Objective: To assess the impact of a computer based educational program on outpatient asthma care. Design: Quasi-experimental design with pre- and post-intervention chart audit. Setting: Primary care practice at an urban teaching hospital (treatment group) and a neighborhood health center (comparison group). Patients: Children aged 2-14 with at least one outpatient, emergency, or inpatient encounter for asthma during the periods April 1, 1993 to March 31, 1994 and April 1, 1995 to March 31, 1996 were eligible. Charts were selected at random for 145 and 73 patients at the two sites prior to the program, and for 128 and 37 patients following the program. All visits for asthma were reviewed. Intervention: Asthma Clinician Algorithm Training System (CAST) is a computer program developed to teach about the diagnosis and management of chronic asthma based on a local adaptation of NHLBI guidelines. Measures: Criteria assessed a range of practices for each visit recommended by the guidelines. Less stringent criteria assessed minimal adherence over the course of one year. Results: Adherence to guideline recommendations at the treatment site increased across several dimensions between the pre- and post-intervention period, especially when the less stringent per year criteria were used, e.g., assessing adherence (+29%), assessing triggers (+13%), patient education (+9%), and use of a peak flow meter (+9%). Similar changes also occurred during this time period at the non-intervention site. Changes in the treatment group relative to the comparison group were not significantly different. Conclusions: Although adherence to guidelines improved in several important dimensions during the study period, this change cannot be attributed to the computer educational system. Efforts at behavioral change should focus on multiple steps in the care process, and not rely solely on physician education.