BackgroundHigh quality of surveillance systems for surgical site infections (SSIs) is the key to their usefulness. The Norwegian Surveillance System for Antibiotic Consumption and Healthcare-Associated Infections (NOIS) was introduced by regulation in 2005, and is based largely on automated extraction of data from underlying systems in the hospitals.MethodsThis study investigates the quality of NOIS-SSI’s denominator data by evaluating completeness, representativeness and accuracy compared with de-identified administrative data for 2005–2010. Comparisons were made by region, hospital type and size, age and sex for 4 surgical procedures.ResultsThe completeness of NOIS improved from 29.2 % in 2005 to 79.8 % in 2010. NOIS-SSI became representative over time for most procedures by hospital size and type, but not by region. It was representative by age and sex for all years and procedures. Accuracy was good for all years and procedures by all explanatory variables.ConclusionsA flexible and incremental implementation strategy has encouraged the development of computer-based surveillance systems in the hospitals which gives good accuracy, but the same strategy has adversely affected the completeness and representativeness of the denominator data. For the purpose of evaluating risk factors and implementing prevention and precautionary measures in the individual hospitals, representativeness seems sufficient, but for benchmarking and/or public reporting it is not good enough.