RESUMO
We explored the feasibility, suitability, and reliability of using controls recruited among members of a non-probabilistic online panel ('panel controls') in a case-control study (CCS) to investigate a Salmonella Braenderup outbreak in Germany. For comparison, another control group was recruited via random digit dialling ('classical controls'). Panel members received questionnaires by email; classical controls were interviewed by phone. Both control groups were frequency-matched to cases by age and sex; the classical controls also by federal state. Cases and controls were queried mainly about fruit consumption since melons were the suspected infection vehicle. We calculated adjusted odds ratios (aOR) and 95% confidence intervals (CIs) using single-variable and multivariable logistic regression. The study included 32 cases, 81 panel controls and 110 classical controls. Analyses identified melons, particularly Galia melons, as the most likely infection vehicle using either control group (panel controls - aOR 12, CI 2.7-66; classical controls - aOR 55, CI 8-1100). Recruitment of panel versus classical controls required substantially less person-time (8 vs. 111 hours) and was about 10 times less expensive. We recommend this timely and reliable control recruitment method when investigating diffuse foodborne outbreaks with CCS.