Behavioural Health Screening and Service Use in a Statewide Sample of Medicaid-Eligible Pediatric Outpatients

Objective: This study explored site-level rates of behavioural health (BH) screening, positive screening, and BH service use in statewide data from the largest Medicaid Health Maintenance Organization in Massachusetts. Methods: Screening rates at annual well-child visits (WCVs) were assessed across 908 sites (practices) for patients ages 4-17 between 2014-2018. The primary analytic sample included WCVs in 2016 (N=76,752) and was restricted to sites with at least 30 WCVs (N=304 sites; N=72,842 patients). Use of BH services was assessed six months before and after the index WCV. Results: The mean WCV screening rate across the analytical sample was 71.2% (SD=31.3; range=0.0100.0%) and the mean positive screening rate was 7.2% (SD=12.7, range=0-100%). Using intra-class correlations, small, but meaningful differences, were found between sites in rates of overall (r=0.38; 95% CI=0.25-0.50) and positive (r=0.10; 95% CI=0.00-0.29) screening. Although the relationship between a site’s rate of screening and the rate of BH treatment failed to reach statistical significance, there was a nonsignificant correlation (r=0.08, p=0.17) in the predicted direction and sites that screened at or above the mean screening rate (71.9% of their WCVs) were significantly more likely to have 6.0% or more of their patients receive subsequent BH services than were sites with screening rates below 71.9% (67.2% vs. 51.2%, p<.05). Conclusion: The current study documented a high level of continued compliance with the statewide mandate for routine psychosocial screening after more than a decade, although there were some relatively small decreases in compliance in recent years. The study also found that there were significant differences between sites in rates of BH screening and positive screening and a relationship between site-level rates of screening and service use. © 2020 J. Michael Murphy. Hosting by Science Repository. All rights reserved.


Introduction
Screening for psychosocial problems has long been recommended by national standard-setting organizations as a routine part of pediatric care.
The original guidelines for the U.S. Medicaid program specified that mental health screening was a required part of all well-child visits (WCVs) [1]. This guideline, which was a component of the Early and Periodic Screening, Diagnosis and Treatment (EPSDT) program, was reaffirmed in the Medicaid Omnibus Budget Reconciliation Act of 1989 [2][3][4]. In 1990, the American Academy of Pediatrics (AAP) created Bright Futures as a resource for evidence-based best practices in pediatric primary care which included psychosocial screening recommendations using validated screening measures [5]. More comprehensive screening recommendations have followed over the past twenty-five years [6][7][8].
Early investigations demonstrated that behavioural health (BH) screening rates were low among Medicaid-insured pediatric populations [9]. In 2006, a class action suit ruling in Massachusetts (Rosie D. vs. Patrick) found that the state was out of compliance with EPSDT's BH standards which led to a court-ordered remedy -the Children's Behavioural Health Initiative (CBHI) [10][11][12]. CBHI provides guidelines for practices, clinician training, and an expanded array of BH services [12]. As a part of CBHI, the state required that all Medicaid-insured youth receive a brief validated psychosocial screening questionnaire at their WCV. Each screen is reimbursed, and providers are required to bill for each screen and enter a modifier code to indicate whether the screen was positive or negative, so compliance can be monitored.
Prior to CBHI, only 4.0% of Medicaid-eligible pediatric patients had evidence of a standardized BH screen at the annual WCV [13]. However, after implementation of the CBHI, rates of screening rose to 53.6% in year two and 73.9% in year five of the program [13]. Studies based on statewide billing data from the early years of the program provided evidence of its positive impact on psychosocial outcomes. One 2014 study found that screening was associated with the identification of patients at risk for psychosocial problems who were not previously in BH treatment and another found that having a positive screen increased a patient's likelihood of receiving BH services six months after screening [14,15]. A third study which compared rates of screening and BH service use in Massachusetts and California found that rates of outpatient BH therapy increased in Massachusetts post-CBHI in comparison to California over the same period of time [16]. These findings, among others which have demonstrated the feasibility and impact of large scale psychosocial screening, were key factors in the National Quality Forum's endorsement of routine psychosocial screening in pediatrics as a national standard for quality care [17].
As of the current study, there have been no studies of statewide samples after the first few years of the program nor of statewide samples that analyse screening outcomes on a clinic-or practice-level. Since one important characteristic of adherence to quality standards is the demonstration of site-specific effects on patient outcomes, the current study explored these effects in a more recent sample, addressing the following research questions using both patient and site-level screening and BH service use data from the program's seventh to eleventh years (2014-2018): i.
Have rates of statewide BH screening and positive screening increased or decreased since the early years of CBHI? ii. Is there a relationship between a site's screening rate and the rate of BH service utilization? iii. What is the relationship between patient demographics, BH screening outcomes, and BH service use?

I Design
Administrative claims data were collected for pediatric patients who completed a routine WCV between 2014-2018 and were enrolled in Neighbourhood Health Plan (NHP; now AllWays Health Partners), the largest of the six Massachusetts Medicaid health maintenance organizations (HMOs). Data included billing claims for WCVs, BH screens, BH screen modifiers, BH service utilization records, and patient-level demographics (age, sex, WCV location).

II Study Sample
A total of 287,042 pediatric patients (4.00-17.99 years) presented for a WCV between 2014-2018 across 908 sites statewide. The sample was restricted to include the patients' first WCV of each year. We focused on a single year (2016) that had an index WCV screen and whose immediate prior and subsequent years had available BH service data. Therefore, the primary analytical sample included patients seen for a WCV in 2016 and used BH service data from as early as 2015 and as late as 2017 to assess BH service use six months before and after the 2016 WCV visit. Within the small number of patients with multiple WCVs in 2016, only the patient's first WCV was considered to assess BH service use six months before and after the patient's first 2016 WCV. In order to conduct meaningful site-level analyses, we further restricted the analytical sample to only sites with at least 30 WCVs to allow for appropriatelysized units of analysis, creating a sample of 304 clinics [18].

III Data
A WCV was identified for study inclusion via the following billing CPT codes: 99384, 99385, 99394, or 99395. BH screens were identified through 96110 or 96127 CPT codes which indicate that a screen had been administered and the result of the BH screen was identified through modifier codes. The U1, U3, or U5 modifier codes indicate that no BH need was identified (negative screen) and the U2, U4, and U6 codes indicate a BH need was identified (positive screen). A recent study suggests that BH claims data has a high concordance rate (94.3%) with the presence of a completed BH screen in the chart and that BH claim modifiers correspond to the screen's outcome according to scored screening data (72.9% agreement between BH risk modifier code and risk is shown on screening data) [19].
Although CPT billing codes make it possible to track the frequency and result of screens, they do not indicate which screen was completed. However, in the Massachusetts Medicaid program, only eight screens were approved for use in 2016. These included the Pediatric Symptom Checklist (PSC; parent-and youth-report measures in both the 35-and 17-item versions), the Parent's Evaluation of Developmental Status (PEDS), the Modified Checklist for Autism in Toddlers (M-CHAT), the Ages and Stages Questionnaire (ASQ), the Brief Infant Toddler Social Emotional Assessment (BITSEA), the CRAFFT Screening Test, the Patient Health Questionnaire-9 (PHQ-9), and the Strengths and Difficulties Questionnaire (SDQ) [13]. A 2016 study found that the PEDS, PSC, and MCHAT were the most frequently used pediatric screening tools in an evaluation of BH screening in Massachusetts in 2010 and 2012, with the PSC accounting for 80.0% of all screens in the 4-17-year-old age range [13].
BH service use six months before and after the patient's WCV was evidenced by 90791-90882 CPT codes (psychiatric diagnostic evaluation, individual therapy, family psychotherapy, group psychotherapy, pharmacologic management, transcranial magnetic stimulation, electroconvulsive therapy, individual psychotherapy with biofeedback, hypnotherapy, or environmental intervention for medical management purposes on a psychiatric patient's behalf) and 99201-99420 psychiatric evaluation and management codes.

IV Data Analysis
Analyses were performed using SPSS version 24 and R version 3.6.1. All hypothesis tests performed were two-sided, and significance was set at 0.05. Chi-square tests were used to compare rates of categorical BH screening, positive screening, and service use across sites. The Cochran-Armitage test was utilized to assess for directional trends in categorical BH screening and positive BH screening across five years (2014-2018). The relationship between site-level BH screening rates and rates of subsequent BH service use were assessed both continuously (Pearson bivariate correlation) and categorically (chi-square test). Rates of BH screening and subsequent BH service use were dichotomized according to the following criteria: BH screening was recoded as falling below or above the mean WCV screening rate of 71.99% in 2016 according to CBHI data reports, and BH service use was recoded as falling above or below 6.0%, which is the approximate mean percent of Medicaideligible pediatric patients who utilized CBHI services between 2015-2017 [20,21].
Binary logistic regression was used to explore which variables predicted an increased likelihood of BH service utilization six months post-WCV, including prior BH service use (yes/no), completion of a BH screen (yes/no), result of the screen (positive/negative), patient sex (male/female), and patient age (child: 4.00-12.99 years/adolescent: 13.00-17.99 years). Overall and positive screening rate reliability across sites was assessed using the intra-class correlation coefficient (ICC), which assesses the proportion of total variance in screening rate that is accounted for by site variation. A higher ICC implies that the betweensite variation (signal) is higher than the within-site variation (noise) [22,23]. The R version 3.6.1 package "ICCbin" was used to compute Fleiss-Cuzick estimates of the ICC and 95% Confidence Intervals [22,23].

I BH Screening: 2014-2018
Rates of BH screening were assessed from January Given that this proportion is ≥10.0%, the observed between-site screening rate differences can be considered meaningful [22,23]. The percent of cases that screened positive was 7.5% overall, with a range of 0-100%. The ICC for positive screening rates in the 245 sites that also had screening results available was 0.10 (95% CI=0.00-0.29). The proportion of total variance in screening rates accounted for by site variation was 10.0%, just within the lower bound of what is considered a meaningful between-site variation.

III Relationship Between Site-Level BH Screening and Service Use
To explore the relationship between site-level overall screening, positive screening, and BH service utilization rates in the six months following the BH screen, we first ran a correlation between site-level BH screening rates and rates of subsequent BH service use. A positive but statistically non-significant relationship between a site's BH screening rate and the rate of subsequent BH service use was found (Pearson's r=0.08, p=0.17). However, when overall site screening rates were dichotomized to contrast sites falling below (n=108 sites) with those falling above (n=196 sites) the mean screening rate of 71.9%, (as determined by CBHI) and when subsequent BH service use was dichotomized as falling below (n=45 sites) or above (n=259 sites) 6.0% (the mean rate of BH treatment in CBHI), sites that screened above 71.9% were significantly more likely to have BH treatment rates of 6.0% or above when compared to sites with less than 71.9% screening (88.3% vs. 79.4%;  2 =4.34; df=1, p<.05), suggesting a positive relationship between BH screening rates and rates of subsequent BH service use at the site-level (Table 2). Since there was a wide range in the volume of WCVs provided across the 304 sites (N=30-1,880 WCVs), we also assessed the relationship between the number of WCVs per site and BH screening and service use. For the following analyses, the number of WCVs per site was dichotomized using a median split: sites with higher volumes of WCVs (n=153 sites with 126 to 1,880 WCVs; n=10,087 patients) and sites with smaller volumes of WCVs (n=151 sites with 30-125 WCVs; n=62,755 patients). Larger volume sites billed for screens for a significantly larger proportion of patients (78.0%) compared to smaller volume sites (67.6%;  2 =518.50, df=1, p<.001). Prior BH service use was significantly higher (9.6%) in sites with a larger volume of WCVs compared to prior BH use in sites with a smaller volume of WCVs (8.9%;  2 =5.20, df=1, p<.05).

Discussion
The current study assessed compliance six to eleven years after the start of the CBHI screening mandate in the largest Medicaid HMO in Massachusetts by evaluating screening rates and BH service billing using site-and patient-level data. The study found a wide range of overall and positive screening rates across sites, and an ICC analysis found meaningful site by site differences in both rates. Categorical assessments of site-level screening rates and BH service use demonstrated that practices with WCV screening rates higher than the mean state-level screening rate had a higher proportion of subsequent BH service use. Additionally, in the sample of patients with no prior BH service use, evidence of a BH screen significantly predicted BH service use following the screen. Taken together, these findings suggest that sitelevel differences in adherence to BH screening mandates may be related to a small but significantly greater likelihood of BH service use for pediatric patients, specifically those who screen positively and who are, therefore, presumably in greater need of services. However, results also suggest a disparity in access to BH services since the rates of BH screening and of prior and subsequent BH services varied widely across sites.
The shared goal of the court and state Medicaid authority in establishing CBHI was to increase the identification and treatment of children at an early point in the development of BH problems. The current study provides evidence showing that even after more than a decade, in Massachusetts, that goal continues to be met, with hundreds of thousands of children screened each year and thousands receiving BH services. However, despite this decade-long success story, the current study also provided some cause for concern, with the screening rates reported in this study and in statewide data decreasing in the last few years while positive screening rates were increasing [21]. Therefore, there is a need for more research related to the following questions: Why are screening rates decreasing and why are rates of positive screens increasing? What should the clinical or administrative response be if no more than onequarter of positively screened children receive treatment? Should BH screening scores in individual patients be tracked to assess their progress (or lack thereof) over time? Are there specific barriers to receiving treatment in this population? What will be the impact of COVID-19 on screening and service use? Much of the information needed to answer these questions is available now and analysing it could serve to increase the efforts to continuously improve the quality of the CBHI screening program.

Highlights
i.
Although on a statewide basis more than 70.0% of Medicaidinsured pediatric patients had evidence of a completed BH screen more than a decade after the start of the Children's Behavioural Health Initiative (CBHI), rates of behavioural health (BH) screening and positive screening were found to significantly differ across sites. ii. Patients with a positive BH screen were three times more likely to receive BH services in the six months following their screen compared to those who did not score at-risk, further supporting previous evidence suggesting that BH screening informs BH service referrals for children in need. iii. Since small decreases in screening rates were noted across the study period, a quality assurance program might shed light on new challenges facing CBHI compliance.