BACKGROUND: An estimated one in six Americans experience foodborne illness (FI) annually; most are neither diagnosed nor reported to health departments. New York City (NYC) has approximately 24,000 restaurants and >8.5 million residents, of which 78% report eating food prepared outside of the home at least once per week. Annually, the NYC Department of Health and Mental Hygiene (DOHMH) receives over 3,000 reports of possible restaurant-associated FI via the citywide complaint system (311), which identify approximately 30 outbreaks. Given the small number of outbreaks identified, it is probable that many incidents remain unreported. In 2012, DOHMH collaborated with Columbia University to routinely detect FI reports on Yelp. However, DOHMH is still likely not receiving all reports of restaurant-associated FI incidents. We sought to estimate the total number of NYC restaurant-related FI incidents and evaluate the DOHMH FI complaint system.
METHODS: FI reports submitted May 1, 2014–April 30, 2016 from 311 and Yelp were matched to identify distinct incidents reported for the same restaurant within 30 days of each other. We performed a capture-recapture analysis using the results of the match and an unbiased Lincoln-Petersen method to estimate the total number of restaurant-associated FI incidents and derive the number not reported.
RESULTS: A total of 8,152 restaurant-associated FI reports were submitted during the study period; 6,506 and 1,646 were reported via 311 and Yelp, respectively. The match identified 7,819 distinct incidents; 128 (2%) reported to both sources, 1,455 (19%) reported only on Yelp and 6,236 (80%) reported only to 311. The Lincoln-Petersen method estimated 78,155 (95% confidence interval: 65,410, 90,901) total incidents, suggesting that 71,791 (92%) and 76,572 (98%) were not reported to 311 and Yelp, respectively. DOHMH complaint systems were estimated to capture 10% of suspected restaurant-associated FI incidents, missing an estimated 70,336 incidents.
CONCLUSIONS: There was little overlap between incidents identified via Yelp and 311; estimates indicated many incidents remained unreported. An important limitation of capture-recapture is the method assumes each incident has the same chance of being reported; it is plausible that those who choose to report to 311 or Yelp have different characteristics compared with the overall affected population, which may affect the probability of an incident being reported to each system and may result in an overestimation of incidents. Nevertheless, the results are consistent with other FI research indicating gross underreporting. Efforts are needed to increase complaint system awareness and identify other reporting sources.