To assess the effect of the Court of Appeals opinion on the prevalence of juvenile court ordered psychiatric evaluations, we used an interrupted time-series design [7]. Interrupted time series designs are the standard for the evaluation of policy changes which are unfeasible to study using randomized designs [8].
We used de-identified data from the Tennessee Department of Mental Health and mid-year population estimates from the U.S. Census Bureau for this study. Administrative data from the Tennessee Department of Mental Health was used to estimate the number of juvenile court ordered inpatient and outpatient psychiatric evaluations in Tennessee. Publicly available data from the U.S. Census Bureau was used to estimate the number of 12 to 19 year olds in Tennessee.
Study month was the unit of analysis. The study time period included the 48 months from July 1, 2006, through June 30, 2010, divided into 3 time periods. The period before the Court of Appeals opinion included the 24 months from July 1, 2006, to June 30, 2008. The transition period was defined as the 2 months between when the Court of Appeals opinion was issued (e.g., July 1, 2008) and when it went into effect (e.g., August 31, 2008). The period after the Court of Appeals opinion included the 22 months from September 1, 2008, to June 30, 2010.
For each study month, we calculated the prevalence of juvenile court ordered inpatient and outpatient psychiatric evaluations per 100,000 youth (aged 12 to 19 years) in the Tennessee population. Prevalence rates were calculated using data from the Tennessee Department of Mental Health on the number of juvenile court ordered inpatient and outpatient psychiatric evaluations in Tennessee during the study time period, and U.S. Census data to estimate the number of youth who were 12 to 19 years old in Tennessee [9, 10]. Given the small number of events in each study month and the need to protect against the possibility of deductive identification of individual youth, demographic information was not available for inclusion in the current study.
Separate segmented linear regression models were fit for inpatient and outpatient evaluations. The regression models included a term for the effect of the implementation of the Court of Appeals opinion and adjusted for linear trends over time prior to the Court opinion and post-opinion, as well as the 2-month transition period. Correlograms were used to assess serial autocorrelation and, since there was no detectable autocorrelation, the data were modeled assuming independence. The study was reviewed and approved by the Tennessee Department of Mental Health and the Vanderbilt University Institutional Review Board.