Valid measures of severity are crucial in asthma pharmacoepidemiological research. This study reports the development and validation of a severity measure in childhood asthma for application to health care administrative data. A drug treatment-based asthma severity measure was developed following the stepped care approach to treatment, and this was applied to a cohort of 16,862 children who met a case definition for asthma drug prescription use between January 1995 and March 1996. Assessments were made of the measure's reliability, validity, and responsiveness to change over time. The drug treatment-based asthma severity measure classified 42% of children as having mild asthma, 37% as having moderate asthma, 19% as having moderate-severe asthma, and 2% as having severe asthma. Agreement on severity classification between two successive time periods was excellent (kappa = 0.82). Children classified as having severe asthma were significantly more likely than children with mild-moderate asthma to have previous asthma hospitalizations, to visit asthma specialists, to have high physician utilization, and to require hospital critical care. They were more likely to be reclassified as having severe asthma 2 years later. These findings show that a drug treatment-based severity measure in childhood asthma, which can be applied to prescription data, has good reliability and validity, and is responsive to changes in asthma severity over time.