Aggregating Interview Data to Estimate Background Frequencies for Food and Animal Exposures of Cases of Salmonella and Stec Infections in Oregon, 2007-2014

BACKGROUND: Culture-confirmed cases of Salmonella and Shiga toxin-producing Escherichia coli (STEC) infections are reported to local and state health departments and subsequently investigated to assess risk factors and epidemiologic links. In Oregon, these cases are all further subtyped using Pulsed-Field Gel Electrophoresis (PFGE); when PFGE subtypes are indistinguishable, clusters are identified and investigated as applicable. The Oregon Public Health Division has been using its homegrown “Shotgun” hypothesis-generating questionnaire to investigate PFGE clusters since 2003. Interview responses are tallied and compared with background exposure frequencies to calculate cumulative binomial probabilities in Bernoulli trials, generating hypotheses that can direct investigations. The questionnaire has gone through much iteration but has consistently asked about case exposures to over 400 food and animal sources.

METHODS: Between 2007-2014, 650 cases were interviewed using 12 successive versions of the “Shotgun”; exposures were compared across all versions. Because selection criteria were different, cases were segmented in to two methodological periods. Between 2007-2012, cases were interviewed only if identified as part of a PFGE cluster (period A); after December 2012, all cases of Salmonella and E. coli O157 infections were interviewed, regardless of association with a cluster (period B).   A master list of food and animals exposures was compiled in 2013 and exposure data from interviews conducted prior to then were recoded to correspond with this list. Recoded interviews were entered into Oregon’s “Shotgun” data system to be aggregated into background frequencies of food and animal exposures.

RESULTS:   The ensuing datasets contain food and animal exposure data from 650 case interviews, 183 in period A and 467 in period B. In period A, data from cluster-associated cases were aggregated from a total of 412 unique exposures; interviews were included from 75 cases of STEC infections and 108 cases of salmonellosis. In period B, data from sporadic and cluster-associated cases were aggregated from a range of 462 to 1026 exposures (depending on questionnaire version); interviews were included from 146 cases of STEC infections and 321 cases of salmonellosis.

CONCLUSIONS: Collecting these exposure data in a standardized way allows for ongoing data aggregation to be used for binomial probability calculations during cluster investigations. Estimating BP for the observed exposure among cases is easily done and can facilitate rapid evaluation of numerous food exposures; segmenting data sets by selection criteria will allow for future comparisons of between surveillance methodologies. These exposure data can be stratified demographically or by pathogen, illness onset period or outbreak association.