In this population-based retrospective cohort study, we examined the frequency, severity, and prediction of post-discharge surgical site infections (SSIs). We evaluated all patients admitted for their first elective surgical procedure in Ontario, Canada, between 1 April 2002 and 31 March 2008. Procedure and patient characteristics were derived from linked hospital, emergency room and physician claims databases within Canada's universal healthcare system. The 30 day risk of SSI was derived from the initial hospital admission, outpatient consultations, return emergency room visits and readmissions. The cohort included 622@?683 patients, of whom 84@?081 (13.5%) were diagnosed with SSI, and more than half (48@?725) were diagnosed post-discharge. Post-discharge infections were associated with an increased risk of reoperation (odds ratio: 2.28; 95% confidence interval: 2.11-2.48), return emergency room visit (9.08; 8.89-9.27), and readmission (6.16; 5.98-6.35). The most common risk index predicted incremental increases in the risk of in-hospital SSI, but did not predict increases in the risk of post-discharge infection. Patients with post-discharge infections had baseline characteristics more akin to uninfected patients than patients with in-hospital infections. Predictors of post-discharge infection included shorter procedure duration, shorter length of stay, rural residence, alcoholism, diabetes and obesity. Post-discharge SSIs are frequent, severe, scattered over time and location, and hard to predict using common risk indices. They represent an important hidden burden in our healthcare system.