Identifying Precise Targets to Improve Child Mental Health Care Equity

INCREASING TARGET PRECISION WHILE ENVISIONING SOCIAL DETERMINANTS OF HEALTH AS MODIFIABLE

Reducing exposure to social determinants of health—social, economic, environmental, and community conditions—may have a stronger influence on population health compared with services delivered by clinicians and health care delivery organizations.¹ However, often child mental health research is encouraged to focus on identifying targets that are mutable or “potentially modifiable.”² This juxtaposition raises questions including (1) what is a modifiable target? and (2) if research identifies precise child and neighborhood-level characteristics that are predictive of poor access, quality, and outcomes of child mental health care, will these findings translate to the political will necessary to invest in extensive transformation of the child mental health care system and communities? Examining variation in child mental health service use and quality by race and ethnicity signals, does not identify, precise underlying causes of disparities. Sociodemographic characteristics alone do not integrate a developmental perspective that is essential to examine how timing, duration, and intensity of exposure to social determinants of health impacts child mental health disparities and outcomes over time.

The pressing need for innovation in child mental health disparities research is evident by the lack of progress over decades to reduce inequity in risk for devastating clinical outcomes among children. Recommendations from the Institute of Medicine’s (IOM) 2003 landmark report, Unequal Treatment,³ has not been achieved, especially for child mental health access, quality, and outcomes.⁴ The IOM framework defined disparity to be any difference in treatment or access not justified by differences in health status or preferences of the groups.^3,5 Using data from the Medical Expenditure Panel Survey, significant Black–White and Hispanic–White disparities in child mental health care have persisted in the United States for 15 years.^6,7 In addition, between 2008 and 2020, the rate of death from suicide among adolescents aged 12 to 17 years increased by 70.3%, with greater worsening of unmet need for depression treatment among Hispanic compared to White teens (relative difference 23.8) and children living in poverty versus not (relative difference 30.3).¹ From 2003 to 2017, Black youth experienced a significant upward trend in suicide, with the largest annual percentage change among youth aged 15 to 17 years and girls (4.9% and 6.6%, respectively).⁸

Higher child mental health need during the COVID-19 pandemic escalated the strain on a child mental health care system that is often inaccessible, ineffective, and inequitable.{Cummings, 2019 #3087} The need for child mental health care substantially rose⁹ and remained high during the subsequent 1 to almost 2 years.¹⁰ Pediatric emergency department visits and hospitalizations for primary mental health disorders disproportionately increased,{Leeb, 2020 #2661;Zima, 2022 #2931} especially for suicide or self-harm among girls and adolescents.{Zima, 2022 #2931} Improving access, quality, and equity of child mental health care was identified as a national priority area by a declaration of a state of emergency in child mental health,¹¹ US Surgeon General Advisory reports,¹² and the National Disparities Report.¹ In this context, Benton and colleagues warn that the long-term impact of the COVID-19 pandemic on children will be “more devastating without urgent action,” and prioritizing child mental health care will demand a transformation in societal drivers and system-level solutions.¹³

The disproportionate impact of the COVID-19 pandemic on families also demands critical reflection. Early literature identified a disproportionate negative impact of the pandemic on children from minoritized race and ethnic groups as well as immigrant communities.{Advisory, 2021 #2916} Rates of COVID-19 hospitalizations and deaths were 3 to 5 times higher among Black, Hispanic, and Asian patients compared to White patients,¹⁴ placing minoritized children at greater risk for premature loss of their primary caregiver and other family members.¹⁵ Prolonged length of stay for mental health emergency department visits also increased, especially for Hispanic children,¹⁶ and Black and Hispanic children were more likely to be brought to mental health care by police.¹⁷ Toxic stress resulting from racial and social inequities was magnified during the pandemic, with implications for poor physical and mental health and socio-economic outcomes, underscoring the need for a trauma-informed social justice response.¹⁸

The proposed mechanisms for worsening of disparities in need for child mental health care posit disproportionate exposure to social determinants of health among racial or ethnic minoritized children, such as parental COVID-19 mortality/morbidity, poverty, poor health care access, parental unemployment, parental jobs that cannot be performed remotely, loss of insurance, racial segregation, housing and food insecurity, maltreatment, exposure to traumatic events, and poor Internet access.^14,15 Exposure to a broad set of adverse childhood experiences is more likely among Black and Hispanic children.¹⁹ COVID-19-related disruption in schooling likely increased child’s risk for mental health problems and educational delays, especially for younger children and those with developmental disorders{Christakis, 2020 #2692} but findings are mixed.{Xiao, 2023 #3098} Further, structural racism has a profound impact on many social determinants of health, especially as children are dependent on their family system and environment to provide space for healthy development.²⁰ Availability of school-based mental health services to reduce disparities is promising, as racial/ethnic differences in school-based service use are less as compared to significant differences in other clinical settings.²¹ During the COVID-19 pandemic, living in economically deprived neighborhoods, disrupted mental health treatment, having separated parents, and inability to afford food were associated with poor child mental health.²² In addition, mental health workforce and availability of outpatient mental health clinics vary by high and low community income quartiles.²³ Nevertheless, little is known about how neighborhood exposure by type, intensity, and duration to social determinants of health contribute to child mental health disparities in service use and quality over time to identify precise targets to effectively reduce disparities.

Even among studies that have been a beacon to support the public health significance to reduce child mental health disparities, there are significant gaps in child mental health disparities research as well as potential solutions (Table 1). Limitations include the lack of a developmental perspective in the IOM disparities framework, restriction to select disorders, poor representation of community-based mental health programs and safety-net hospitals, and variable data quality. Recent studies using composite indexes of child social determinants of health suggest that higher area measures of child opportunity (ie, greater gaps in the quality of resources and conditions in a child’s neighborhood that matter for healthy development) are associated with repeat mental health emergency department visits,²⁴ pediatric hospitalizations for asthma,²⁵ ambulatory sensitive conditions,²⁶ pediatric emergency department (ED) visits with medical complexity,²⁷ and abrupt decline in pediatric ED use following onset of the COVID-19 pandemic.²⁸ However, a composite index signals but does not identify type, intensity, or duration of child exposure to specific neighborhood-level social determinants of health, such as violent crime, food insecurity, or lack of green space.

Table 1.

Child mental health disparities research gaps and potential solutions

Research Gaps	Potential Solutions
Statistical models to apply the IOM “need/non-need” framework to examine mental health disparities have been generally limited to adults using self-reported mental health symptoms, substance abuse, and service use from survey data.49	Integrate a developmental framework, such as the Life Course Health Development Framework,50 with the IOM disparities framework.3
Use of aggregated (vs individual child level) service use data51,52 and inclusion of only one health care system.53 The cumulative impact of community-level social determinants of health and individual disparities in child mental health use and need in the postpandemic era has been hypothesized but not yet empirically investigated.13	Use longitudinal cohort study designs with study enrollment sensitive to onset and aftermath of COVID-19 pandemic and integrate electronic health record data across health care systems linked with public use data files to track unique children over time.
Restriction to children enrolled in Medicaid.54	Include private, public, and no insurance status and change in status/time per child using structured electronic health record data.
Restriction to only a few child mental health disorders.54 Reliance on parent- and/or child-report surveys using screening questions for mental health symptoms and broadly defined prior service use.22	Build a more detailed clinical profile for each child encounter over time using ICD-10 codes to identify suicide/self-injury and 23 mental health disorders (including autism, substance use)55 and standardized approaches to assess medical and psychiatric comorbidity.52
Restriction to a few target social determinants of health and outcome is poor child mental health, not mental health service use or quality.22 Linkage often relies on variables conveniently extrapolated from adult health conditions that may miss key contributors of children’s mental health use (eg, access to child mental health providers and school-based resources). A composite index of child opportunity may signal but does not identify type, intensity, or duration of child neighborhood-level exposure to specific social determinants of health predictive of child mental health disparities over time.24	Include a larger number of environmental social determinants of health from public data files (child’s neighborhood level), organized by the National Institute of Minority Health and Disparities (NIMHD) Framework,56 that are linked temporally and geographically to electronic health care record data (child level).
Linkage of health data with publicly available sources on local geographic levels has scarcely been examined as a means to improve risk prediction of child mental health outcomes and not been applied to child mental health through a health equity lens.	Include future risk prediction models of disparities in child mental health service use and quality data on detailed clinical profile for each encounter and a larger number of neighborhood-level social determinants of health.
Interaction between community-level and individual-level vulnerabilities to bias and inequity in pediatric mental health is not yet well understood in light of concerns raised for misprediction of risk among people of color.	During the data analyses, evaluate the extent of model biases, make appropriate adjustments to address, and acknowledge limitations in generalizing models to produce unbiased predictions in different populations.

LEVERAGING ADVANCEMENTS IN CLINICAL RESEARCH INFORMATICS

Identification of drivers of disparities in child mental health service use and quality within an actionable timeframe is an important but difficult task. Clinical research informatics (CRI) is a specialized field in biomedical informatics that focuses on the development, use, and evaluation of information systems and technologies to support clinical research and health delivery.²⁹ The field involves the design and implementation of databases and data management systems to collect and store clinical research data, ensuring data integrity, security, and compliance with regulatory standards. Moreover, CRI includes development of information retrieval tools and techniques to query electronic health records (EHRs), clinical trial databases, and other health care data sources to support research inquiries.³⁰ Clinical decision-making may be supported by CRI by incorporating clinical decision support systems to assist health care providers in making decisions about evidence-based interventions for patient care.³¹ Capacity to geolink clinical data with public use data files has accelerated the application of CRI approaches to model social vulnerability and drivers of disparities across health outcomes.

Bridging advancements in clinical informatics with measurement of child mental health service use and quality holds promise for discovery of contributors of disparities at the child and neighborhood level.³² Integration of child mental health data from primary and specialty mental health care into EHRs enables a view into children’s longitudinal history, including health assessments, treatments, and outcomes. CRI has enabled the creation of data warehouses that store this information as well as application of analytical methods, such as machine learning, to detect trends and risk factors associated with mental health and neurodevelopmental conditions in children. Examples include EHR-based early autism detection at age 30 days predicting autistic spectrum disorder (ASD) diagnosis at 1 year,³³ prediction of involuntary mental health detainment use and continuation among children 10 to 17 years,³⁴ and detection of adolescent suicide attempts using discharge summaries.³⁵ First developed to enable genomic and biomedical data analysis in large-scale multisite studies, informatics now encompasses indicators of social vulnerability at the individual patient and, linked by geography, the neighborhood level. Linkage of EHRs with publicly available datasets has demonstrated the potential of pediatric care use and quality by geography, such as geospatial identification of neighborhood hot spots associated with pediatric intensive care use, green space and changes in pediatric obesity during the pandemic, and neighborhood opportunity and primary mental health rehospitalization.³⁶

The use of CRI and associated computational methods has only recently been applied to child mental health service use and quality, with most applications focusing on adults.³⁷ Recent work has demonstrated the “in vitro” feasibility of EHR-based risk prediction to pediatric suicide attempts.³⁸ Another study examined emergency department visits by children aged 10 to 17 years and discovered poor sensitivity of International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) and chief complaints in detecting self-injurious thoughts and behaviors, with significantly poorer detection among male children (vs female children) and preteens (vs adolescents).³⁹ Real-world implications of predictive models for bias and inequity in pediatric care are not yet well understood despite concerns raised for misprediction of risk among people of color.³¹ A 2021 comparison of EHR-based suicide risk prediction among adults discovered a substantial gap in sensitivity (62.2% vs 10.0%) comparing White and Black patients.⁴⁰

To “first do no harm,” scrutiny of new analytical and methodological approaches tailored for child mental health applications is necessary when examining the complex multifactorial pathways and cumulative impacts responsible for disparities over time.⁴¹ Also important is to establish a foundation for data standardization in this process.⁴² Standard common data models delineating discrete data readily available for inclusion in analytical models currently omit key variables relevant to child mental health (eg, primary caregiver, school type and performance, justice system involvement, and child protective service contact). Table 2 describes the potential toolset offered by CRI to investigate contributors to disparities in child mental health service use and quality, along with foreseeable challenges to this application.

Table 2.

Examples of advancements In clinical Informatics with potential to Increase precision and equity In child mental health care

	Potential Gains	Challenges
Data Integration	Integration of data sources, including electronic health records, publicly available data on social determinants of health (school district, exposure to violence, police contact, access to firearms, green space, climate, and natural disaster exposure), large-scale survey data, and other population health datasets.	Variation in data format, standards, and quality. Ensuring data consistency and resolving discrepancies are a complex task.
Demographic Profile	Creation of detailed demographic profiles of child populations, such as primary language, sexual orientation and gender identity, primary caregiver, religion, pronouns, and caregivers.	Obtaining accurate information, especially for vulnerable populations may be hindered by incomplete, false, or outdated information.
Geographic Mapping	Geographic information systems can geospatially represent child mental health service utilization patterns and reveal neighborhood-level disparities and correlation with social determinants.	Connecting individual patient level-data to larger geography may lose granularity or insufficiently represent the distinction between individual identity and the individual’s community.
Longitudinal Analysis	Longitudinal tracking of individual child mental health trajectories, uncovering disparities in continuity and quality of care over time.	Requires access to historical data, which may not be consistent or available, especially for populations receiving care across multiple settings (eg, clinical and school-based).
Natural Language Processing	Analysis of unstructured clinical notes to identify social determinants, contributing factors, and patient experiences, shedding light on disparities not captured by structured data alone.	Difficulty deciphering complex and nuanced information in unstructured clinical notes requires contextual understanding and handling of medical jargon. Documentation may reflect biases of the provider and vary by type of encounter.
Machine Learning	Machine learning algorithms enable detection and prediction of child mental health service utilization patterns and quality outcomes incorporating numerous variables (features) across domains.	Supervised approaches require high-quality labeled data, which can be limited or biased, potentially affecting model accuracy. Unsupervised approaches may be biased toward majority classes.
Intervention Development	Development of targeted clinical decision support interventions by identifying specific modifiable contributors to disparities in clinical workflow patterns.	Identification of specific modifiable contributors is a multifaceted challenge and interventions may need to address complex social, economic, and cultural factors that may vary widely.
Equity Metrics	Creation of metrics to measure progress in reducing child mental health disparities and benchmark effectiveness of targeting modifiable factors over time.	Metrics necessitate consensus on conceptual frameworks, measurement standards and definitions of disparities. Defining an equitable outcome can be subjective and context dependent.
Stakeholder Collaboration	Integration of perspectives from health care providers, community organizations, and public health agencies to promote interdisciplinary collaboration and comprehensively address child mental health disparities.	Involves addressing privacy concerns, data-sharing agreements, data governance and priorities between health care providers, organizations, and agencies that may have differing levels of resources for data infrastructure and oversight.

LEVERAGING THE INSIGHTS OF LIVED EXPERIENCE

Participatory action research (PAR) and other community-engaged research approaches have become increasingly important in the field of mental health service studies. PAR is an umbrella term for research designs, methods, and frameworks that use systematic inquiry in direct collaboration with those affected by the issue being studied for the purpose of action or change.⁴³ PAR can promote representation of underserved groups in informatics research and practice, improve the relevance and effectiveness of interventions, and drive study approaches in all phases including the identification of relevant study variables, risk and protective factors, and data interpretation.⁴⁴ Drawing on community knowledge creates space for translating lived experiences into research ideas and for principal investigators to be aware of their privilege and promote power sharing.⁴⁵ This results in research questions and outcomes that align with the needs and priorities of the community, increasing the likelihood of successful implementation and sustainability of resulting mental health interventions that address disparities. A related approach is participatory health informatics (PHI), which is a multidisciplinary field that uses information technology as provided through the web, smartphones, or wearables to increase participation of individuals in their care process, shared decision-making and even for establishing research priorities and analysis.⁴⁶ The following brief example illustrates the potential synergy that can occur when combining PAR and informatics approaches.

By Youth, For Youth Study (Fortuna and Kataoka)⁴⁷:

With their unrivaled ability to reach youth, school-based and pediatric primary care services are ideal hubs to provide mental health, health care, social services, and prevention to students and families who otherwise face barriers to care. Mobile technology approaches are gaining empirical support and hold great potential for enhancing mental health navigator models within these settings. Using community partnered participatory informatics methods, the University of California psychiatry research centers with Los Angeles Trust for Children’s Health are conducting a study to (1) fully co-design (with youth, caregivers, clinicians, and other stakeholders) an innovative mental health digital app, called 4Youth, to implement algorithmic and machine learning supported mental health and social determinants screening and delivery of mental health tools plus navigator support to help the primary care clinical workforce within school centers and pediatric services and (2) research the implementation of the app plus family navigator support for improving access to care and matching youth to the right level of care and supports. This project was initiated with youth aged 13 to 22 years and family and community members through schools and primary care clinics that serve mostly Black, Hispanic, and Asian youth. PHI approaches included youth and caregivers in the prioritization of health outcome variables, designing social determinants of health tools, designing app features and content.⁴⁶ A successful outcome of the project will be a PHI-developed intervention implementable in school-based and pediatric primary care services, for improving mental health services and prevention resources access for minoritized youth.

PAR and community-engaged research require time, resources, and strong partnerships with community organizations. Researchers must navigate ethical considerations and power dynamics within the research process and realize the importance of sharing authority, authentic partnership, and responsibility in the research process. There is potential for cultural understanding to diverge, even between researchers and stakeholders from similar minoritized backgrounds. Therefore, the ethics and principles of collaborative research with minoritized and traditionally underserved communities include transparency, 2 way communication, and commitment to social justice, trustworthiness, and equity.⁴⁴

SUMMARY AND RELEVANT CLINICAL POINTS

Going forward, addressing escalating disparities in child mental health care has emerged as a pressing concern, necessitating a paradigm shift in approach. “Humility is essential” in delineating a national strategy to reduce child mental health disparities for the postpandemic years.⁴⁸ To effectively combat this challenge, we can use the following data-driven approaches to improve research and clinical care.

• Integrate new computational methods and the systematic analysis of large-scale health data, including granular publicly available data linked to geography, to identify precise targets to improve child mental health.

• Harness the power of innovative technological solutions to enable health care providers to deliver more timely and tailored interventions and mitigate the cumulative impact of disparities in children’s mental well-being.

• Incorporate insights from individuals with lived experiences to shape a holistic understanding of these disparities and ensure that interventions are not only data-driven but also culturally sensitive, patient-centered and relevant, so as not to replicate existing disparities.

• By synergizing these elements, we embark on a transformative journey toward a more equitable future for child mental health care, where evidence, innovation, and empathy converge to address the challenges posed by the pandemic’s aftermath.

KEY POINTS.

• Higher child mental health need during the COVID-19 pandemic escalated the strain on a child mental health care system that is often inaccessible, ineffective, and inequitable.

• Identification of the drivers of disparities in child mental health service use and quality within an actionable timeframe requires innovative and efficient research methods.

• Clinical research informatics is a specialized field in biomedical informatics that focuses on the development, use, and evaluation of information systems and technologies to support clinical research, including disparities research, and health delivery.

• Community participatory research approaches promote representation of underserved groups in informatics research and practice and may improve the relevance and effectiveness of these interventions to address mental health needs and disparities.