![[ Visit Client Website ]](images/banner.jpg)
BACKGROUND: While the case-control study (CCS) has long been a fundamental tool of foodborne disease epidemiology, binomial probability calculations (i.e., Bernoulli trials) have emerged as a valuable companion (and sometimes substitute) for CCS during outbreak investigations. Binomial probability calculations (BP) have identified food vehicles in pulsed-field gel electrophoresis (PFGE) clusters of infections by Salmonella, Shiga toxin-producing E. coli (STEC), and Listeria. The FoodNet Population Survey (Pop Survey) has estimated U.S. background rates of food consumption, but its data are no longer current and capture fewer foods than does Oregon’s “Shotgun” hypothesis-generating questionnaire (HGQ) for PFGE cluster investigations. There is no plan to conduct a sixth Pop Survey, but a need remains for up-to-date background estimates of consumption of a wide variety of foods for use during PFGE cluster investigations. To meet this need, the Oregon Public Health Division (OPHD) built a data system to capture and aggregate food exposures ascertained routinely during interviews of reported cases of culture-confirmed Salmonella or STEC infection.
METHODS: Using FileMaker ProTM database software, Oregon designed a data system containing a library of standardized food exposures; potential exposures available for addition to an HGQ were assigned unique identifiers. Data from each unique exposure were aggregated from all previous interviews, yielding background estimates of food consumption. Additionally, background estimates from the Pop Survey were mapped to applicable exposures, enabling comparison with Oregon HGQ-generated estimates.
RESULTS: Oregon’s exposure library houses >1150 unique food exposure variables, all of which are available for use on HGQs and BP calculation through the analysis tool embedded in the system. Of these exposures, 288 correspond to Pop Survey variables, allowing for multiple BP calculations in one report. Epidemiologists can interpret BP to generate hypotheses that may be confirmed by tracing back common food exposures or further tested using case-control studies.
CONCLUSIONS: Oregon’s data management system and standardized HGQ supports rapid collection and aggregation of background food consumption data. Estimating BP for the observed exposure among cases is easily done and can facilitate rapid evaluation of numerous food exposures. This type of analysis can quickly focus investigation on likely food vehicles before controls could be recruited for a case-control study. Binomial probability calculations are an important and emerging tool for epidemiologists during PFGE cluster investigations, and generating reliable food consumption background estimates is an essential component of this process.