Estimated numbers of community cases of illness due to Salmonella, Campylobacter and verotoxigenic Escherichia coli: Pathogen-specific community rates

all countries (1-8). Traditionally, reportable disease data have been the main source of information used to describe the burden of specific pathogenic causes of GI in the population (9-14). However, such data capture only a fraction of the true number of cases (3,15,16). In Canada, the National Notifiable Disease registry (NND) is a legislated mechanism that captures information on culture-confirmed cases of nationally reportable enteric pathogens. Case capture in this registry depends on many steps: patients must be sick enough to visit a physician; the physician must request a sample; the sample must be submitted to a laboratory; the sample must test positive; and the positive result must be reported to the local health authority, to the provincial health authority, and finally to the NND. Although this registry has the advantage of being highly specific, it is subject to significant under-reporting. However, if the extent of under-reporting was known, registry data could be adjusted to provide an estimate of the true number of cases in the population. A recent study by Voetsch et al (17) from the United States estimated the number of cases of salmonellosis in the population using nationally reported disease data, combined with

OBJECTIVE: To estimate the annual number of cases of illness due to verotoxigenic Escherichia coli (VTEC), Salmonella and Campylobacter in the Canadian population, using data from the National Notifiable Disease registry (NND), estimates of underreporting derived from several National Studies on Acute Gastrointestinal Illness, and the literature. METHODS: For each of the three pathogens (VTEC, Salmonella and Campylobacter), data were used to estimate the percentage of cases reported at each step in the surveillance system. The number of reported cases in the NND for each pathogen was then divided by these percentages. In cases where the pathogen-specific estimates were unavailable, data on acute gastrointestinal illness were used, accounting for differences between those with bloody and nonbloody diarrhea. RESULTS: For every case of VTEC, Salmonella and Campylobacter infection reported in the NND, there were an estimated 10 to 47, 13 to 37, and 23 to 49 cases annually in the Canadian population, respectively. CONCLUSIONS: The authors estimate that a significant number of infections due to VTEC, Salmonella and Campylobacter occur each year in Canada, highlighting the fact that these enteric pathogens still pose a significant health burden. Recognizing the significant amount of under-reporting is essential to designing appropriate interventions and assessing the impact of these pathogens in the population. estimates of under-reporting derived from population-based surveys, laboratory level surveys and the literature. In Canada, the Public Health Agency of Canada's National Studies on Acute Gastrointestinal Illness (NSAGI) initiative has performed similar population-and laboratory-level surveys. Thus, using methods similar to those outlined by Voetsch et al (17), the objective of the present study was to estimate the number of cases of illness due to verotoxigenic Escherichia coli (VTEC), Salmonella and Campylobacter in the Canadian population, using data from the NND, estimates of under-reporting derived from several NSAGI studies, and the literature.

Analytical approach
For each of the three pathogens (VTEC, Salmonella and Campylobacter), data from various sources were used to estimate the percentage of cases reported at each step in the surveillance system. In cases where pathogen-specific data were unavailable, data on GI were used, accounting for differences between those with bloody and nonbloody diarrhea. The number of reported cases in the NND for each pathogen was then divided by these percentages using the equation in the Appendix. Because there were two or more available estimates for the percentage of cases reported (eg, test sensitivity) at some steps in the reporting chain, the estimated number of community cases was calculated for both the most conservative and liberal of scenarios. By calculating a range of values, the authors attempted to reflect the variability and uncertainty of the estimates. An example calculation is provided in the Appendix to illustrate the methods used. Final estimates are reported as the number of community cases, ranges of annual rates per 1000 population, and ratios per reported case. Data were analyzed in SAS 9.1 (SAS Institute Inc, USA) and Microsoft Excel 2000 (Microsoft Corporation, USA).

Data
Percentage of cases reported at each step in the surveillance system: Table 1 outlines the data sources used to estimate the percentage of cases reported at each step in the surveillance system, by pathogen. These data sources are described here in detail, starting at the first step in the reporting chain.
The percentage of patients that visit the physician, the percentage of patients visiting a physician who are requested to submit a sample, and the percentage of patients who actually submit the sample (Table 1, steps a to c) were all estimated using data from the NSAGI Population Surveys, which were conducted in Hamilton, Ontario in 2001/2002 (2) and in British Columbia in 2002/2003 (18). Briefly, these two studies collected information over a 12-month period on the incidence and health care burden of GI using retrospective telephone surveys of randomly selected community members. The combined sample size was 8108. For this analysis, a case was defined as someone experiencing diarrheal illness before being interviewed, the same definition used in the study by Voetsch et al (17). Diarrheal illness was defined as diarrhea that lasted longer than 24 h, and diarrhea was defined as three or more loose stools or stools with abnormal liquidity in 24 h. Individuals were excluded from the case group if they reported that their symptoms resulted from a chronic or other medical condition previously diagnosed by a physician (eg, pregnancy, Crohn's disease, irritable bowel syndrome).
Because the percentage of cases who (a) visited a physician, (b) were asked to submit a sample and (c) submitted a sample was different for those with bloody diarrhea versus nonbloody diarrhea, and because the proportion of cases expected to have bloody diarrhea was thought to vary for the three pathogens, separate percentages for those with bloody and nonbloody diarrhea were used for these three steps accordingly. The results of recent Canadian and international outbreak investigations (19-37) were used to obtain An extensive review of the international microbiology literature, circa 2000, on pathogen-specific test sensitivity (39)(40)(41)(42)(43)(44) was conducted for the percentage of cases for which the test was found to be positive. For each pathogen, the highest and lowest test sensitivities reported in the literature were used (Table 1, step e).
The percentage of Canadian laboratories that reported confirmed cases of illness due to VTEC, Salmonella and Campylobacter to the local public health authority at least 80% of the time, as reported in the NSAGI Laboratory Survey (described above), was used as the estimate of the percentage of cases reported from the laboratories to the local health authorities (Table 1, step f).
The percentage of cases reported from the local health authorities to the provinces was estimated using data from the NSAGI Public Health Reporting Survey (38). Briefly, this study was a mail survey conducted with all local health authorities in Ontario and British Columbia between 2000 and 2001. For each of the three pathogens, the percentage of cases reported from the health units to the provincial health authority was used (Table 1, step g).
A reporting rate of 100% was assumed for the percentage of cases reported from the provinces to the NND because cases of infection of VTEC, Salmonella and Campylobacter are all required by legislation to be reported nationally (Table 1,

RESULTS
The estimated annual number of community cases and community rates per 1000 population due to infection by VTEC, Salmonella and Campylobacter are shown in Table 3. For every case of VTEC infection reported in the NND, there were an estimated 10 to 47 cases annually in the Canadian population. Similarly, for every case of Salmonella and Campylobacter infection reported in the NND, there were an estimated 13 to 37 cases and 23 to 49 cases annually in the Canadian population, respectively.

DISCUSSION
Due to factors such as underdiagnosing and under-reporting, the actual number of cases of illness due to VTEC, Salmonella and Campylobacter infection in the Canadian population is unknown. However, by combining data from the NND with estimates of the percentage of cases reported at each step in the reporting chain, the number of community cases can be estimated. Here, we estimated that approximately 2.5 to 6.9 cases of salmonellosis and 9.   study highlight the fact that these three pathogens represent a potential significant health burden in the Canadian population Our estimated range of the ratio of the number of salmonellosis cases in the community per reported case (13 to 37 cases) is very similar to previous estimates of 38.6 (17) and 39 (47) from studies conducted in the United States that used similar methodology. However, our estimate is higher than that reported in a previous study conducted in England (3), where an estimated 3.2 community cases occurred per reported case. Similarly, our estimated range of the ratio of the number of campylobacteriosis cases in the community per reported case (23 to 49 cases) is greater than that reported in the English study, where an estimated 7.6 community cases occurred per reported case (3). Such differences may be due to methodological differences (eg, calculations, surveillance systems, laboratory practices and case definitions used in population surveys), differences in the health care-seeking behaviour of patients, or possibly true differences in the epidemiology of these pathogens.
Interestingly, our estimated range of the ratio of the number of cases due to VTEC in the community per reported case (10 to 47 cases) is higher than the previously reported Ontario estimate of four to eight community cases per reported case (16). It is possible that this difference reflects a true difference in the reporting of VTEC infection in Ontario compared with the overall Canadian rates, but it is more likely that the difference in the estimates is due to methodological differences in the analyses, particularly in the data sources that have become available since the study, which was reported in 2000, was performed. Data used in steps b and c of Table 1 of our analysis were obtained from the NSAGI Population Surveys (2,18); in the Ontario study (16), however, these steps were combined and data from the literature were used to estimate the percentage of symptomatic patients from whom stool samples were obtained. Similarly, the percentage used in step d of Table 1 was obtained from the NSAGI Laboratory Survey, whereas the Ontario study relied on the assumption that 100% of submitted stool samples were tested. Here, the percentages used in steps f and g of Table 1 were obtained from the NSAGI Laboratory and Public Health Reporting Surveys, whereas the Ontario study combined these steps and used a value between 99% and 100% obtained through personal communication. It is likely that these differences in data used are the reasons for differences in the final estimates.
The main limitation of this type of analysis is that it requires several assumptions. First, we assumed that the behaviour of patients with respect to seeking care and submitting samples for testing was the same regardless of pathogen because, unfortunately, such pathogen-specific information is expensive to obtain and is currently unavailable for Canada. In an attempt to address this issue, however, we did account for differences in these behaviours by severity of illness (ie, bloody diarrhea), as well as for differences in severity by pathogen. Second, we used international literature, circa 2000, to obtain the sensitivities of the pathogen tests because information specifically on test sensitivity for Canadian laboratories was unavailable. However, our range of sensitivity estimates of 52.5% to 100% for VTEC were within the range of that used in the Ontario study (mean ± SD 51.8±10.4%) (16). Our range of sensitivity for Salmonella (85% to 100%) was slightly higher than that used by Voetsch et al (70%) (17). Third, the percentages obtained from the NSAGI Population Surveys for those who submitted stool samples (Table 1, step c), sorted by bloody and nonbloody diarrhea, were based on small numbers, as was the case in the study by Voetsch et al (17). Given this potential limitation, we reran the analysis with a pooled number of 93% for both those with bloody and nonbloody diarrhea. Using the pooled numbers made little difference to the results, producing estimates of eight to 43, 11 to 37, and 20 to 50 per reported case of VTEC, Salmonella and Campylobacter, respectively. Finally, the NSAGI Population Surveys were conducted in the city of Hamilton, Ontario, and the province of British Columbia; as such, the surveys may not be representative of the entire Canadian population.
In summary, we estimated that a significant number of infections due to VTEC, Salmonella and Campylobacter occur each year in the Canadian population, highlighting the fact that these enteric pathogens pose a significant health burden in Canada. These methods can be replicated in other countries, although it would be necessary to evaluate, on a countryby-country basis, how appropriately the multipliers used here can serve as proxy for each country-specific system. Estimates for other pathogens may be conducted in a similar manner using these methods as well, taking into account inherent differences between pathogens.
To further our understanding of the disease burden, an understanding of the pathogen-specific costs of illness is necessary. As well, recognizing the significant amount of under-reporting of cases that result from these three pathogens is essential to designing appropriate interventions, assessing the impact of pathogens in the population and identifying risk exposures.

ACKNOWLEDGEMENTS:
The authors thank all of the partners involved in the various NSAGI studies for supporting the collection of information crucial to this analysis. The authors also thank Carole Scott from the Public Health Agency of Canada for providing data from the National Notifiable Disease registry. This project was supported by the Public Health Agency of Canada.