Understanding Racial/Ethnic Disparities in Mental Health Care At Each Step in the Episode of Care

Cross-sectional analyses of mental health care disparities may miss important information (e.g., the probability of initiation of care and the duration of care) that is relevant to understanding disparities and identifying policies to decrease disparities. In this paper, we compare disparities measured using longitudinal data with typically used cross-sectional analyses and comment upon the policy implications of differences in measurement and disparities results.

We combine four two-year longitudinal datasets from Panels 9-12 (2004-2008) of the Medical Expenditure Panel Surveys. We assess three different aspects of mental health utilization episodes among non-Latino whites, African-Americans and Latinos: 1) the probability of initiation of care; 2) the intensity (visits, Rx fills, and expenditures) and duration of care (days) during an episode.

We use survival analysis to estimate mental health care initiation and duration of episode accounting for right censoring. Intensity of care is measured using negative binomial regression models to estimate number of visits and number of prescription drug fills and refills. For comparing mental health care expenditures per episode across groups, we use 2-part generalized linear models (part 1 being initiation and part 2 being expenditures per episode given initiation).

Using longitudinal data, we identified racial/ethnic disparities in initiation of care and duration/intensity of care that were greater in magnitude than disparities identified using cross-sectional data. We recommend the use of episodic analyses when longitudinal data is available given that information related to initiation and dropping of care is potentially more useful to the development of policy solutions.