Disease burden studies commonly use data from electronic health records (EHRs) or community surveys. Quantitative bias assessments of these study designs are needed. We compared two studies on acute gastroenteritis (AGE) burden conducted in an integrated healthcare system in Oregon and Washington, USA. EHRs were used to identify AGE patients who sought care during July 2014 - June 2016 and determine the incidence of medically attended AGE (MAAGE). Members from the same health care system were surveyed during September 2016 - September 2017 to estimate community AGE incidence. MAAGE incidence was calculated using the rate of reported healthcare seeking among survey respondents and compared to the estimate derived from the EHR study. Survey respondents' EHR data were used to conduct a bias analysis. MAAGE incidence from survey respondents was 6.1 times higher than the EHR derived MAAGE estimate. Among survey respondents who self-reported contacting KPNW for an AGE episode, 36.3% had an AGE-coded encounter in the EHR during the same timeframe, and among those who reported no contact (either no AGE or AGE without medical attention), 2.6% did have an AGE-coded encounter. Potential noninfectious explanations for symptoms were reported by 35% of ill survey respondents. We quantify misclassification bias in both studies and discuss other potential sources of bias. Researchers should consider these biases when designing disease burden studies and consider including sensitivity analyses in published work.