently associated with degree of urbanization and geographic region. The incidence was the lowest in the North, as was found in the study in 1992 to 1993 (8). This cannot be explained by a lower response rate (Table 2). A higher incidence was found in areas with a high or low degree of urbanization than in areas with an intermediate degree of urbanization. Other studies also report a higher incidence in urban areas (30). Possibly, a high population density causes an increased risk for person-to-person transmission. In rural areas, contact with animals, manure, and raw products could be more frequent than in more urbanized regions, possibly leading to higher levels of exposure to zoonotic pathogens. An analysis of risk factors for each pathogen will be published after an ongoing population-based study is completed. The differences in incidence could also reflect differences in consultation behavior.

Patient-related factors that were independently associated with incidence were patient's gender and age. The incidence was higher for women than for men and was clearly higher in the youngest age groups, consistent with other studies (5,8,28,31–33). Although several studies have reported a higher disease burden for the elderly, no increase in the incidence of gastroenteritis was observed in our study (1). Possibly, the higher risk for gastroenteritis for the elderly is limited to the relatively weaker persons living in nursing homes (25). Because these homes usually have their own GP, the proportion of persons living in nursing homes could be underrepresented in this study.

Pathogens

Bacteria were detected in 16.3% of patients, viruses in 15.4%, and pathogenic parasites in 8.3%. The higher participation of the northern region might have increased the percentage of Salmonella Typhimurium because S. Typhimurium is predominant over S. Enteritidis in this region, in contrast to the rest of the Netherlands (34). The order of relative importance of pathogens in general practices was similar for the Netherlands and England, although the percentages positive for viruses and bacteria in this study were slightly lower than in England and the percentages positive for parasites were slightly higher. Because bacteria and viruses are more often detected in patients with acute diarrhea, whereas parasites tend to cause less fulminant but intermittent and long-lasting symptoms that might lead to delayed consultation, these differences could be due to the exclusion of persons with symptoms lasting longer than 2 weeks in the English study (35,36). In our study, 32% of patients had symptoms >2 weeks' duration (36). In addition, we used formalin-fixed material to detect parasites, and four different preparates were examined, which increased the sensitivity of microscopy examination, compared to the use of nonfixed material and the examination of three different preparates in the English study (37). The lower participation of patients in the last 2 years of the study might have caused an underestimate of viral pathogens because the seasonal viral peak was relatively low in the first year (24,38). In spite of a less sensitive method of testing for Norwalk-like viruses (electron microscopy in England versus reverse transcriptase-polymerase chain reaction in the Netherlands), the percentage of stool samples positive for Norwalk-like viruses is higher in England (36). The low response for the younger age groups might also have reduced the percentage of rotavirus, Sapporo-like viruses, and to a lesser degree Norwalk-like viruses, which are most common in young children (24). Differences in the proportion of specific pathogens in the English study and in our study may also be explained by differences in consultation rates. A lower threshold for consulting a GP might increase the proportion of pathogens that cause relatively mild gastroenteritis, such as Norwalk-like viruses.

Parasites that were (possibly) nonpathogenic were more frequently found in controls than in patients. A more detailed study of differences between patients and controls with these parasites might identify factors related to the development of disease after infection with these parasites.

Diagnostic Deficit

In spite of a diagnostic panel that included most of the known pathogens that can cause gastroenteritis, the percentage of patients in which no pathogen could be detected was 61% (including Sapporo-like viruses in 1998 and excluding D. fragilis and nonpathogenic parasites). Some cases could be noninfectious because an exclusive distinction between infectious and noninfectious intestinal disease cannot be made clinically. The high percentage of patients with chronic gastrointestinal symptoms suggests that a substantial proportion might not be infectious but an expression of other illnesses, such as inflammatory bowel disease. The symptoms might also be caused by intestinal microorganisms not included in this study or not yet known. Several pathogens, such as Campylobacter, verocytotoxin-producing E. coli, and torovirus have only recently been recognized as a cause of gastroenteritis, and it is likely that new pathogens will be added to this list [39]. In addition, we did not screen for some pathogens, such as some pathogenic E. coli (e.g., enterotoxigenic E. coli, enteroinvasive E. coli) and bacterial toxins (Bacillus spp, Clostridium difficile cytotoxin, C. perfringens enterotoxin), which were associated with 15% and 6% of the cases, respectively, in the English study (36). However, all stool samples from our study were stored and are currently being tested for Sapporo-like viruses and can be tested for other pathogens. Also, intestinal symptoms can be caused by nonintestinal infections, such as influenza, not included in our study. Approximately one quarter of gastroenteritis cases in an American population-based study coincided with respiratory disease (31). Detection of pathogens in the study is also influenced by logistics factors and the sensitivity of the testing method. The timing of sampling was not ideal in many cases; because of the relatively long patient delay, some pathogens might no longer have been being excreted in the stool at the time of sampling. Some pathogens, such as Campylobacter, do not grow in culture after a long delay in shipping. At present, we are conducting a study on gastroenteritis in the community to elucidate the role of sampling time and referral behavior on pathogen-specific incidence.