Efficacy of United States’ federally-funded interventions in increasing school capacities to improve student mental health and education outcomes in Tennessee

Abstract

About two in five children and adolescents will meet the criteria for a mental disorder by age 18, and more than half of youth who are accessing mental health services receive them in an educational setting. Yet there is limited evidence on the effectiveness of school-based interventions on children’s mental health and education outcomes. We examine the effectiveness of two key United States’ federally-funded interventions for expanding school-based capacities to improve children’s health and education outcomes—School-Based Health Centers (SBHCs) and Advancing Wellness and Resiliency in Education (AWARE) grants—in a mixed method, longitudinal study of low-income, Tennessee children.

We linked health insurance claims data for children enrolled in Tennessee’s Medicaid program with administrative education records for students attending Tennessee public schools between 2006 and 2019. We also implemented a census of Tennessee school districts to determine which had SBHCs and AWARE grants and their start years, and we conducted semi-structured interviews with each treated district to assess their infrastructure, programs, staffing, partnerships, health services offered, and more. We estimated effects of SBHCs and AWARE grants on school-level rates of mental health conditions, behavioral health conditions, preventive health care visits, absences, chronic absences, and disciplinary incidents using a staggered adoption, difference-indifferences (DiD) approach.

We found a statistically significant reduction in diagnosed mental health conditions among treated schools of 6 percent relative to their baseline prevalence, which our qualitative findings suggest might be related to increased health staffing in schools, earlier detection of mental health needs, and greater use of prevention strategies. We saw larger effects in some school districts with more extensive mental health infrastructure. We did not satisfy model assumptions for estimating causal effects on preventive health care visits, absences, chronic absences, and disciplinary incidents, although associations were in the expected direction.

1. Introduction

Approximately one in five children and adolescents are diagnosed with a mental or behavioral health disorder each year, and about two in five will meet the criteria for a mental disorder by age 18 (Bitsko et al., 2022). Since 2007, the number of children visiting emergency departments for mental health-related issues has risen considerably, with a 60 percent increase in visits for mental health conditions, a 159 percent surge in visits related to substance use, and a 329 percent escalation in cases of deliberate self-harm (Lo et al., 2020). In addition, between 2016 and 2020, rates of diagnosed anxiety and depression among children increased by 29 percent and 27 percent, respectively (Lebrun-Harris et al., 2022). The COVID-19 pandemic, in turn, exacerbated mental health struggles for children, with some estimates indicating that rates of depression and anxiety symptoms in children have doubled since early 2020 (Benton et al., 2022; Racine et al., 2021). At the same time, healthcare access and utilization deteriorated, including in school-based settings (Sullivan et al., 2021; Damian and Oo, 2022).

Given the extensive time that children and adolescents spend in schools and proactive interventions to support their mental health, schools serve as the most common institutional entry point to mental health services for youth (Keeton et al., 2012; Crosby, 2015; Love et al., 2019). A 2019 study using data from the 2012–2015 National Survey on Drug Use and Health found that among adolescents who received mental health services, 58 percent accessed them in an educational setting, and more than one-third—who were more likely to be low-income, publicly insured, or from minoritized racial and ethnic groups—did so in an educational setting only (Ali et al., 2019). In fact, the School-Based Health Centers (SBHCs) Reauthorization Act of 2020 described SBHCs as a “powerful tool for achieving health equity among children and adolescents who experience disparities in health outcomes because of ethnicity, race, and/or family income.”¹

Existing research generally distinguishes two primary types of SBHCs: on-site, fixed facility and off-site or school-linked centers, the latter of which include mobile health and telehealth (Doll et al., 2017). The definition used by the School-Based Health Alliance (SBHA) requires an SBHC to offer primary care services, although the percentage of SBHCs also offering behavioral health services has increased from about 33 percent in 2010 to over 80 percent (Soleimanpour et al., 2023). Federal funding for expanding school-based health interventions—from Medicaid expansion in the 1990’s, the Affordable Care Act (ACA) in 2010, and the School-Based Health Centers Reauthorization Act of 2020—has been critical to bridging gaps between healthcare needs and access for school-aged children. Approximately 70 percent of SBHCs also report receiving state funding for their operations (Arenson et al., 2019). Recent estimates suggest that there are more than 2500 SBHCs across the U.S. that provide access to care for 8–10 percent of public-school students (Soleimanpour, 2020; Haeder, 2021). SBHCs have typically opened in areas with higher proportions of low-income and underserved children; nearly half are in urban areas and about one-third are in rural settings (Arenson et al., 2019).

In addition, between 2014 and 2023, the Substance Abuse and Mental Health Services Administration (SAMHSA) awarded more than 80 Advancing Wellness and Resiliency in Education (AWARE) grants (up to $1,950,000 per year for five years) to state and local educational agencies to support school-based mental health programs and services. Similar in mission to SBHCs, these grants typically pass through state agencies to school districts to enhance awareness of mental health issues among youth, equip school staff with skills to recognize and address students’ mental health needs, and ensure that students with behavioral health needs are connected to appropriate services.

Despite the proliferation of SBHCs and AWARE grants, gaps in our understanding of their effectiveness in improving children’s mental health and education outcomes persist (Heinrich et al., 2023). The research base, including a systematic review of 46 studies of on-site SBHCs (Knopf et al., 2016), is generally consistent in showing that SBHCs increase children’s access to preventive health care services (e.g., immunizations, well child visits) and support the management of chronic health conditions (e.g., reducing symptoms and incidents related to asthma) (Arenson et al., 2019; Adams et al., 2020; Hussaini et al., 2021). The only mental health outcome examined among the studies was self-reported mental health, with mixed findings.

Two studies (Bains et al., 2017; Zhang et al., 2020) found that students who had access to mental health services through an SBHC were less likely to report depressive episodes and suicidal ideation or suicide attempts, compared to those in schools without SBHCs. More recently, Golberstein et al. (2023) found increased use of outpatient mental health services through SBHCs in a Minnesota county and reduced self-reported suicide attempts. This study also examined education outcomes and found no evidence of effects on average school attendance or standardized test scores. Other studies of SBHCs have found reductions in school suspensions and high school dropout rates and increases in student grade point averages, grade promotion, and educational attainment (Kerns et al., 2011; Knopf et al., 2016; Arenson et al., 2019; Paschall et al., 2019; Thomas et al., 2020; Westbrook et al., 2020).

There is a paucity of research on AWARE grants, with most of the evidence descriptive and from only a handful of school districts (Haggerty et al., 2019). For instance, one evaluation of an AWARE grant that funded mental health professionals, school social workers, and a Family Resource Center reported ensuing reductions in suspensions, substance use, and suicide ideation among students but lacked a comparison group.² The only study of AWARE grants that included a comparison group assessed their effects in a West Virginia county and concluded that they reduced absences and disciplinary referrals and increased grade point averages among students who received AWARE-funded services (Wilson, 2018).

In this research, we employ a mixed methods research design to investigate how the introduction of SBHCs³ and AWARE grants, both of which are used to increase school capacities for addressing children’s mental health needs, affect the mental health and educational outcomes of school-aged children. We use linked, longitudinal health and education data on children in Tennessee who were enrolled in the state’s Medicaid program between 2006 and 2019 to conduct a quasi-experimental, difference-in-differences (DiD) event study. This approach leverages temporal variation in the introduction of SBHCs and AWARE grants to estimate average treatment effects of these interventions on education, mental health, and health care utilization outcomes. We also draw on qualitative data collected in semi-structured interviews with school personnel and school district site visits to illustrate how schools are using SBHCs and AWARE grants to expand health infrastructure and services to better respond to students’ mental health needs.

2. Material and methods

2.1. Data sources

We use longitudinally linked health and education data on low-income, school-aged children in Tennessee. This dataset was assembled through a research-practice partnership between Vanderbilt University’s Peabody College, the Vanderbilt University Medical Center (VUMC), the Tennessee Education Research Alliance, the Tennessee Departments of Education and Health, and Tennessee’s Medicaid Agency (TennCare). The dataset includes all Tennessee public-school students who were enrolled in Medicaid between 2006 and 2019 and links demographic characteristics, academic outcomes, health conditions, and utilization measures at the individual level for each academic year. The dataset contains 1,575,411 unique children and 9,705,840 student-year observations, representing 67.5 percent of students in Tennessee enrolled in a public school between 2006 and 2019. Students in these data are enrolled in 2453 unique schools, totaling 24,248 school-academic year observations between 2006 and 2019.

2.2. Measures

2.2.1. Demographic and outcome measures

Measures of students’ demographic characteristics include race/ethnicity (White, Black, Hispanic and other), sex, economic disadvantage,⁴ special educational needs, housing status, and if English is not the child’s first language. Education outcomes of interest include school absences (based on daily attendance data), chronic absence rates (absent 10 percent or more of instructional days in one school year),⁵ and disciplinary incidents in the school primarily attended. Disciplinary incidents are disciplinary offenses (e.g., fighting, violation of school rules, and violence) and disciplinary actions (e.g., in- and out-of-school suspensions, placement in alternative facilities, and expulsion) recorded for students during the academic year.

Health outcomes assessed include the diagnosis of attention deficit and hyperactivity disorder (ADHD) and other conduct disorders; affective disorders (depression and anxiety); bipolar disorders; substance use disorders; and eating disorders. We also identified self-harm, suicidal ideation, and suicide attempts. These conditions were identified using the Centers for Medicare and Medicaid Services’ Chronic Conditions Warehouse (CCW) criteria for each condition using diagnosis and procedure codes from inpatient and outpatient claims and prescription medication use from pharmacy claims. Following CCW guidance, we required one inpatient or two outpatient claims with a diagnosis code or at least one pharmacy fill for a medication used to treat the condition. Additionally, we identified a measure of healthcare utilization – annual well-child screenings – in the claims data. Online Appendix 1 provides additional details on the coding and construction of the education and health outcomes.

2.2.2. Treatment (exposure) measures

To identify the presence of an SBHC (including school-based or school-linked health centers) in a school district, we developed a census of Tennessee school districts to determine SBHC status and their operational years. We started with a mapping tool from the School-Based Health Alliance (SBHA)⁶ and data from the Tennessee Department of Education (TDOE) survey of school districts. We then conducted outreach to each of Tennessee’s 147 school districts; we confirmed the SBHC status for 142 (97%) of these districts. To identify districts that received an AWARE grant, we used publicly available⁷ information to document the presence and years of AWARE-funded resources in counties during our study period.

The study period includes 42 treated school districts—35 with an SBHC, 7 with an AWARE grant, one of which has both an SBHC and AWARE grant—between 2006 and 2019. In our empirical analysis, three school districts with SBHCs that began before 2006 and one that started in 2006 are excluded because we do not have baseline data for them, yielding a total of 41 treated school districts and 1277 treated schools in the analysis.

Table 1 presents the number of unique schools potentially served by each SBHC and AWARE grant (by year introduced) and the number of school-by-year observations in the sample by year. Whereas the SBHC exposure measure is defined by the SBHC’s first year in operation, AWARE grants were allocated to help districts develop infrastructure like an SBHC over a 5-year period. As described by AWARE grant recipients in interviews, receipt of the grant typically provided the first opportunity to hire staff to identify students with mental health needs, which might initially contribute to higher rates of diagnosed conditions. In addition, the types of physical and mental health services made available to students through SBHCs and AWARE grants varies across SBHCs/AWARE recipients and over time (Larson et al., 2017), as well as between schools within districts. For example, in a site visit to one large school district, we observed two SBHCs, one at an elementary school and one at a middle school, the latter of which was on a campus that afforded high school students some access to the clinic. The district also offered behavioral health services to all students in the district through contracts with community-based providers. Student-level services utilization data are not routinely collected in school districts, and billing for services is also not a consistent practice; therefore, our data only capture services provided in schools that are billed to Tennessee’s Medicaid program, TennCare. For these reasons, the effect of the adoption of an SBHC or AWARE grant in Tennessee on student mental health outcomes at the school level is unclear a priori. We draw on qualitative data from interviews and site visits about services offered through SBHCs and AWARE grants to aid in interpreting our empirical findings.

Table 1.

Distribution of SBHCs and AWARE grants in schools over time.

	SBHC start year	AWARE grant start year	Number of unique schools treated	Number of treated schools across time	Percent	Pre-treatment periods	Post-treatment periods
Never treated	0	0	1054	11,852	47.83
Treated before 2006 a	3	0	111	1413	5.7
2006	1	0	11	154	0.62	0	11
2008	3	0	60	733	2.96	1	10
2009	4	0	654	3822	15.42	2	9
2010	2	0	12	162	0.65	3	8
2012	5	0	214	2360	9.52	4	7
2013	1	0	9	115	0.46	5	6
2014	0	3	40	507	2.05	6	5
2015	2	0	91	1140	4.6	7	4
2016	4	0	27	342	1.38	8	3
2017	3	0	27	372	1.5	9	2
2018	6	0	51	671	2.71	10	1
2019	5	4	92	1139	4.6	11	0
Total	39	7	2453	24,782	100

^aSchools treated before 2006 and 11 schools missing data in 2006 (shaded above) are not included in the estimation sample, yielding 1277 treated schools and 1054 never treated schools.

2.3. Analytic sample

Table 2 presents school-level descriptive statistics (demographics and health and education outcomes) for our analytic sample (overall), as well as separately for schools with access to an SBHC or AWARE grant in the year before the intervention was introduced and for schools not yet treated in that year. It shows that at baseline (before intervention), students in schools with access to an SBHC or AWARE grant during our study period were more likely to be students of color, economically disadvantaged, and to speak a language other than English at home than those without access to these interventions, although they were less likely to be identified for special educational services.⁸

Table 2.

Demographic characteristics and health and education outcomes by treatment status.

Measure (school level)	Overall	Treated (Baseline)	Not Yet Treated	t-test p-value
Demographic characteristics a
Female	48.92%	48.84%	48.93%	0.626
Black	18.21%	42.49%	16.90%	<0.001
Hispanic	5.26%	6.50%	5.20%	<0.001
White	74.95%	49.38%	76.33%	<0.001
Other race	1.58%	1.64%	1.58%	0.488
Economically disadvantaged	63.79%	74.73%	63.14%	<0.001
Inclusion in Special Education	18.51%	17.69%	18.56%	<0.001
English is not first language	5.77%	7.83%	5.66%	<0.001
Health outcomes
Any behavioral condition	0.426%	0.429%	0.426%	0.955
Drug use disorder	0.139%	0.124%	0.139%	0.552
Eating disorder	0.029%	0.028%	0.029%	0.885
Pregnancy	0.498%	0.559%	0.495%	0.236
STI	0.057%	0.091%	0.056%	<0.001
Tobacco misuse	0.236%	0.217%	0.237%	0.596
Any mental health condition	10.012%	9.719%	10.028%	0.107
ADHD	7.948%	7.963%	7.948%	0.923
Anxiety	2.102%	1.881%	2.114%	0.003
Bipolar disorder	1.063%	0.968%	1.067%	0.135
Depression	2.975%	2.685%	2.991%	0.005
Self-harm or suicidal ideation	0.143%	0.145%	0.143%	0.888
Well child visit	26.042%	30.547%	26.042%	<0.001
Education outcomes
Absence rate	6.299%	6.620%	6.282%	0.003
Chronically absent	17.643%	18.637%	17.589%	0.020
Any disciplinary outcome	10.378%	14.645%	10.147%	<0.001
Suspended	9.994%	14.444%	9.754%	<0.001
In-school suspension	6.205%	7.459%	6.138%	<0.001
Out-of-school suspension	5.573%	9.933%	5.338%	<0.001
Expelled	0.204%	0.231%	0.203%	0.578
Alternate program placement	0.930%	0.576%	0.950%	<0.001

^aCoding: = 1 if characteristic is present, = 0 otherwise.

Among the baseline mental health conditions (ADHD, anxiety, bipolar disorder, depression, and self-harm, suicidal ideation, or suicide attempt), rates of diagnosed anxiety and depression were lower in treated schools (vs. not yet treated schools) before intervention, which our interviews suggested might be related to a lack of access to mental healthcare professionals, particularly in rural areas. The baseline rates of diagnosed behavioral health conditions (substance use disorders, tobacco use, eating disorders, sexually transmitted infections, and pregnancy) indicate only one statistically significant difference between treated and not yet treated groups, the rate of sexually transmitted infections, which has a very low prevalence rate overall (less than 0.1%). Finally, absence rates, chronic absence rates, and rates of disciplinary incidents were statistically higher in treated schools, reflecting the intentional targeting of health resources to students in schools experiencing these educational difficulties.

We estimate the effects of SBHCs and AWARE grants on two health summary measures: any mental health condition and any behavioral health condition, and on the rate of preventive health care visits among children. For education outcomes, we estimated the effects of SHBCs and AWARE grants on absence and chronic absence rates for students at all school levels. We limit our analysis of disciplinary incidents to high school students—for which baseline prevalence rates are about twice those of middle school students and four times of those of elementary students—and estimate the effects of SBHCs and AWARE grants on rates of any disciplinary incidents and suspensions (the most frequent disciplinary action). In addition, in sensitivity analyses, we estimated the same quasi-experimental models for these outcomes separately by school level (high school, middle school and elementary school).⁹

2.4. Empirical methods

The introduction of SBHCs and AWARE grants occurred unevenly over a long period in our study, with some academic years having larger school districts or larger numbers of schools newly treated. This reflects, in part, the timing of disbursements of federal resources (i.e., ACA and SAMHSA funding) and also that funding for school-based health interventions has been allocated to bridge gaps between healthcare needs and access. Analysis of our interview data substantiates that funding for SBHCs and AWARE grants has typically been used to increase health staffing to address physical, mental, and behavioral health needs at schools, as well as to expand supports for students district-wide (including for families in some districts, such as through family resource centers). We accordingly estimate treatment effects at the school level, employing a quasi-experimental, DiD approach developed by Callaway and Sant’Anna (2021) specifically for circumstances with variation in treatment timing across multiple time periods.

This DiD method assumes parallel trends—in the absence of treatment, the average outcome for the treated group would have evolved in parallel to the comparison group (i.e., constant difference in outcomes over time)—after conditioning on observed covariates. This implies that causal effects may be identified even if differences in observed characteristics create non-parallel outcome dynamics between groups (those treated in different time periods). As shown in Table 1, schools in our sample differ in the number of pre-treatment and post-treatment periods available for examining pre-treatment and post-treatment trends in outcomes, depending on the timing of the introduction of SBHCs or AWARE grants. Using this DiD approach, we can estimate a single overall treatment effect parameter (like the average treatment on the treated effect estimated in a conventional DiD model), but also partial aggregations to estimate how average treatment effects vary by group and length of exposure to treatment (i.e., event-study-type estimation).

Following Callaway and Sant’Anna, we define $D_{idt}$ to be a binary variable equal to one if a school $i$ in district $d$ is treated in period $t$ (for $\mathrm{t}=1,\dots ,\mathrm{T}$), and equal to zero otherwise. Once a school becomes treated, we assume it remains treated in subsequent periods (the staggered adoption assumption). In turn, $G$ is defined as the period when a school first becomes treated, where $G_g$ is a binary variable that is equal to one if a school is first treated in period $g$, and $C$ is a binary variable that is equal to one for schools that are not treated in any period. The group-time average treatment effect—average treatment effect for schools that are members of a particular group g at a particular time period t—is then denoted by: ${\text{ATT}}_{(\mathrm{g},\mathrm{t})}=\mathrm{E}[{\mathrm{Y}}_{\mathrm{t}(\mathrm{g})}-{\mathrm{Y}}_{\mathrm{t}(0)}\mid {\mathrm{G}}_{\mathrm{g}}=1]$.

Estimating group-time average treatment effects allows us to investigate whether specific schools (or time periods of treatment) experience differential effects in health and education outcomes after the introduction of SBHCs and AWARE grants. As indicated above, we compare schools in districts first treated in period g with schools not-yet-treated at that time, so that our comparison group consists of both schools in districts that are never-treated over the study period and schools in districts that are treated in an academic year after period g.¹⁰ In standard DiD, notation, the equation we estimate is:

$$Y_{idt}={\beta }^{\star }[SBH{C}_{idt}orAWAR{E}_{idt}]+{\alpha }_i+{\delta }_t+\gamma {\mathit{X}}_{it}+{\epsilon }_{idt}$$

where Y is the outcome; i represents a school in a district d; t is time in years; ${\alpha }_{\mathrm{i}}$ are school fixed effects; ${\delta }_{\mathrm{t}}$ are year fixed effects, and ${\mathrm{X}}_{\text{it}}$ are time-varying school characteristics.

Of the three alternatives to model estimation described by Callaway and Sant’Anna (2021), we employ the doubly robust approach. This approach combines the outcome regression and inverse probability weighting approaches, but it only requires the correct specification for one of the two, minimizing the consequences of potential model misspecifications. The doubly robust estimation procedure requires a first step estimation of the generalized propensity score and the outcome regression, and a multiplier bootstrap procedure is then used to construct confidence bands for the group-time average treatment effects, (i.e., $ATT(g,t)$ estimates). We used this doubly robust DiD estimator with stabilized inverse probability weighting and wild bootstrap standard errors. We adjust for some school characteristics, including school level (elementary, middle and high), demographic composition of the study body, and proportion of students for whom English was not their first language, as these characteristics may co-vary with both the probability of adopting an SBHC or AWARE grant and our outcomes.

2.5. Qualitative methods

To inform and contextualize our quantitative results, we use data from semi-structured interviews (n = 58) conducted between September 2023 and May 2024. Interviewees were primarily directors of coordinated school health within each district. We first prioritized school districts identified in our census as having an SBHC or AWARE grant for the interviews. We completed interviews with each of the seven AWARE grantees, and approximately two-thirds of the other districts interviewed reported having active SBHCs. The interview protocols, shown in online Appendix 2, include questions about the types of physical, mental, and behavioral health services offered to students, personnel and partnerships with external agencies, funding sources, identification and referral processes for students needing health services, staff and school trainings, and barriers encountered in providing and sustaining services. Transcripts and informal reflexive memos were developed for each interview, followed by multiple rounds of coding and sub-coding using NVivo programming software (Miles et al., 2020). In addition, we conducted eight district site visits to observe different types of SBHCs—school-based, school-linked, mobile health and telehealth—in both urban and rural contexts, including two districts that received AWARE grants. The purpose of the site visits was to better understand the health services infrastructure, scope of services, and the contexts within which they operate.

3. Findings on effects of school-based health centers and AWARE grants

Table 3 presents a summary of the findings of our doubly robust DiD analyses for mental health, healthcare utilization, and education outcomes. We distinguish estimated effects of SBHCs and AWARE grants that we interpret as plausibly causal because we satisfy the conditional parallel trends assumption from estimated associations that do not satisfy this assumption (shown in italics). Statistically significant estimates (at $\alpha <0.05$) are reported in boldface. In addition, if we satisfied the conditional parallel trends assumption for a given outcome, we also reported any statistically significant group average effects (where a group is defined by the year the SBHC or AWARE grant began).

Table 3.

Summary of effect estimates (or associations) for any SBHC or AWARE grant.

	Estimates for any SBHC or AWARE grant
Outcome (school-level)	Baseline average (percent)	Intervention effect (percentage pts.)	Percent change relative to baseline
Any (diagnosed) mental health conditions – all (n = 19,151)	6.9	−0.415 (0.172)	−6.0
Group average effects
Start year 2009	5.0	−1.001 (0.323)	−20.0
Start year 2012	8.1	−0.553 (0.249)	−6.8
Start year 2014 (3 AWARE grant recipients)	11.8	0.983 (0.433)	8.3
Any (diagnosed) behavioral health conditions – all (n = 19,151)	8.0	0.000584 (0.071)	0.007
Group average effects
Start year 2015	0.95	−0.240 (0.106)	−25.0
Preventive care visits (n = 19,151)	20.7	2.429 (0.424)	11.7
Absence rate (n =19,109)	6.4	−0.395 (0.716)	6.2
Chronic absence rate (n = 19,245)	19.7	−0.436 (0.617)	2.2
Any disciplinary incident rate - HS only (n = 4236)	16.4	−0.985 (1.433)	6.0
Suspension rate – HS only (n = 4295)	14.5	−1.437 (1.408)	9.9

Notes: effect estimates in boldface are statistically significant at $\alpha <0.05$; estimates in italics do not meet the conditional parallel trends assumption.

3.1. Mental and behavioral health conditions

The DiD model results suggest that having an SBHC or AWARE grant reduces the prevalence of mental health conditions in these schools compared to those that do not introduce these interventions. The average treatment effect on the treated (ATT) for the group of schools treated at time $g$ in academic year $t$ for this outcome is −0.415 percentage points and statistically significant, with a standard error of 0.172. Given the baseline average rate of mental health conditions in the study sample of 6.9 percent, this indicates that school districts with SBHCs and/or AWARE grants saw a 6.0 percent overall reduction (−0.415/6.9) in the rate of diagnosed mental health conditions after their introduction, compared to school districts without these interventions. The conditional parallel trends assumption is satisfied in this analysis.

Fig. 1 displays the model results graphically for group-time (school district-academic year) average treatment effects, including 95 percent confidence interval bands (via wild bootstrap standard errors). The figure distinguishes between pre-treatment estimates of the average outcomes and post-treatment estimates of the effects of SBHCs and AWARE grants on any mental health diagnosis in the post-treatment periods. Some of the standard error bars on the post-treatment effect estimates are noticeably larger, likely reflecting smaller sample sizes for some cohorts/start years or greater variation in effects within groups.

Fig. 1.

Group-time (school-academic year) average treatment effects: any diagnosed mental health conditions.

We investigate how average treatment effects vary by group (SBHC/AWARE grant start dates) using the partial aggregations generated by this estimation approach. In Table 3, we report the statistically significant group average treatment effects, which includes school districts with interventions starting in 2009 and 2012 and three school districts that received AWARE grants in 2014. The effect estimate for districts treated starting in 2009 (in which 86 percent of the schools were in Memphis-Shelby County) was larger—a 1 percentage point reduction in the rate of any diagnosed mental health conditions (on average) compared to schools without these interventions—and represented a 20 percent reduction in diagnosed mental health conditions from the baseline rate of any diagnosed mental health conditions in schools first treated in 2009. The ATT for schools in five school districts with interventions starting in 2012—one of which was Davidson County (Nashville) that accounted for about 89 percent of schools first treated in this academic year—was also larger, indicating a 0.553 percentage point reduction (or a 6.8 percent reduction relative to baseline) in the rate of any diagnosed mental health conditions on average compared to school districts without SBHCs or AWARE grants. Alternatively, the effect estimate for schools that first received an AWARE grant in 2014 indicates that diagnoses of mental health conditions rose by nearly 1 percentage point (0.983) on average (or 8.3 percent from baseline), potentially reflecting newly developed capacities and staffing for identifying students with mental health needs.

We did not identify a statistically significant average treatment effect of having an SBHC or AWARE grant on the rate of behavioral health conditions (compared to schools without these interventions over our study period). We did, however, identify a statistically significant group average treatment effect for schools with interventions starting in 2015, consisting of mostly (95 percent) schools in Hamilton County, the Chattanooga metro area. The estimated effect for schools first treated in this year was −0.240 ppts., reflecting a 25 percent reduction relative to the baseline rate of any diagnosed behavioral health conditions for schools with an intervention that started that year.

In Appendix 2, we present the results for the rate of diagnosed mental health conditions and the rate of diagnosed behavioral health conditions estimated by school level. Although in the estimation with these smaller, school-level subsamples, we do not satisfy the DiD parallel trends assumption, there are notable differences in the estimated associations by school level. Specifically, the reduction in the rate of diagnosed mental health conditions is substantially larger among middle school students, and the decreases in the rate of diagnosed behavioral health conditions is largest among high school students. The latter result is expected, given that school-based interventions addressing these behaviors (substance use disorders, tobacco use, eating disorders, sexually transmitted infections, and pregnancy) are more frequently targeted to high school students.

3.2. Preventive care visits

We also estimated the DiD model to understand the average treatment effect of having an SBHC or AWARE grant on the rate of Medicaid-covered preventive health care visits, although our qualitative research suggests some caution is in order in interpreting the results. For many school districts, garnering additional resources to expand school-based health services has allowed them to extend services to students regardless of their insurance status. Some school staff explained that they were not billing Medicaid for these services, in part because of the administrative burdens associated with doing so and their preference to use school-based health staff for healthcare. Thus, these estimates could understate any increases in preventive health care access. In addition, this model did not satisfy the conditional parallel trends assumption, so we interpret the result as an association. The statistically significant estimate reported in Table 3 indicates that the average association of the introduction of SBHCs and AWARE grants with the rate of preventive health care visits is 2.429, implying that these interventions are associated with an 11.7 percent increase in preventive health care visits over the baseline rate (compared to schools without these interventions). Online Appendix 3 shows that when estimated by school level, the estimated association of SBHCs and AWARE grants with the rate of preventive health care visits is largest for elementary students (3.376), which is consistent with the routine practice of conducting annual, preventive health screenings for students in kindergarten, 2nd, 4th, 6th and 8th grades.

3.3. Education outcomes

Our analysis focused on education outcomes of critical concern for school districts today: absence and chronic absence rates and rates disciplinary incidents and suspensions (estimated for high schools). As shown in Table 3, we did not find any statistically significant average effects of having an SBHC or AWARE grant on either absence rates or discipline rates. We also did not satisfy the conditional parallel trends assumption for these outcomes. The estimated coefficients are all negative, that is, in the direction that would imply a reduction in absences and disciplinary incidents. However, we cannot draw any conclusions given the lack of precision in the estimates and absence of conditional parallel trends prior to treatment. The same is true for the estimates by school level (shown in Appendix 3).

4. Discussion and study limitations

4.1. Qualitative insights on study findings

We leveraged variation in the timing of the introduction of SBHCs and AWARE grants in Tennessee school districts to estimate average treatment effects of these interventions, including group averages by the year they were introduced. The results indicated a statistically significant reduction in diagnosed mental health conditions among treated schools of 6 percent relative to their baseline prevalence. From our interviews, we know that schools typically use these monetary resources to increase health staffing in schools—such as counselors, social workers, therapists, and behavioral health specialists—that enable more students experiencing mental health challenges to interact with appropriate professionals who can address their needs. Early identification of risk factors, detection of mental health needs, and effective implementation of prevention strategies can prevent conditions from reaching a diagnosable level (Cabral and Patel, 2020). In addition, by helping children manage conditions that do not meet the criteria for an official diagnosis, schools may reduce the likelihood of an official diagnosis and the development of comorbid conditions. In our interviews, school district staff frequently described preventive efforts that included equipping students with tools to understand and manage their emotions through individual and group counseling sessions and classroom lessons on emergent behavioral and mental health concerns.

For the two years in which the largest school districts in Tennessee introduced SBHCs, we identified larger post-treatment reductions in diagnosed mental health conditions. In 2009 (Memphis/Shelby County) and 2012 (Nashville/Davidson County; along with three smaller county school districts served by a mobile health provider), the effect estimate was 42% and 20% larger than the overall ATT, respectively.

In our interviews, we learned that Memphis-Shelby County School District has had as many as four SBHCs and currently has two SBHCs and three full-service mental health wellness centers. The SBHCs not only offer comprehensive primary care services to all students, families, and community members (year-round and outside of school hours), but the district also has a large staff of mental health professionals—more than 60 social workers, 10 psychologists, and licensed drug and alcohol counselors—who serve students year-round. Referrals and telehealth are used to provide access to specialized care. Mental health staff are assigned to schools in teams (including counselors, social workers, behavioral specialists, etc.) and cultivate relationships with students and other staff to reduce barriers to engaging with mental health supports and to target and refer students for services. Metro Nashville-Davidson County schools have similarly drawn on partnerships with local nonprofit hospitals and healthcare organizations to offer school-based and school-linked primary care and mental health services, as well as mobile health and telehealth services.

We also observed a statistically significant estimated increase in diagnosed mental health conditions after AWARE grants were awarded to three county school districts in 2014. We have interviewed staff from the seven Tennessee school districts that received AWARE grants that are included in the study period. We consistently heard throughout the interviews that these grants usually provide the first opportunity for the districts to hire mental health staffing or to establish partnerships that facilitate identifying and serving students’ mental health needs, which might explain the rise in diagnosed conditions. As one Project AWARE coordinator expounded: “This is honestly the first year [after receiving the grant] we have been capable of serving students’ mental health needs, and we already need more therapists. I mean, caseloads are already more than they can handle, and there’s nobody here in rural northwest Tennessee to refer to or even hire.” Although only one of the AWARE grantees in our study sample created an SBHC with the support of the grant, school district staff described their efforts to sustain mental health personnel and programming after the grant’s end. For example, one of the AWARE grantees in our study sample was able to retain three of four social workers that were hired through the grant with district-provided funding.

Although we did not find any statistically significant average effects of having an SBHC or AWARE grant on the rate of behavioral health conditions in Tennessee schools, we found a comparatively large group average treatment effect for schools in the Chattanooga metro area (Hamilton County). Interviewees in this school district described partnerships with seven different providers of mental and behavioral health services that launched since the SBHC’s introduction in 2015, as well as coordinated processes for student referrals and robust training for and collaboration among mental health staff and school counselors, social workers, and nurses. Similar to Memphis-Shelby County Schools, Hamilton County developed student success teams to provide health supports that are personalized for each child, with a “whole child” focus, as well as integrated into the curriculum.

One other statistically significant association in our analysis suggested that SBHCs and AWARE grants might contribute to an increase in preventive health visits among school-aged children in Tennessee. Preventive programming and services were a focus in nearly every interview with school district staff, and most SBHCs were partnered with federally qualified health centers, children’s hospitals, university training programs, or other providers to deliver regular preventive health screenings and programming to students on campus. Mental and behavioral health were increasingly key components of the preventive health visits or programs, including on topics such as suicide awareness and prevention and bullying prevention.

Our findings on how SBHCs or AWARE grants have affected education outcomes in Tennessee—absences, chronic absences, and disciplinary incidents—were also described as associations because we did not satisfy the conditional parallel trends assumptions for these outcomes, and we also did not observe any statistically significant associations. This may reflect, in part, selection into or targeted allocation of funding for school-based interventions among school districts struggling with especially high absence rates and disciplinary problems. In interviews, school districts with SBHCs or AWARE grants described concerted efforts to address chronic absences and adopt trauma-informed approaches and training to reduce exclusionary discipline practices. For example, one district created “reset spaces” at middle and high schools to allow students time and space to process and reorient themselves mentally and emotionally, with the aim to reduce absences and suspensions. Students are also required to check in with counselors when entering and leaving a reset space, which helps in tracking their mental health needs.

4.2. Study limitations

Our study focused on an analytic period of 2006 through 2019, before the COVID-19 pandemic, which disrupted school-based health services. The 2019 end date to our data precludes us from estimating effects for SBHCs and AWARE grants established after 2019, and it also limits our ability to understand potential effects for those with only one year of treatment. In addition, our data only capture health outcomes (diagnoses) for low-income students that are billed to Tennessee’s Medicaid program.

Unobserved differences between our treatment and comparison groups are another important limitation of our analysis. These differences preclude us from satisfying the conditional parallel trends assumption and generating causal estimates for healthcare utilization and education outcomes. Strategies employed to “detrend” these outcomes in advance of estimation did not fully resolve the selection issues. This may reflect the careful investment of resources for health services in areas with the greatest need-to-resource gaps. In fact, the pre-period differences in outcomes and other observable characteristics by SBHC or AWARE grant status highlight the success of efforts to target resources in areas of high need.

5. Conclusions

In this study, we found that the implementation of SBHCs and AWARE grants reduced mental health diagnoses among children in school districts with these resources in Tennessee between 2006 and 2019, advancing knowledge on the effects of these programs. Specifically, we expanded on outcomes used in prior work by using linked administrative data sources instead of student self-report data. Overall, our results suggest that SBHCs and/or AWARE grants may reduce the prevalence of mental health conditions in school districts that introduce these interventions (compared to those that do not). We were not able to draw causal conclusions about effects on preventive/wellness visits and education outcomes, although the estimates (interpreted as associations) were in the expected direction, e.g., districts with these interventions saw increases in wellness visits.

More generally, we heard in every interview with school districts that timely and adequate infusions of external resources are critical for expanding school-based capacities to address children’s health and education needs. Our qualitative findings stressed the importance of federal resources in supporting the hiring, training, and retention of health staff in public schools, especially staff to address children’s mental and behavioral health needs. Many school district staff conveyed that in the absence of these funds, they would not have had the resources to identify students that need these services, and they also emphasized how challenging it is to sustain the services. As one school health staff member explained, “it’s money, money, money. Funding is a major barrier to providing comprehensive primary care and mental or behavioral health services,” and healthcare organizations will not partner with school districts to provide services if they are not making enough money. Given that schools are a gateway to youth access to mental health services and frequently serve as a “medical home” for the least advantaged, our findings suggest that substantial, multi-year commitments of funding—such as the five-year, $9.5 million federal School-Based Mental Health Services (SBMH) grant that one Tennessee school district recently received—should be reauthorized and expanded by federal legislation to serve students across the U.S.