Objectives: To estimate metabolic screening rates, predictors of screening, and incidence of metabolic disturbances in children initiating second-generation antipsychotic (SGA) drug treatment. Design: A retrospective, new-user cohort study (between July 1, 2004, and June 30, 2006) using Medicaid claims data. Settings: California, Missouri, and Oregon. Patients: A total of 5370 children (aged 6-17 years) without diabetes mellitus taking SGA drugs and 15 000 children without diabetes taking albuterol (control) but no SGA drugs. Intervention: Findings 1 year after recommendations from the American Diabetes Association and American Psychiatric Association called for metabolic screening of patients receiving SGA drugs. Outcome Measures: Serum glucose and lipid testing, 6-month incidence of diabetes, and dyslipidemia disturbances. Results: Glucose screening was performed in 1699 (31.6% [95% confidence interval (CI), 30.4%-32.9%]) SGA-treated children vs 1891 (12.6% [12.1%-13.2%]) control individuals. Lipid testing was performed in 720 (13.4% [95% CI, 12.5%-14.4%]) SGA-treated children vs 458 (3.1% [2.8%-3.3%]) controls. In multivariate logistic regression analysis, children with serious and/or multiple psychiatric diagnoses and those who used health care services more intensively were more likely to receive metabolic screening. The case incidence of glucose and lipid disorders was higher in SGA-treated vs albuterol-treated children (8.9 per 1000 children [95% CI, 6.6%-11.8%] vs 4.9 per 1000 children [3.9%-6.2%]; and 9.7 per 1000 children [95% CI, 7.2%-12.7%] vs 4.6 per 1000 children [95% CI, 3.6%-5.8%], respectively). Conclusion: Most children starting treatment with SGA medications in this public sector sample did not receive recommended glucose and lipid screening.