Equity in Digital Mental Health Interventions in the United States: Where to Next?

Health care technologies have the ability to bridge or hinder equitable care. Advocates of digital mental health interventions (DMHIs) report that such technologies are poised to reduce the documented gross health care inequities that have plagued generations of people seeking care in the United States. This is due to a multitude of factors such as their potential to revolutionize access; mitigate logistical barriers to in-person mental health care; and leverage patient inputs to formulate tailored, responsive, and personalized experiences. Although we agree with the potential of DMHIs to advance health equity, we articulate several steps essential to mobilize and sustain meaningful forward progression in this endeavor, reflecting on decades of research and learnings drawn from multiple fields of expertise and real-world experience. First, DMHI manufacturers must build diversity, equity, inclusion, and belonging (DEIB) processes into the full spectrum of product evolution itself (eg, product design, evidence generation) as well as into the fabric of internal company practices (eg, talent recruitment, communication principles, and advisory boards). Second, awareness of the DEIB efforts-or lack thereof-in DMHI research trials is needed to refine and optimize future study design for inclusivity as well as proactively address potential barriers to doing so. Trials should incorporate thoughtful, inclusive, and creative approaches to recruitment, enrollment, and measurement of social determinants of health and self-identity, as well as a prioritization of planned and exploratory analyses examining outcomes across various groups of people. Third, mental health care advocacy, research funding policies, and local and federal legislation can advance these pursuits, with directives from the US Preventive Services Taskforce, National Institutes of Health, and Food and Drug Administration applied as poignant examples. For products with artificial intelligence/machine learning, maintaining a "human in the loop" as well as prespecified and adaptive analytic frameworks to monitor and remediate potential algorithmic bias can reduce the risk of increasing inequity. Last, but certainly not least, is a call for partnership and transparency within and across ecosystems (academic, industry, payer, provider, regulatory agencies, and value-based care organizations) to reliably build health equity into real-world DMHI product deployments and evidence-generation strategies. All these considerations should also extend into the context of an equity-informed commercial strategy for DMHI manufacturers and health care organizations alike. The potential to advance health equity in innovation with DMHI is apparent. We advocate the field's thoughtful and evergreen advancement in inclusivity, thereby redefining the mental health care experience for this generation and those to come.

User Engagement Clusters of an 8-Week Digital Mental Health Intervention Guided by a Relational Agent (Woebot): Exploratory Study.

BACKGROUND: With the proliferation of digital mental health interventions (DMHIs) guided by relational agents, little is known about the behavioral, cognitive, and affective engagement components associated with symptom improvement over time. Obtaining a better understanding could lend clues about recommended use for particular subgroups of the population, the potency of different intervention components, and the mechanisms underlying the intervention's success. OBJECTIVE: This exploratory study applied clustering techniques to a range of engagement indicators, which were mapped to the intervention's active components and the connect, attend, participate, and enact (CAPE) model, to examine the prevalence and characterization of each identified cluster among users of a relational agent-guided DMHI. METHODS: We invited adults aged 18 years or older who were interested in using digital support to help with mood management or stress reduction through social media to participate in an 8-week DMHI guided by a natural language processing-supported relational agent, Woebot. Users completed assessments of affective and cognitive engagement, working alliance as measured by goal and task working alliance subscale scores, and enactment (ie, application of therapeutic recommendations in real-world settings). The app passively collected data on behavioral engagement (ie, utilization). We applied agglomerative hierarchical clustering analysis to the engagement indicators to identify the number of clusters that provided the best fit to the data collected, characterized the clusters, and then examined associations with baseline demographic and clinical characteristics as well as mental health outcomes at week 8. RESULTS: Exploratory analyses (n=202) supported 3 clusters: (1) "typical utilizers" (n=81, 40%), who had intermediate levels of behavioral engagement; (2) "early utilizers" (n=58, 29%), who had the nominally highest levels of behavioral engagement in week 1; and (3) "efficient engagers" (n=63, 31%), who had significantly higher levels of affective and cognitive engagement but the lowest level of behavioral engagement. With respect to mental health baseline and outcome measures, efficient engagers had significantly higher levels of baseline resilience (P<.001) and greater declines in depressive symptoms (P=.01) and stress (P=.01) from baseline to week 8 compared to typical utilizers. Significant differences across clusters were found by age, gender identity, race and ethnicity, sexual orientation, education, and insurance coverage. The main analytic findings remained robust in sensitivity analyses. CONCLUSIONS: There were 3 distinct engagement clusters found, each with distinct baseline demographic and clinical traits and mental health outcomes. Additional research is needed to inform fine-grained recommendations regarding optimal engagement and to determine the best sequence of particular intervention components with known potency. The findings represent an important first step in disentangling the complex interplay between different affective, cognitive, and behavioral engagement indicators and outcomes associated with use of a DMHI incorporating a natural language processing-supported relational agent. TRIAL REGISTRATION: ClinicalTrials.gov NCT05672745; https://classic.clinicaltrials.gov/ct2/show/NCT05672745.

Fearlessness and low social affiliation as unique developmental precursors of callous-unemotional behaviors in preschoolers.

BACKGROUND: Early callous-unemotional (CU) behaviors identify children at risk for severe and persistent aggression and antisocial behavior. Recent work suggests that fearlessness and low social affiliation are implicated in the etiology of CU behaviors, although more research is needed to clarify these etiological pathways, as well as the role of parenting. METHOD: Using a sample of preschoolers (N = 620), we examined pathways between observed fear in response to social and non-social stimuli and observed social affiliation during social interactions at age 3 and increases child CU behaviors and oppositional-defiant behaviors from ages 3 to 5. To elucidate the role of parenting in exacerbating or buffering the relationships between low fear and social affiliation and CU behaviors, we tested whether parental harshness or low warmth moderated these pathways. RESULTS: Fearlessness and low social affiliation uniquely predicted increases in CU behaviors, but not oppositional-defiant behaviors, from ages 3 to 5. Moreover, there was evidence for differential moderation of the fear pathway by harsh parenting, such that harsh parenting predicted increases in CU behaviors in fearless children but increases in oppositional-defiant behaviors in fearful children. CONCLUSIONS: Fearlessness and low social affiliation contribute to the development of CU behaviors. Harsh parenting can exacerbate the risky fearlessness pathway. Preventative interventions aimed at reducing risk for CU behaviors and persistent aggression and antisocial behavior should target socioaffiliative processes and provide parents with strategies and training to manage and scaffold rule-compliant behavior when children show low fearful arousal.

Less imitation of arbitrary actions is a specific developmental precursor to callous-unemotional traits in early childhood.

OBJECTIVE: Callous-unemotional (CU) traits in early childhood explain heterogeneity within conduct problems and are associated with higher risk for later diagnoses of childhood disruptive behavior disorders and antisocial behavior in adulthood. Emerging research implicates impairments in affiliative processes in the etiology of CU traits. The current study tests whether the imitation of intentional actions with no functional significance -a behavior that supports the acquisition of social conventions and affiliative bonds, is a specific developmental precursor to CU traits in early childhood. METHODS: Data came from a longitudinal twin study of 628 children (Age 2: 47% females; Age 3: 44.9% females) with observations of arbitrary (i.e., nonfunctional actions) and instrumental (i.e., functional actions) imitation and parent reports of CU traits and oppositional defiant (ODD) behaviors at ages 2 and 3. RESULTS: Lower arbitrary imitation at age 2, but not instrumental imitation, was related to increases in CU traits from ages 2 to 3 (ß = -.10, p = .02). CONCLUSIONS: These findings establish early social and affiliative processes in the etiology of CU traits, highlighting that novel personalized treatment and intervention strategies for CU traits may benefit from targeting these processes to help reduce CU traits and risk for persistent conduct problems in children.

Effects of socioeconomic status and executive function on school readiness across levels of household chaos.

Isolating child attributes and familial characteristics that support school readiness in children on the upper half of the socioeconomic spectrum can complement existing research on lower-socioeconomic status (SES) children and facilitate a more complete understanding of how children's performance varies across the full SES spectrum. This study examined if relations between SES, two components of executive function (EF; set-shifting and inhibitory control), and school readiness vary as a function of household chaos in 564 four-year-old children, primarily from middle-to upper-middle class families in the Northeast Region of the United States. Structural equation modeling of direct and indirect effects revealed three major findings: 1) higher levels of EF were related to better school readiness regardless of level of household chaos; 2) SES had an indirect effect on school readiness through set-shifting; and 3) household chaos was negatively associated with school readiness.

Parent- and Observer-Rated Positive Affect in Early Childhood: Genetic Overlap and Environmental Specificity.

The sources of individual differences in both observed and parent-rated positive affect (PA) were examined in a sample of 304 3-year-old twin pairs (140 MZ, 164 DZ). Based on model-fitting analyses, individual differences in observed PA were attributed to moderate genetic and high nonshared environmental factors, but not shared environmental factors. In contrast, shared environmental effects accounted for over half of the variance in parent-rated PA and genetic and nonshared environmental effects were more modest. The genetic correlation across the two measures was high, indicating substantial overlap between genetic factors influencing the two. It was these overlapping genetic effects that fully explained the phenotypic correlation between both measures. There was no significant covariance between the environmental influences on parent rated and observed PA. Thus, the two measures of PA in early childhood have common genetic underpinnings, whereas environmental influences are measure-specific. Measurement implications are discussed.

Genetic and environmental links between motor activity level and attention problems in early childhood.

Cross-lagged biometric models were used to examine genetic and environmental links between actigraph-assessed motor activity level (AL) and parent-rated attention problems (AP) in 314 same-sex twin pairs (MZ = 145, DZ = 169) at ages 2 and 3 years. At both ages, genetic correlations between AL and AP were moderate (ra2 = .35; ra3 = .39) indicating both overlap and specificity in genetic effects across the two domains. Within- and across-age phenotypic associations between AL and AP were entirely due to overlapping genetic influences. There was a unidirectional effect of AL at age 2 predicting later AP. For AP, genetic and environmental influences from age 2 were transmitted to age 3 via stability effects and from AL. For AL, across-age effects were transmitted only via stability. These results suggest that overactivity in late infancy may impact the later development of problems related to inattention, and that genetic factors explain the association between the two domains.

Tipping points? Curvilinear associations between activity level and mental development in toddlers.

BACKGROUND: The Theory of Optimal Stimulation (Zentall & Zentall, Psychological Bulletin, 94, 1983, 446) posits that the relation between activity level (AL) and cognitive performance follows an inverted U shape where midrange AL predicts better cognitive performance than AL at the extremes. METHODS: We explored this by fitting linear and quadratic models predicting mental development from AL assessed via multiple methods (parent ratings, observations, and actigraphs) and across multiple situations (laboratory play, laboratory test, home) in over 600 twins (2- and 3-year olds). RESULTS: Only observed AL in the laboratory was curvilinearly related to mental development scores. Results replicated across situations, age, and twin samples, providing strong support for the optimal stimulation model for this measure of AL in early childhood. CONCLUSIONS: Different measures of AL provide different information. Observations of AL which include both qualitative and quantitative aspects of AL within structured situations are able to capture beneficial aspects of normative AL as well as detriments of both low and high AL.

Callous-unemotional behaviors in early childhood: Genetic and environmental contributions to stability and change.

Callous-unemotional (CU) behaviors demonstrate meaningful individual differences in early childhood, even in nonclinical samples with low mean levels of CU, but the factors underlying this variation have not been examined. This study investigated genetic and environmental contributions to individual differences and to sources of continuity and change in CU in toddler twins (145 monozygotic, 169 dizygotic) assessed at ages 2 and 3 years. CU, as assessed by the Child Behavior Checklist 1.5-5 (Achenbach & Rescorla, 2000), was moderately stable across age (r = .45, p < .0001). Longitudinal biometric analyses revealed genetic and nonshared environmental influences on CU at both ages, with no significant contribution from shared environmental factors. Stability from age 2 to 3 was due to genetic factors, whereas change was due to both genetic and nonshared environmental influences. This genetic and nonshared environmental change was substantial, suggesting malleability of CU in early childhood. Over 50% of the genetic influences and 100% of the nonshared environmental influences on CU at age 3 were independent of those that operated at age 2. Implications of novel sources of variance across age are discussed.

Infant hair cortisol: associations with salivary cortisol and environmental context.

Early chronic stress has enduring implications for physical and mental health outcomes. Hair cortisol concentration (HCC) has emerged as a marker of cumulative cortisol exposure, yet HCC in infants is not well understood. We examined how infant HCC relates to widely used basal salivary cortisol measures, maternal HCC, and environmental context in 111 infants assessed at 6 and 12 months of age. Maternal HCC at 6 and 12 months was correlated with infant HCC at 12 months. At 12 months, infant HCC was positively associated with waking salivary cortisol concentration (SCC), evening SCC, and area under the curve (AUC), but was independent of diurnal slope. Breastfeeding was associated with lower HCC, whereas increased sleep disruption was related to flatter slope. Reduced nighttime sleep duration was related both to higher HCC and to flatter slope. A person-focused analysis indicated that the combination of high HCC and flattened slope was associated with more environmental risks, highlighting the importance of investigating the interplay between HCC and diurnal cortisol slope. Results support the validity of HCC as a marker of cumulative cortisol exposure in infancy, while emphasizing the value of including multiple cortisol measures assessing distinct aspects of hypothalamic-pituitary-adrenal (HPA) function.

Improvements in Work Productivity and Activity Impairment Among Adults With Anxiety or Depressive Symptoms Participating in a Relational Agent-Delivered Digital Mental Health Intervention.

OBJECTIVES: The study aimed to assess changes between baseline and end of treatment in work-related absenteeism, presenteeism, productivity, and nonwork-related activity impairment and estimate cost savings associated with observed improvements. METHODS: Data from 91 employed adult participants who enrolled in a single-arm, exploratory study of a relational agent-delivered digital mental health intervention and completed Work Productivity and Activity Impairment assessments were analyzed; overall work productivity improvement was multiplied by the overall and education-adjusted US median annual salary to arrive at potential cost savings estimates. RESULTS: Adjusted models indicated more than 20% improvements in presenteeism, work productivity impairment, and activity impairment, yielding cost-savings estimates between $14,000 and more than $18,000 annually. CONCLUSIONS: Relational agent-delivered digital mental health interventions may be associated with improvements in work productivity and activity impairment, which could result in a sizable cost savings.

Engagement, Satisfaction, and Mental Health Outcomes Across Different Residential Subgroup Users of a Digital Mental Health Relational Agent: Exploratory Single-Arm Study.

BACKGROUND: Mental illness is a pervasive worldwide public health issue. Residentially vulnerable populations, such as those living in rural medically underserved areas (MUAs) or mental health provider shortage areas (MHPSAs), face unique access barriers to mental health care. Despite the growth of digital mental health interventions using relational agent technology, little is known about their use patterns, efficacy, and favorability among residentially vulnerable populations. OBJECTIVE: This study aimed to explore differences in app use, therapeutic alliance, mental health outcomes, and satisfaction across residential subgroups (metropolitan, nonmetropolitan, or rural), MUAs (yes or no), and MHPSAs (yes or no) among users of a smartphone-based, digital mental health intervention, Woebot LIFE (WB-LIFE). WB-LIFE was designed to help users better understand and manage their moods and features a relational agent, Woebot, that converses through text-based messages. METHODS: We used an exploratory study that examined data from 255 adults enrolled in an 8-week, single-arm trial of WB-LIFE. Analyses compared levels of app use and therapeutic alliance total scores as well as subscales (goal, task, and bond), mental health outcomes (depressive and anxiety symptoms, stress, resilience, and burnout), and program satisfaction across residential subgroups. RESULTS: Few study participants resided in nonmetropolitan (25/255, 10%) or rural (3/255, 1%) areas, precluding estimates across this variable. Despite a largely metropolitan sample, nearly 39% (99/255) resided in an MUA and 55% (141/255) in an MHPSA. There were no significant differences in app use or satisfaction by MUA or MHPSA status. There also were no differences in depressive symptoms, anxiety, stress, resilience, or burnout, with the exception of MUA participants having higher baseline depressive symptoms among those starting in the moderate range or higher (Patient Health Questionnaire-8 item scale≥10) than non-MUA participants (mean 16.50 vs 14.41, respectively; P=.01). Although working alliance scores did not differ by MHPSA status, those who resided in an MUA had higher goal (2-tailed t203.47=2.21; P=.03), and bond (t203.47=1.94; P=.05) scores at day 3 (t192.98=2.15; P=.03), and higher goal scores at week 8 (t186.19=2.28; P=.02) as compared with those not living in an MUA. CONCLUSIONS: Despite the study not recruiting many participants from rural or nonmetropolitan populations, sizable proportions resided in an MUA or an MHPSA. Analyses revealed few differences in app use, therapeutic alliance, mental health outcomes (including baseline levels), or satisfaction across MUA or MHPSA status over the 8-week study. Findings suggest that vulnerable residential populations may benefit from using digital agent-guided cognitive behavioral therapy. TRIAL REGISTRATION: ClinicalTrials.gov NCT05672745; https://clinicaltrials.gov/study/NCT05672745.

When the Process Is the Problem: Racial/Ethnic and Language Disparities in Care Management.

OBJECTIVES: Achieving health equity requires addressing disparities at every level of care delivery. Yet, little literature exists examining racial/ethnic disparities in processes of high-risk care management, a foundational tool for population health. This study sought to determine whether race, ethnicity, and language are associated with patient entry into and service intensity within a large care management program. DESIGN: Retrospective cohort study. METHODS: Subjects were 23,836 adult patients eligible for the program between 2015 and 2018. Adjusting for demographics, utilization, and medical risk, we analyzed the association between race/ethnicity and language and outcomes of patient selection, enrollment, care plan completion, and care management encounters. RESULTS: Among all identified as eligible by an algorithm, Asian and Spanish-speaking patients had significantly lower odds of being selected by physicians for care management [OR 0.74 (0.58-0.93), OR 0.79 (0.64-0.97)] compared with White and English-speaking patients, respectively. Once selected, Hispanic/Latino and Asian patients had significantly lower odds compared to White counterparts of having care plans completed by care managers [OR 0.69 (0.50-0.97), 0.50 (0.32-0.79), respectively]. Patients speaking languages other than English or Spanish had a lower odds of care plan completion and had fewer staff encounters than English-speaking counterparts [OR 0.62 (0.44-0.87), RR 0.87 (0.75-1.00), respectively]. CONCLUSIONS: Race/ethnicity and language-based disparities exist at every process level within a large health system's care management program, from selection to outreach. These results underscore the importance of assessing for disparities not just in outcomes but also in program processes, to prevent population health innovations from inadvertently creating new inequities.

Cost, Utilization, and Patient and Family Experience With ACO-Based Pediatric Care Management.

BACKGROUND AND OBJECTIVES: Children and Youth with Special Health Care Needs have high healthcare utilization, fragmented care, and unmet health needs. Accountable Care Organizations (ACOs) increasingly use pediatric care management to improve quality and reduce unnecessary utilization. We evaluated effects of pediatric care management on total medical expense (TME) and utilization; perceived quality of care coordination, unmet needs, and patient and family experience; and differential impact by payor, risk score, care manager discipline, and behavioral health diagnosis. METHODS: Mixed-methods analysis including claims using quasi-stepped-wedge design pre and postenrollment to estimate difference-in-differences, participant survey, and semistructured interviews. Participants included 1321 patients with medical, behavioral, or social needs, high utilization, in Medicaid or commercial ACOs, and enrolled in multidisciplinary, primary care-embedded care management. RESULTS: TME significantly declined 1 to 6 months postenrollment and continued through 19 to 24 months (-$645.48 per member per month, P < .001). Emergency department and inpatient utilization significantly decreased 7 to 12 months post-enrollment and persisted through 19 to 24 months (-29% emergency department, P = .012; -82% inpatient, P < .001). Of respondents, 87.2% of survey respondents were somewhat or very satisfied with care coordination, 56.1% received education coordination when needed, and 81.5% had no unmet health needs. Emergency department or inpatient utilization decreases were consistent across payors and care manager disciplines, occurred sooner with behavioral health diagnoses, and were significant among children with above-median risk scores. Satisfaction and experience were equivalent across groups, with more unmet needs and frustration with above-median risk scores. CONCLUSIONS: Pediatric care management in multipayor ACOs may effectively reduce TME and utilization and clinically provide high-quality care coordination, including education and family stress, with high participant satisfaction.

Longitudinal Links Between Callous-Unemotional Behaviors and Parenting in Early Childhood: A Genetically Informed Design.

OBJECTIVE: Most research on callous-unemotional behaviors (CU) and parenting does not focus on directions of effect, and work that does so has not been genetically informed. The present study is the first to examine potential reciprocal effects between parenting and CU in a community sample of early childhood. Use of a twin sample also allows us to distinguish child-based genetic effects from environmentally driven effects, which is necessary before translating this research to interventions. METHOD: The present study used biometric cross-lagged models to investigate the relation between CU and parenting in twins at 2 and 3 years of age (monozygotic = 145, dizygotic = 169 twin pairs). CU was assessed using the parent-reported Child Behavior Checklist 1.5-5. Scores were residualized to control for conduct problems assessed on the Revised Rutter Parent Scale for Preschool Children. Parents' reports of negative and positive parenting were obtained using parent ratings of discipline and parent feelings from the Parent Feelings Questionnaire. RESULTS: CU and negative parenting were significantly correlated at both ages. Cross-lagged analyses revealed a unidirectional effect with CU at age 2 years predicting negative parenting at age 3 years. These child-driven effects were primarily genetically mediated, although there were modest nonshared environmental contributions. CU and positive parenting were not consistently correlated, and further biometric analyses were not performed. CONCLUSION: Children's genetically influenced CU behaviors can have an impact on the parenting that they experience. Child-driven CU effects, although less examined in the literature, are important and should receive more attention in future work.

Do Genetic Factors Explain the Links Between Callous-Unemotional, Attention Hyperactivity and Oppositional Defiant Problems in Toddlers?

Research demonstrates that callous-unemotional (CU) behaviors, Attention Deficit Hyperactivity Disorder (ADHD) and Oppositional Defiant Problems (ODD) are related, but little is known about the sources of covariation among the three externalizing behaviors. The present study looked at genetic and environmental links between all three behavioral domains in twins at ages 2 and 3 years (MZ = 145, DZ = 169), a time when CU behaviors are beginning to emerge. CU, ADHD, and ODD behaviors as assessed using the Child Behavior Checklist 1.5-5 (Achenbach and Rescorla 2000) were strongly interrelated at both ages. Genetic factors primarily explained the covariation among the three behavioral domains via a common externalizing factor; however, there were also genetic factors unique to each behavior. Furthermore, the majority of nonshared environmental influences on each externalizing behavior were behavior-specific. The heritable externalizing factor was highly stable across age, largely due to genetic factors shared across ages 2 and 3 years. Despite their extensive phenotypic and genetic overlap, CU, ADHD, and ODD behaviors have unique genetic and nonshared environmental influences as early as toddlerhood. This supports phenotypic research showing that the three are related but distinct constructs in very young children.