The goal of this research application is to develop a new risk adjustment measure for indigent persons suffering from severe mental illness, schizophrenia. More specifically, the researchers intend to develop and validate a schizophrenia specific risk adjustment index based on combined information from prescription drugs, diagnoses codes, and demographic/eligibility information commonly found in administrative data bases. Potentially, a diseases specific risk adjustment index using diagnoses codes and drug markers found in administrative claims data will outperform claims derived indices based on demographics combined with diagnosis based information alone. The theoretical basis of the proposed model is derived from previous risk adjustment techniques based on comorbidities derived from ICD-9-CM codes developed by Elixhauser, Deyo, and Charlson and drug exposure data that can compliment comorbity information and delinieate disease and comorbidity severity initially organized by VonKorf. Other candidate variables specific to schizophrenia are also proposed. A retrospective longitudinal review of administrative Medicaid claims data linked with Georgia mental health institutional inpatient data for approximatley 15,000 persons suffering from schizophrenia between 1995 and 1998 will be used to build a prospective diagnosis based and a combined risk adjustment model including the diagnosis, demographic, and drug based information. The risk adjustment models will adjust for two outcomes of interest, total one year resource utilization and resource utilization specific to mental health and substance abuse. The risk adjustment models will be prospective in nature, where information obtained in one year will be used to predict expenditures the year after. The models will be developed and validated using a split halves validation procedure, where candidate variables will be screened and internally validated using bootstrap methods on a random half of the sample and measures of discrimination and validation will be calculated using the validation sample. Expert clinical judgement will used to supplement statistical candidate variable screening. The results of this research will provide payers of mental health a means to adjust capitated payment rates and enable researchers to stratify varying levels of risk in quasi-experimental retrospective comparitive evaluations. Ultimately this research could be extended to develop a risk adjustment model for all mental disorders. This application is responsive to Healthy People 2000's priority area for economic evaluation in Mental Disorders Prevention.
