The Impact of Emergency Mobile Psychiatric Services for Youth on Emergency Department Utilization: A Propensity Score Matching Analysis

Background:  Pediatric behavioral health visits to emergency departments (EDs) have skyrocketed in recent years.  EDs are expensive, often lack providers with sufficient expertise and typically do not provide the kind of follow-up care necessary to effectively address youth behavioral health problems. Connecticut’s Emergency Mobile Psychiatric Services (EMPS) sends trained mental health clinicians to homes, schools and community locations throughout the state to provide in-person crisis stabilization services and linkages. EMPS has a primary aim of preventing youth with behavioral health diagnoses and acute treatment needs from obtaining services through hospital EDs.  Accordingly, we designed a study using propensity score matching (PSM) to investigate whether an EMPS system implemented statewide reduced use of ED services for youth with behavioral health problems.  

Methods:  Using state Medicaid claims records, EMPS youth were matched with youth who received behavioral health treatment in hospital EDs in the same time-period (fiscal year 2014). The data recorded behavioral health treatment for both groups - EMPS youth and ED comparison youth (EDC) in the 18-month period leading up to the 2014 index episode and in the subsequent 18 months.  Using SPSS Version 23, propensity scores (with assignment to EMPS as the dependent variable) were generated for EMPS  and EDC youth using region, race/ethnicity (with missing values imputed), gender, any pre-index episode ED behavioral health visits, and diagnosis as covariates.  Propensity scores were then used to generate quintiles (Guo & Fraser, 2015), which were employed as separate strata in 1) a negative binomial regression to investigate the relative risk for the number of post-index episode ED behavioral visits and 2) a logistic regression to investigate the relative odds of any post-index episode ED behavioral health visit.  Risk and and odds ratios were aggregated (using Stata Metanalysis procedures) across strata to generate average treatment effects for the entire sample.

Results:  Post PSM covariate balance was adequate across groups with one exception.  Since prior behavioral health ED visits were more prevalent in the EMPS group, even after PSM rebalancing, regression models controlled for “any prior ED visits.”  Negative binomial regression on number of post-index episode behavioral health ED visits yielded an incident risk ratio of .78 (95% CI: .71, .87), suggesting that EMPS youth had a 22% reduction in new post-service ED episodes compared to EDC youth.  Logistic regression on any post-index episode behavioral health visits yielded an odds ratio of .75 (95%CI: .66, .84), suggesting that EMPS youth had a 25% reduction in odds of any post-service ED use compared to EDC youth.         

Conclusions/Implications:  Compared to other similar youth with behavioral health problems seen in ED settings, EMPS youth were at significantly reduced risk for subsequent behavioral health related visits in the ED.  When implemented statewide, EMPS services may be an effective model for addressing acute behavioral health needs for youth in the community.  Research examining 1) the cost-effectiveness of EMPS, 2) strategies for enhancing implementation of EMPS services, and 3) the generalizability of the effects found here to other states is warranted.