Asthma is an important chronic childhood illness. A population-based surveillance program could measure the burden of illness, but first, the validity of an administrative diagnosis of asthma must be confirmed. The objective was to evaluate the accuracy of population-based outpatient administrative data in identifying children with asthma for the purpose of on-going asthma surveillance and research. Twenty-one primary care physician (PCP) clinics in Ontario participated. Patients under 18 yr old were categorized into three diagnosis categories according to administrative data diagnosis codes: asthma, asthma-related, and non-asthma. In each PCP clinic, for each diagnosis category, 10 charts were randomly selected for abstraction. A panel of experts (blind to the code) reviewed the abstracted charts and identified them as asthma or non-asthma. The reviewers’ diagnosis was considered the gold standard. The accuracy of the administrative data diagnosis coding was analyzed using the concepts of diagnostic test evaluation. Six hundred and thirty patient charts were abstracted and reviewed. Overall agreement between the diagnosis provided by expert chart review and the administrative data diagnosis code was 84.8% (p < 0.001), and was 60.2%, 94.8% and 99.5% for the asthma, asthma-related, and non-asthma categories, respectively. Additionally, the sensitivity and specificity were 91.4% and 82.9%, respectively. Agreement between the administrative data diagnosis code and the PCP chart diagnosis was 99.4% (p < 0.001). An administrative data diagnosis code of asthma is sensitive and specific for identifying asthma. By using the results of this study as a starting point, future research will create a cohort of children with asthma to be used for population-based surveillance and research.