Foodborne Cluster Detection Using Geospatial Analytical Techniques, Austin, Texas 2015

BACKGROUND: The objective of our study was to identify geographic clustering of foodborne illness complaints in our community.  Austin/Travis County has over 5000 food establishments serving a population of over one million.    The Austin/Travis County Health Department receives over 250 foodborne complaints each year, with such limited data it is important to analyze this data fully to identify outbreaks early.  In our analysis we utilized geospatial cluster analysis to identify whether there were areas in our community where foodborne illness clustered. 

METHODS: This is a cross sectional study utilizing 2015 Foodborne complaint and inspection data in Travis County. Foodborne complaint data is collected via interview utilizing a standard intake form which includes obtaining a three-day food history.  Complaint data and inspection results were data entered into Epi Info 7 and frequency analysis was performed.   All geospatial analysis was performed using ARCGIS 10.2.  Restaurant and complainant addresses were geocoded and cluster analysis was performed on complainant’s address, all restaurants listed in the three-day food history, restaurants directly implicated by the complainant, and restaurants that had violations noted on their subsequent inspection.

RESULTS:    Frequency analysis indicates that only 25% of our complainants sought medical attention with only 5% of our complaints having a confirmed pathogen.  Median age of complainants was 43 years.  On average complainants took 5.3 days to make a complaint after symptom onset reporting and incubation of 8.9 hours on average and with 42% reporting less than 4 hours.  The most common symptoms were diarrhea (81.6%) and vomiting (71.7%). Inspections indicated that the top violations were improper handwashing techniques observed (21.5%), evidence of insects (12.3%), risk of contamination (10%), and improper temperatures (18.5%).  Preliminary geographical analysis identified four geographical clusters in the analysis of all of the restaurants listed in the food history as compared with the three clusters from the suspect restaurant only analysis.  The violation analysis indicated only one cluster. 

CONCLUSIONS: This analysis demonstrates how complaint data may be used to identify clusters of illness.   Geographic clusters were noted for all restaurants listed in the food history (4 clusters), implicated restaurants (3 clusters), and restaurants with violations (1 cluster).  One of the weaknesses of this study is the small sample size and some selection bias inherent in using complaint data.  In the future this analysis may also be expanded to include restaurants listed in the three-day food histories from interviews conducted for persons with salmonellosis and campylobacteriosis.