Methods: Using data from EMPS program records merged with Connecticut state Medicaid data records, we examined information about behavioral health service use among 2,532 EMPS youth in the 18 months prior to and following the 2014 index service episode. We identified 78 potential variables as predictors of subsequent BHEDs from 4 domains – prior service use, demographic and background characteristics, child functioning, and EMPS episode characteristics. To identify the best predictors of BHEDs, we used “Classification and Regression Tree” (C&RT) analysis (Brieman, et al., 1984) incorporating all 78 predictors in two models – one predicting “any subsequent BHEDs” (ABH) and one predicting the “number of subsequent BHEDs” (NBH). Using the top predictors from the CR&T analysis, we performed regression models to evaluate the effect of these predictors on NBH (negative binomial regression) and ABH (logistic regression) simultaneously controlling for the other covariates.
Overall, 43.5% of EMPS clients had any BHED post-index; they averaged 1.19 (sd=2.09; median=0) BHED visits post-index.
Of the top NBH predictors from C&RT, being referred to inpatient services was most associated with higher NBH (IRR=1.57; 95% CI: 1.28-1.94). This was followed by being from the eastern region (vs. Hartford region; IRR=1.54 ; 95% CI=1.28-1.85), being a current EMPS client (IRR=1.47; 95% CI=1.29-1.87), the number of pre-index BHEDs (IRR=1.19; 95% CI: 1.16-1.22), and the number of types of trauma identified during the EMPS episode (IRR=1.14; CI 1.09-1.20). Being from the southwest region (vs. Hartford) was associated with having fewer NBH (IRR=0.68; 95%CI: 0.56-0.81).
The top ABH predictors from C&RT were the number of pre-index BHEDs (OR=1.55; 95% CI: 1.22-1.98), having prior ABH, followed by the number of pre-index BH services (ED or inpatient; (OR=1.23; 95% CI: 1.14-1.33), lifetime admissions to inpatient psychiatric services (OR=1.21; 95% CI: 1.08-1.37), and the Ohio Scale worker rating of family severity (OR=1.02; 95%CI: 1.02-1.03). Longer length of stay in EMPS was associated with a lower likelihood of having ABH (OR=0.99; 95%CI 0.985-0.996). We noted that many of the variables associated with BHED use are highly correlated with each other even though they retained significance in regression models.
Not surprisingly, youth with more extensive behavioral health service history are at higher risk for repeated involvement with EDs. Additionally, youth at risk for subsequent BHED visits present with more clinical and family risk than other EMPS youth. These findings, which will be shared with providers, can potentially be used to develop a risk-screening tool that could be used by providers statewide and nationally to inform treatment service delivery so that high-risk youth can more effectively be treated in the community.