This proposal seeks support for five lines of work focused on the NCS-A survey, a national survey of the mental health of adolescents. NCS-A is interviewing 10,000 12-17 year olds and administering questionnaires to their parents. DSM-IV disorders are being asessed with an adolescent version of the WHO CIDI. A small (n = 500 adolescents and their parents) K-SADS clinical reappraisal study is included as part of NCS-A. (1) We propose to integrate the 500 K-SADS clinical reappraisal interviews with the CIDI-A interviews by using prediction equations with item-level CIDI-A data in the calibration sample to generate multiple imputations (MI) of predicted K-SADS diagnoses for all 10,000 adolescents. MI methods will then be used to carry out parallel analyses of CIDI-A and predicted K- SADS diagnoses. The multiple imputations and the MI analysis programs used to analyze them will be made publically available as part of the NCS-A public use dataset. (2) We propose to refine the CIDI-A to improve its concordance with the K-SADS for use in future studies by revising pivotal low-concordance CIDI-A questions. These revisions will be based on targeted cognitive debriefing reinterviews with reappraisal sample respondents aimed at determining whether lack of clarity or imprecision in the CIDI-A questions help explain low concordance of particular CIDI- A/DSM-IV criterion-level asessments with parallel K-SADS assessments. (3) We propose to develop short self-report and parent-report screening scales of any disorder and SED based on analysis of an extensive battery of screening questions in the NCS-A. (4) We propose to write and distribute a computer program that can be used by school counselors, pediatrians, and others to assign individual-level predicted probabilities of SED from adolescent self-report and parent-report screening scores using on Bayesian methods that incorporate information about subpopulation prevalences based on screening scale distributions. (5) We propose to generate small area (State and County) estimates of the prevalence of SED for purposes of mental health treatment planning. Bayesian small-area estimation methods will be used here to borrow strength from screening questions that are included both in the NCS-A and in the US National Health Interview Survey as well as from geocoded data that can be mapped onto the NCS-A (e.g., Area Resources File, CPS, etc.).
Showing the most recent 10 out of 28 publications
