Assessing the predictability of Andersen's behavioral model to explain mental health outcomes in South Carolina using administrative data

Mental healthcare (or behavioral healthcare) in the United States historically has been delivered under the auspices of the states, rather than the private sector or as a single national policy. Fifty state policies for mental health care have resulted in a fragmented delivery system for mental health patients. This study measures the effect of converging changes in the location of care and changes in the financing mechanism for public mental health providers presents by using Andersen's Behavioral Health Model to evaluate mental healthcare services.;The objective of this study is to explore the utilization of the Behavioral Health Model to predict "bad outcomes" or policy failures for public mental health patients using only administrative data. The Traditional and Vulnerable domains are tested independent to determine the ability to predict bad outcomes and finally are combined to see if there is an improvement ability to predict bad outcomes.;From central South Carolina, a cohort of mental health patients were drawn in 2002 and followed for two years. Logistic regression models are used to evaluate the ability of Andersen's Behavioral Health Model to predict policy failures for patient served by the public mental health sector using only administrative data.;Results demonstrate that administrative data can be used in Andersen's Behavioral Health Model to predict bad outcomes for patients served by the public health sector.