Correlates of using mental health services: Implications of using alternative definitions

Background. Studies using varying definitions of mental health visits result in widely divergent estimates. This analysis examines the stability of a predictive model using three varying definitions based on provider type, diagnosis, psychotherapy, and psychotropic medication use. Methods. Interview and questionnaire data and claims records from the RAND Health Insurance Experiment are used to test these models among 3138 enrollees. Results. Estimates of visits, and factors associated with them, are highly sensitive to definitions. Depression was the only symptom/life situation variable, and education the only sociodemographic measure, predictive across all three models. Risk indicators such as suicide thoughts and drinking problems were only significant for the traditional (mental health specialty) model. While patients within the traditional model definition were significantly younger than other enrollees, those within the model using the most expansive definition were significantly older. Varying the definition also led to different results in respect to experimental manipulations, geographic sites, and some specific types of comorbidity. Conclusions. A reasonable definition, consistent with medical standards, requires, at least, a mental health diagnostic judgment and some form of psychotherapeutic or drug treatment. Studies of the content of mental health care are needed.