No abstract

This study describes the rate that Medicaid encounter data on gender, race/ethnicity, and diagnosis matched information in the medical record, among a statewide sample of Medicaid children who received ongoing care for attention deficit hyperactivity disorder (ADHD), conduct disorder (CD), and major depression (MD) in outpatient specialty mental health clinics in 1998-1999. The match rate for gender was 99%; and for race/ethnicity it was 71.8%, 90.5%, and 89.7% for Caucasian, African American, and Hispanic children, respectively. Misidentified Caucasian children were more likely to be recorded as African American or Hispanic than misidentified minority children to be recorded as Caucasian. Diagnosis match rates were high (ADHD: 98%, CD: 89%, MD: 89%). If the California Department of Mental Health relied solely on Medicaid encounter data, misclassification of African American or Hispanic children as Caucasian could produce an underestimate of their service use.

This article explores the S-plus statistical computing environment in the context of an analysis of Duncan's occupational prestige data. The article highlights the modeling and graphics capabilities of S-plus and introduces basic programming concepts. In the course of the analysis, the authors develop functions for bootstrapping and kernel density estimation. In addition, they display some of the extensive interactive graphics capabilities of S-plus.