Established methods of recruiting population controls for case–control studies in infectious disease outbreak investigations are resource- and time-intensive, and are often subject to bias. The online panel have recently gained interest as an easy and timely method to select controls. We examined the feasibility, suitability and reliability of using an online panel to select controls for case–control studies as part of investigations of diffuse food and waterborne outbreaks. In January 2019, we deployed a web survey by email to the 277 members of a non-probabilistic online panel in Lower Saxony, Germany. We questioned them on basic sociodemographic characteristics and eating habits. They were frequency matched to cases on sex and age. Their food exposures were compared to those of traditionally recruited controls of four historical case–controls studies, which successfully investigated food and waterborne outbreaks. We used logistic regressions to assess the association between the food exposures and the disease (odds ratios). The use of a control panel successfully led to the identification of the food items in three of the four historical outbreak investigations, and their recruitment benefitted from increased speed and limited costs. Timely outbreak investigations would enable rapidly implementing control measures. We recommend the further evaluation of using panellists as controls in parallel case–control studies and case–panel studies.

Current methods of control recruitment for case-control studies can be slow (a particular issue for outbreak investigations), resource-intensive and subject to a range of biases. Commercial market panels are a potential source of rapidly recruited controls. Our study evaluated food exposure data from these panel controls, compared with an established reference dataset. Market panel data were collected from two companies using retrospective internet-based surveys; these were compared with reference data from the National Diet and Nutrition Survey (NDNS). We used logistic regression to calculate adjusted odds ratios to compare exposure to each of the 71 food items between the market panel and NDNS participants. We compared 2103 panel controls with 2696 reference participants. Adjusted for socio-demographic factors, exposure to 90% of foods was statistically different between both panels and the reference data. However, these differences were likely to be of limited practical importance for 89% of Panel A foods and 79% of Panel B foods. Market panel food exposures were comparable with reference data for common food exposures but more likely to be different for uncommon exposures. This approach should be considered for outbreak investigation, in conjunction with other considerations such as population at risk, timeliness of response and study resources.