Times are displayed in (UTC-06:00) Central Time (US & Canada) Change
Symposium 18: The Industry Connections Committee and Digitial Health SIG present: Real-World Evidence from Industry: Integrating Scientific Perspectives to Evaluate and Scale Digital Mental Health Solutions
Real-World Evidence from Industry: Integrating Scientific Perspectives to Evaluate and Scale Digital Mental Health Solutions
Time: 09:00 AM - 09:50 AMTopics: Digital Health, Mental Health
Significant gaps in mental healthcare access and delivery continue to challenge efforts to improve population mental health worldwide. Digital mental health platforms developed in the private sector are innovating to meet this challenge, offering scalable models to deliver evidence-based care. But to optimize these services, ensure quality and safety, and evaluate their true impact, we need data not just from clinical trials, but from the conditions in which care is actually delivered.
Real-world evidence (RWE) derived from data collected during routine care offers insights into how mental health interventions function across diverse populations and real-life contexts. Unlike controlled clinical trials, RWE reveals the complexity, variability, and practical realities of routine care, facilitating a deeper understanding of safety, effectiveness, and implementation at scale. For the field of behavioral medicine, RWE represents a tool to accelerate the integration of scientific knowledge into healthcare systems, ensuring that research informs real-world practice and decision-making at scale.
This symposium brings together behavioral medicine scientists working across four industry organizations implementing and evaluating scalable digital mental health solutions using RWE. Presentations will share lessons learned and findings from real-world studies and highlight the value of RWE for informing product development and optimization, clinical validation, and external evidence generation for healthcare stakeholders. The studies presented use a range of real-world data (RWD) sources including app usage, survey data, biometric monitoring, electronic health records, and insurance claims. Study designs that will be discussed include prospective cohort studies, psychometric validation, pragmatic hybrid approaches, and retrospective analyses.
Dr. Sara Sagui Henson from Modern Health will open with findings from longitudinal cohort studies of an employer-sponsored digital mental health platform, examining care patterns, effectiveness, and equity. Dr. Lydia Roos from Walking on Earth will present on the development and validation of a multidimensional resilience index embedded into a digital stress reduction platform, integrating longitudinal psychometric and biometric data at scale. Dr. Caitlin Stamatis from Slingshot AI will share real-world clinical outcomes and engagement data from a purpose-built mental health AI. Dr. Eduardo Bunge from ParenteAI will present RWD on a novel hybrid behavioral parent training program for children with disruptive disorders, combining human therapists and an AI agent. Together, these talks will illustrate how integrating RWE methodologies accelerates healthcare innovation, fosters new clinical discoveries, informs product enhancement, and ultimately improves access to evidence-based mental healthcare.
Dr. Cynthia Castro Sweet, a recognized SBM leader with deep experience in both academia and industry, will serve as discussant and facilitate audience Q&A. Attendees will gain practical insight into how RWE is generated, used, and translated within industry settings, and why it is a critical scientific approach for advancing discovery, accelerating mental healthcare innovation, and driving better outcomes for diverse patient populations.
Additional details on each presentation are provided below:
Presentation 1: From Data to Impact: Real-World Evidence for Innovation and Scale in Employer-Sponsored Digital Mental Health
Presenter: Sara Sagui Henson, PhD
Co-Authors: Komal Kumar, MS; Kristen Van Swearingen, MA; Valerie Hoffman, PhD; Dori Steinberg, PhD; Jessica Watrous, PhD
Background: The global mental health burden requires innovative care models that can scale. Employer-sponsored digital mental health platforms offer scalable solutions through stratified, blended care models. Yet, real-world implementation and impact often differ from controlled settings. Real-world evidence (RWE) collected while employees are receiving care can offer essential insights into use, effectiveness, and value of these platforms. This presentation will share implementation lessons and scientific discoveries from conducting RWE research in a commercial industry digital mental health setting.
Methods: Since 2021, Modern Health has conducted three prospective, longitudinal cohort studies evaluating its technology-enabled platform, which offers personalized mental health services, including self-guided digital content, group psychoeducation, coaching, therapy, and psychiatry. Employees who enrolled through workplace benefits and opted to participate in these studies were followed for 6 to 12 months. Studies integrated de-identified platform data (e.g., engagement, satisfaction, therapeutic alliance) with survey measures of mental health symptoms, well-being, transdiagnostic factors, workplace functioning, and social determinants of health. Studies were approved by a commercial IRB and designed collaboratively across clinical research, product, and commercial teams.
Results: Across studies, over 1,900 employees enrolled with follow-up rates near 65%. We will describe implementation insights including how the timing of employee benefit launches affects recruitment, how quota sampling improves representation, and how flexible but rigorous measurement strategies balance data quality with user burden. Insights have driven product optimization (e.g., assessment redesign), supported clinical discovery (e.g., 72% of moderate-risk coaching users improved), generated population health insights (e.g., 70% of employees follow their care recommendations; People of Color and white employees reported equitable outcomes), and informed key commercial activities (e.g., research webinars attended by 250+ prospective clients). Study results have generated five peer-reviewed publications and more than a dozen scientific conference presentations. Results are used across commercial, clinical, and product teams to inform decision-making, accelerate adoption through commercialization, and ensure integration of scientific evidence into real-world business strategies.
Conclusion: RWE embedded in digital mental health care yields critical insights about real-world effectiveness, equity, and innovation that are not observable in controlled trials. This presentation will highlight the scientific and commercial value of conducting rigorous research in real-world industry settings, which bridges efficacy and effectiveness, informs clinical validation, and accelerates scalable access to evidence-based mental health care.
Presentation 2: Balancing Scientific Validity and User Value for Longitudinal Research: Embedding a Multidimensional Stress Resilience Measure in a Digital Health Platform
Presenter: Lydia Roos, PhD
Co-Authors: Destiny Gilliland, Kelsey Julian, MA, Reeva Misra, MA
Background: Integrating behavioral medicine evidence into digital platforms is essential to accelerate discovery in stress and resilience science. Traditional measures are often lengthy, static, or siloed, limiting scalability. Walking on Earth (WONE) is a digital wellbeing platform distributed via employer and payer partnerships that combines stress recovery interventions with resilience training through personalized programs. WONE developed and validated the WONE Index, a multidimensional measure of stress burden and resilience resources. Unlike survey add-ons, the Index is embedded as a core app feature, powering insights, recommendations, and tracking. This dual role enhances user value while enabling longitudinal evidence collection.
Methods: Validation used two studies (Study 1: N=1,005 US/UK employees; Study 2: N=306 US employees). Analyses included exploratory and confirmatory factor analyses, internal consistency (CR ≥.74), test–retest reliability (ICCs=.77–.90), and convergent/discriminant validity against gold-standard measures (PSS-4, CD-RISC, BRS, PROMIS-SF-8). Incremental validity was tested via regressions predicting depression, anxiety, and well-being. The Index is delivered monthly via self-report and paired with daily biometric data (HRV, sleep, activity) through device-agnostic integrations.
Results: A 10-factor structure was supported, spanning stress (work stress, personal stress, burnout) and resilience resources (emotion regulation, social connection, purpose, sleep, physical activity, diet, perseverative thinking). The Index showed strong reliability and validity, with incremental predictive utility beyond existing measures. By embedding the Index into onboarding and via monthly check-ins, yielding personalized insights for users, it is designed to function as a core app experience that encourages high completion rates and ongoing engagement while enabling continuous multimodal data collection. Importantly, pairing the Index with daily biometrics (HRV, sleep, activity) enables multimodal, longitudinal evidence generation that static survey tools cannot provide.
Conclusion: This work shows how validated measurement can serve as core infrastructure in digital platforms, bridging academic rigor with real-world scalability. By integrating psychometrics and biometric streams, the WONE Index captures how stress and resilience unfold in daily life, linking psychological and physiological processes in real-world contexts. More broadly, the development process offers a model for balancing psychometric standards with user-centered design, advancing sustainable, evidence-generating health technologies.
Presentation 3: Clinical Improvement and Engagement Phenotypes with a Purpose-Built Mental Health AI: Evidence From Two Real-World Studies
Presenter: Caitlin A. Stamatis, PhD
Co-Authors: Ting Su, PhD, Matteo Malgaroli, PhD, Thomas D. Hull, PhD
Background: Generative AI mental health tools may expand access to scalable psychological support, yet real-world evidence on clinical outcomes, safety, and engagement-response patterns remains limited. This work synthesizes findings from two real-world evidence (RWE) studies evaluating a purpose-built mental health AI. RWE is especially important in digital mental health, where outcomes depend on heterogeneous, self-directed use and engagement is episodic and context-dependent.
Methods: Study 1 was a prospective naturalistic cohort of U.S. adults with baseline PHQ-9 or GAD-7 ≥10 (N=305; 82% female; mean age 41.7) using a mental health AI tool between May-Sept 2025, completing outcome measures biweekly through 6 weeks and follow-up at 10 weeks. Study 2 used platform data to derive engagement clusters in a large-scale naturalistic cohort (~100k users). Cluster definitions were applied to a clinical analytic sample (n=224) who completed baseline and week 3 in-app PHQ-9/GAD-7 and consented to conversation data sharing; baseline working alliance was examined across clusters, and symptom change was compared by cluster.
Results: In Study 1, users demonstrated significant reductions in depression and anxiety at 6 weeks (PHQ-9 d=.80; GAD-7 d=.77), sustained at 10-week follow-up (PHQ-9 d=.93; GAD-7 d=.79). Social and behavioral functioning improved, including loneliness (d=.67) and perceived social support (d=.55). Engagement was high (mean 9.02 hours), and safety guardrails flagged 1.02% of sessions for potential crisis content with clinical review/escalation. In Study 2, k-means clustering identified five usage phenotypes (early dropout, high sessions/long-term, long sessions/high engagement, sparse/long-term, frequent/short-term). In the clinical sample (n=224), depression/anxiety change distributions varied by cluster, with higher-engagement users generally showing greater symptom decreases and improved social relationships. Working alliance was relatively consistent across clusters, but moderated outcomes. Notably, 38.6% of conversations happened between 9pm-5am, underscoring the value of “always available” support that may extend beyond traditional care hours.
Conclusions: Across two real-world studies, a mental health AI was associated with clinically meaningful reductions in depression and anxiety, concurrent improvements in social functioning, low observed safety-event flag rates, and heterogeneous engagement-outcome profiles. Lessons for RWE in digital health include: (1) engagement is multidimensional and should be modeled as phenotypes rather than simple dose counts, (2) linking passively collected behavioral metrics with repeated standardized outcomes enables scalable evaluation, and (3) safety should be operationalized as continuous risk monitoring with clinician escalation, not a one-time pre-launch validation.
Presentation 4: Innovating in Behavioral Parent Training: Combining Human Therapists and AI
Presenter: Eduardo Bunge, PhD
Co-Authors: Felipe Rivera Cepeda Mgr., Blanca Pineda, EdD. Karin Mostovoy, MS
Background: Approximately 20% of children experience a diagnosable mental health disorder each year, with the most common being disruptive behavioral disorders. Behavioral Parent Training (BPT) programs have been shown to be effective in the treatment of disruptive behaviors, but limited availability of trained providers, scheduling conflicts, and high costs reduce access to traditional BPT. Artificial
ParenteAI offers an AI platform to augment therapists' reach and outcomes, though its use in mental health is still in the early stages. Thus, the safety and efficacy of its implementation as an adjunct to individual and group BPT should be assessed in research settings and the real world
Methods: Study 1. Assess treatment fidelity and safety. Study 2. Assess the efficacy of ParenteAI as an adjunct to individual therapy. Study 3: A retrospective cohort analysis of therapy attitudes inventory, therapeutic alliance, and user engagement. Nine BPT modules were delivered during 8 live sessions with a mental health professional, and an AI conversational agent (PAT) provided support and complemented the learning between sessions. Participants were 70 caregivers of children ages 4-14 years seeking services for disruptive behaviors.
Intervention: Study 1 (Fidelity/Safety): Compare delivery of core BPT components to established standards; monitor adverse events. Study 2 (Efficacy): Evaluate changes in children’s disruptive behaviors and caregivers’ depression, anxiety, and stress over ~2 months with ParenteAI as an adjunct. Study 3 (Real-world data) Retrospective Cohort of 70 parents attending BPT group sessions with a mental health professional, and PAT provided between-session support and skills practice. Parents' perspectives and usage metrics were examined.
Results: Study 1): treatment fidelity meets fidelity standards comparable to those of highly trained human therapists. Study 2: There were significant pre- to post-intervention decreases in child externalizing (p < 0.001, d = 1.00) and internalizing (p = 0.019, d = 0.63) symptoms, and significant decreases in caregiver depression (p = 0.005, r = 0.68), anxiety (p = 0.002, r = 0.79), and stress (p = 0.008, d = 0.74). Study 3: Caregivers reported Net Promoter Score (NPS) = 71 (excellent by industry benchmarks), strong therapeutic alliance with PAT (M = 4.06), positive treatment attitudes (Therapy Attitude Inventory, M = 4.39), 89% module completion, and 450 messages exchanged on average.
Conclusion: Caregivers reported high satisfaction (NPS), strong alliance, and positive attitudes toward combining a human therapist with an AI agent. Together, trial-style evaluation and real-world evidence highlight the feasibility, usability, and scalable potential of ParenteAI/PAT to address access barriers in parenting programs while also revealing the strengths and implementation challenges that matter for translation into everyday practice.
This symposium is sponsored by the Industry Connections Committee and endorsed by the membership of the 2025-2026 SBM Digital Health Special Interest Group. The authors used OpenAI’s ChatGPT for language refinement and editorial support during abstract preparation.
Keywords: Technology, EffectivenessReal-world evidence (RWE) derived from data collected during routine care offers insights into how mental health interventions function across diverse populations and real-life contexts. Unlike controlled clinical trials, RWE reveals the complexity, variability, and practical realities of routine care, facilitating a deeper understanding of safety, effectiveness, and implementation at scale. For the field of behavioral medicine, RWE represents a tool to accelerate the integration of scientific knowledge into healthcare systems, ensuring that research informs real-world practice and decision-making at scale.
This symposium brings together behavioral medicine scientists working across four industry organizations implementing and evaluating scalable digital mental health solutions using RWE. Presentations will share lessons learned and findings from real-world studies and highlight the value of RWE for informing product development and optimization, clinical validation, and external evidence generation for healthcare stakeholders. The studies presented use a range of real-world data (RWD) sources including app usage, survey data, biometric monitoring, electronic health records, and insurance claims. Study designs that will be discussed include prospective cohort studies, psychometric validation, pragmatic hybrid approaches, and retrospective analyses.
Dr. Sara Sagui Henson from Modern Health will open with findings from longitudinal cohort studies of an employer-sponsored digital mental health platform, examining care patterns, effectiveness, and equity. Dr. Lydia Roos from Walking on Earth will present on the development and validation of a multidimensional resilience index embedded into a digital stress reduction platform, integrating longitudinal psychometric and biometric data at scale. Dr. Caitlin Stamatis from Slingshot AI will share real-world clinical outcomes and engagement data from a purpose-built mental health AI. Dr. Eduardo Bunge from ParenteAI will present RWD on a novel hybrid behavioral parent training program for children with disruptive disorders, combining human therapists and an AI agent. Together, these talks will illustrate how integrating RWE methodologies accelerates healthcare innovation, fosters new clinical discoveries, informs product enhancement, and ultimately improves access to evidence-based mental healthcare.
Dr. Cynthia Castro Sweet, a recognized SBM leader with deep experience in both academia and industry, will serve as discussant and facilitate audience Q&A. Attendees will gain practical insight into how RWE is generated, used, and translated within industry settings, and why it is a critical scientific approach for advancing discovery, accelerating mental healthcare innovation, and driving better outcomes for diverse patient populations.
Additional details on each presentation are provided below:
Presentation 1: From Data to Impact: Real-World Evidence for Innovation and Scale in Employer-Sponsored Digital Mental Health
Presenter: Sara Sagui Henson, PhD
Co-Authors: Komal Kumar, MS; Kristen Van Swearingen, MA; Valerie Hoffman, PhD; Dori Steinberg, PhD; Jessica Watrous, PhD
Background: The global mental health burden requires innovative care models that can scale. Employer-sponsored digital mental health platforms offer scalable solutions through stratified, blended care models. Yet, real-world implementation and impact often differ from controlled settings. Real-world evidence (RWE) collected while employees are receiving care can offer essential insights into use, effectiveness, and value of these platforms. This presentation will share implementation lessons and scientific discoveries from conducting RWE research in a commercial industry digital mental health setting.
Methods: Since 2021, Modern Health has conducted three prospective, longitudinal cohort studies evaluating its technology-enabled platform, which offers personalized mental health services, including self-guided digital content, group psychoeducation, coaching, therapy, and psychiatry. Employees who enrolled through workplace benefits and opted to participate in these studies were followed for 6 to 12 months. Studies integrated de-identified platform data (e.g., engagement, satisfaction, therapeutic alliance) with survey measures of mental health symptoms, well-being, transdiagnostic factors, workplace functioning, and social determinants of health. Studies were approved by a commercial IRB and designed collaboratively across clinical research, product, and commercial teams.
Results: Across studies, over 1,900 employees enrolled with follow-up rates near 65%. We will describe implementation insights including how the timing of employee benefit launches affects recruitment, how quota sampling improves representation, and how flexible but rigorous measurement strategies balance data quality with user burden. Insights have driven product optimization (e.g., assessment redesign), supported clinical discovery (e.g., 72% of moderate-risk coaching users improved), generated population health insights (e.g., 70% of employees follow their care recommendations; People of Color and white employees reported equitable outcomes), and informed key commercial activities (e.g., research webinars attended by 250+ prospective clients). Study results have generated five peer-reviewed publications and more than a dozen scientific conference presentations. Results are used across commercial, clinical, and product teams to inform decision-making, accelerate adoption through commercialization, and ensure integration of scientific evidence into real-world business strategies.
Conclusion: RWE embedded in digital mental health care yields critical insights about real-world effectiveness, equity, and innovation that are not observable in controlled trials. This presentation will highlight the scientific and commercial value of conducting rigorous research in real-world industry settings, which bridges efficacy and effectiveness, informs clinical validation, and accelerates scalable access to evidence-based mental health care.
Presentation 2: Balancing Scientific Validity and User Value for Longitudinal Research: Embedding a Multidimensional Stress Resilience Measure in a Digital Health Platform
Presenter: Lydia Roos, PhD
Co-Authors: Destiny Gilliland, Kelsey Julian, MA, Reeva Misra, MA
Background: Integrating behavioral medicine evidence into digital platforms is essential to accelerate discovery in stress and resilience science. Traditional measures are often lengthy, static, or siloed, limiting scalability. Walking on Earth (WONE) is a digital wellbeing platform distributed via employer and payer partnerships that combines stress recovery interventions with resilience training through personalized programs. WONE developed and validated the WONE Index, a multidimensional measure of stress burden and resilience resources. Unlike survey add-ons, the Index is embedded as a core app feature, powering insights, recommendations, and tracking. This dual role enhances user value while enabling longitudinal evidence collection.
Methods: Validation used two studies (Study 1: N=1,005 US/UK employees; Study 2: N=306 US employees). Analyses included exploratory and confirmatory factor analyses, internal consistency (CR ≥.74), test–retest reliability (ICCs=.77–.90), and convergent/discriminant validity against gold-standard measures (PSS-4, CD-RISC, BRS, PROMIS-SF-8). Incremental validity was tested via regressions predicting depression, anxiety, and well-being. The Index is delivered monthly via self-report and paired with daily biometric data (HRV, sleep, activity) through device-agnostic integrations.
Results: A 10-factor structure was supported, spanning stress (work stress, personal stress, burnout) and resilience resources (emotion regulation, social connection, purpose, sleep, physical activity, diet, perseverative thinking). The Index showed strong reliability and validity, with incremental predictive utility beyond existing measures. By embedding the Index into onboarding and via monthly check-ins, yielding personalized insights for users, it is designed to function as a core app experience that encourages high completion rates and ongoing engagement while enabling continuous multimodal data collection. Importantly, pairing the Index with daily biometrics (HRV, sleep, activity) enables multimodal, longitudinal evidence generation that static survey tools cannot provide.
Conclusion: This work shows how validated measurement can serve as core infrastructure in digital platforms, bridging academic rigor with real-world scalability. By integrating psychometrics and biometric streams, the WONE Index captures how stress and resilience unfold in daily life, linking psychological and physiological processes in real-world contexts. More broadly, the development process offers a model for balancing psychometric standards with user-centered design, advancing sustainable, evidence-generating health technologies.
Presentation 3: Clinical Improvement and Engagement Phenotypes with a Purpose-Built Mental Health AI: Evidence From Two Real-World Studies
Presenter: Caitlin A. Stamatis, PhD
Co-Authors: Ting Su, PhD, Matteo Malgaroli, PhD, Thomas D. Hull, PhD
Background: Generative AI mental health tools may expand access to scalable psychological support, yet real-world evidence on clinical outcomes, safety, and engagement-response patterns remains limited. This work synthesizes findings from two real-world evidence (RWE) studies evaluating a purpose-built mental health AI. RWE is especially important in digital mental health, where outcomes depend on heterogeneous, self-directed use and engagement is episodic and context-dependent.
Methods: Study 1 was a prospective naturalistic cohort of U.S. adults with baseline PHQ-9 or GAD-7 ≥10 (N=305; 82% female; mean age 41.7) using a mental health AI tool between May-Sept 2025, completing outcome measures biweekly through 6 weeks and follow-up at 10 weeks. Study 2 used platform data to derive engagement clusters in a large-scale naturalistic cohort (~100k users). Cluster definitions were applied to a clinical analytic sample (n=224) who completed baseline and week 3 in-app PHQ-9/GAD-7 and consented to conversation data sharing; baseline working alliance was examined across clusters, and symptom change was compared by cluster.
Results: In Study 1, users demonstrated significant reductions in depression and anxiety at 6 weeks (PHQ-9 d=.80; GAD-7 d=.77), sustained at 10-week follow-up (PHQ-9 d=.93; GAD-7 d=.79). Social and behavioral functioning improved, including loneliness (d=.67) and perceived social support (d=.55). Engagement was high (mean 9.02 hours), and safety guardrails flagged 1.02% of sessions for potential crisis content with clinical review/escalation. In Study 2, k-means clustering identified five usage phenotypes (early dropout, high sessions/long-term, long sessions/high engagement, sparse/long-term, frequent/short-term). In the clinical sample (n=224), depression/anxiety change distributions varied by cluster, with higher-engagement users generally showing greater symptom decreases and improved social relationships. Working alliance was relatively consistent across clusters, but moderated outcomes. Notably, 38.6% of conversations happened between 9pm-5am, underscoring the value of “always available” support that may extend beyond traditional care hours.
Conclusions: Across two real-world studies, a mental health AI was associated with clinically meaningful reductions in depression and anxiety, concurrent improvements in social functioning, low observed safety-event flag rates, and heterogeneous engagement-outcome profiles. Lessons for RWE in digital health include: (1) engagement is multidimensional and should be modeled as phenotypes rather than simple dose counts, (2) linking passively collected behavioral metrics with repeated standardized outcomes enables scalable evaluation, and (3) safety should be operationalized as continuous risk monitoring with clinician escalation, not a one-time pre-launch validation.
Presentation 4: Innovating in Behavioral Parent Training: Combining Human Therapists and AI
Presenter: Eduardo Bunge, PhD
Co-Authors: Felipe Rivera Cepeda Mgr., Blanca Pineda, EdD. Karin Mostovoy, MS
Background: Approximately 20% of children experience a diagnosable mental health disorder each year, with the most common being disruptive behavioral disorders. Behavioral Parent Training (BPT) programs have been shown to be effective in the treatment of disruptive behaviors, but limited availability of trained providers, scheduling conflicts, and high costs reduce access to traditional BPT. Artificial
ParenteAI offers an AI platform to augment therapists' reach and outcomes, though its use in mental health is still in the early stages. Thus, the safety and efficacy of its implementation as an adjunct to individual and group BPT should be assessed in research settings and the real world
Methods: Study 1. Assess treatment fidelity and safety. Study 2. Assess the efficacy of ParenteAI as an adjunct to individual therapy. Study 3: A retrospective cohort analysis of therapy attitudes inventory, therapeutic alliance, and user engagement. Nine BPT modules were delivered during 8 live sessions with a mental health professional, and an AI conversational agent (PAT) provided support and complemented the learning between sessions. Participants were 70 caregivers of children ages 4-14 years seeking services for disruptive behaviors.
Intervention: Study 1 (Fidelity/Safety): Compare delivery of core BPT components to established standards; monitor adverse events. Study 2 (Efficacy): Evaluate changes in children’s disruptive behaviors and caregivers’ depression, anxiety, and stress over ~2 months with ParenteAI as an adjunct. Study 3 (Real-world data) Retrospective Cohort of 70 parents attending BPT group sessions with a mental health professional, and PAT provided between-session support and skills practice. Parents' perspectives and usage metrics were examined.
Results: Study 1): treatment fidelity meets fidelity standards comparable to those of highly trained human therapists. Study 2: There were significant pre- to post-intervention decreases in child externalizing (p < 0.001, d = 1.00) and internalizing (p = 0.019, d = 0.63) symptoms, and significant decreases in caregiver depression (p = 0.005, r = 0.68), anxiety (p = 0.002, r = 0.79), and stress (p = 0.008, d = 0.74). Study 3: Caregivers reported Net Promoter Score (NPS) = 71 (excellent by industry benchmarks), strong therapeutic alliance with PAT (M = 4.06), positive treatment attitudes (Therapy Attitude Inventory, M = 4.39), 89% module completion, and 450 messages exchanged on average.
Conclusion: Caregivers reported high satisfaction (NPS), strong alliance, and positive attitudes toward combining a human therapist with an AI agent. Together, trial-style evaluation and real-world evidence highlight the feasibility, usability, and scalable potential of ParenteAI/PAT to address access barriers in parenting programs while also revealing the strengths and implementation challenges that matter for translation into everyday practice.
This symposium is sponsored by the Industry Connections Committee and endorsed by the membership of the 2025-2026 SBM Digital Health Special Interest Group. The authors used OpenAI’s ChatGPT for language refinement and editorial support during abstract preparation.
Authors:
Chair - Sara Sagui Henson, PhD, Modern Health
Presenter - Caitlin Stamatis, PhD, Otsuka Pharmaceutical Companies
Presenter - Lydia Roos, PhD, Walking on Earth
Presenter - Eduardo Bunge, PhD, ParenteAI
Discussant - Cynthia Castro Sweet, PhD, FSBM, Hinge Health
Symposium 18: The Industry Connections Committee and Digitial Health SIG present: Real-World Evidence from Industry: Integrating Scientific Perspectives to Evaluate and Scale Digital Mental Health Solutions
Description
Date: 4/24/2026
Start: 9:00 AM
End: 9:50 AM
Location: Salon A-3
