Two models estimating the proportion of Escherichia coli O157:H7 cases not reported in the Ontario notifiable diseases surveillance system are described. The first model is a linear series of adjustments in which the total number of reported cases is corrected by successive underreporting coefficients. The structure of the second model is based on a relative difference in the proportion of E. coli O157:H7 cases which are hospitalized between the surveillance database and the underlying population. Based on this analysis, the rate of under-reporting of symptomatic cases of E. coli O157:H7 infection in Ontario ranges from 78 to 88% corresponding to a ratio of 1 reported case for approximately 4-8 symptomatic cases missed by the surveillance system. This study highlights the need to increase awareness among public health workers of the potential biases that may exist in the interpretation of routine surveillance data.