We used a recently developed diagnostic adverse drug reaction (ADR) algorithm and an intensive telephone surveillance program to monitor all courses of prescription and nonprescription drug therapy in a general pediatric group practice for 1 year. A total of 3181 different children visited the practice during the year and received 4244 separate courses of drug therapy. Adverse symptoms were noted in 473 (11.1%) of these courses of therapy. Of 534 total adverse symptoms, however, only 24 scored as definite and 176 as probable ADRs. The main ADRs noted were antibiotic-associated gastrointestinal complaints and rashes, and various manifestations of CNS stimulation with bronchodilators. Sociodemographic variables significantly associated with the risk of a definite or probable ADR were socioeconomic status (P less than 0.0001), ethnic origin (P = 0.0015), and age (P less than 0.05). Treatment-related risk factors included treatment by a practitioner outside the study practice (usually during nonoffice hours) (P less than 0.001) and administration of a dosage above the range recommended by the manufacturer (P less than 0.001). Half the ADRs were judged as inconsequential by the children's parents, and most of the remainder resulted in only minor morbidity. Half were judged to be highly or probably preventable. Our results suggest that ADRs do not occur commonly in general pediatric outpatients and that most are mild and self-limited.