Risk-adjusted outcome models for public mental health outpatient programs

Abstract

Objective: To develop and test risk-adjustment outcome models in publicly funded mental health outpatient settings. We developed prospective risk models that used demographic and diagnostic variables; client-reported functioning, satisfaction, and quality of life; and case manager clinical ratings to predict subsequent client functional status, health-related quality of life, and satisfaction with services.

Data sources/study setting: Data collected from 289 adult clients at five- and ten-month intervals, from six community mental health agencies in Washington state located primarily in suburban and rural areas. Data sources included client self-report, case manager ratings, and management information system data.

Study design: Model specifications were tested using prospective linear regression analyses. Models were validated in a separate sample and comparative agency performance examined.

Principal findings: Presence of severe diagnoses, substance abuse, client age, and baseline functional status and quality of life were predictive of mental health outcomes. Unadjusted versus risk-adjusted scores resulted in differently ranked agency performance.

Conclusions: Risk-adjusted functional status and patient satisfaction outcome models can be developed for public mental health outpatient programs. Research is needed to improve the predictive accuracy of the outcome models developed in this study, and to develop techniques for use in applied settings. The finding that risk adjustment changes comparative agency performance has important consequences for quality monitoring and improvement. Issues in public mental health risk adjustment are discussed, including static versus dynamic risk models, utilization versus outcome models, choice and timing of measures, and access and quality improvement incentives.