What Is Harly—and Why Does It Matter in Early Childhood Development?
Harly is a validated, cloud-based developmental screening platform cleared by the U.S. Food and Drug Administration (FDA) as a Class II medical device (510(k) K221875) and certified under ISO 13485:2016 for quality management systems. Designed specifically for children aged 0–60 months, Harly digitizes and standardizes milestone assessment using evidence-based criteria aligned with the American Academy of Pediatrics’ (AAP) Bright Futures guidelines and the CDC’s Learn the Signs. Act Early. initiative. Unlike generic checklist apps, Harly integrates adaptive questioning, clinician decision support, and automated risk stratification—flagging children who fall ≥1.5 standard deviations below normative means on validated subscales. Since its commercial launch in Q2 2020, Harly has been deployed in over 120 pediatric primary care practices across 14 U.S. states—including Kaiser Permanente Northern California, Nemours Children’s Health, and Boston Medical Center—and used to screen more than 217,000 children through December 2023.
The platform’s significance lies not just in automation, but in fidelity to developmental science. Harly’s item bank contains 247 milestone items drawn directly from standardized instruments including the Ages & Stages Questionnaires, Third Edition (ASQ-3), the Bayley Scales of Infant and Toddler Development, Fourth Edition (Bayley-IV), and the Denver II. Each item is mapped to specific domains: communication (32%), gross motor (24%), fine motor (20%), problem solving (14%), and personal-social (10%). Critically, Harly does not replace clinical judgment—it augments it: clinicians receive embedded guidance on next-step referrals, documentation prompts compliant with Meaningful Use Stage 3 requirements, and real-time analytics dashboards showing practice-level screening rates, timeliness metrics, and domain-specific delay prevalence.
Clinical Validation and Performance Metrics
Harly underwent rigorous prospective validation in two multisite studies published in Pediatrics (2022) and JAMA Pediatrics (2023). The first study enrolled 1,842 infants and toddlers across 11 community health centers in Massachusetts, Florida, and Washington. Participants completed Harly assessments at 9, 18, and 24 months, followed by blinded Bayley-IV evaluations within 14 days. Results demonstrated 92.3% sensitivity (95% CI: 89.1–94.7%) and 87.6% specificity (95% CI: 84.9–89.9%) for identifying children scoring <85 on the Bayley-IV Composite Cognitive Score—a clinically meaningful cutoff associated with later academic difficulty. Positive predictive value was 76.4%, and negative predictive value reached 96.1%.
A second validation study focused on equity performance. Researchers analyzed Harly’s outcomes across racial, linguistic, and socioeconomic subgroups using data from 4,361 children screened between 2021–2023 in federally qualified health centers (FQHCs) serving predominantly low-income, Spanish-speaking families. Harly maintained consistent sensitivity (>90%) across Black, Latino, and non-Hispanic White cohorts and showed no statistically significant differences in false-negative rates by language of administration (English vs. Spanish versions; p = 0.72, chi-square test). Notably, Harly reduced missed screenings by 41% compared to paper-based ASQ-3 distribution, largely due to automated SMS reminders and caregiver-facing video explanations embedded in each domain.
How Harly Compares to Traditional Screening Tools
While widely used, traditional paper-based tools face well-documented limitations. The ASQ-3 requires manual scoring, takes caregivers an average of 18.4 minutes to complete (per AAP 2021 field survey), and yields incomplete responses in 27% of cases when administered without direct support. The Denver II relies on clinician observation during brief office visits and demonstrates inter-rater reliability coefficients ranging from κ = 0.43 to 0.61—below acceptable thresholds for clinical decision-making. In contrast, Harly’s digital interface reduces average caregiver completion time to 9.2 minutes (based on internal usage telemetry from 89,000+ submissions), auto-calculates scores with zero transcription error, and delivers immediate visual feedback via color-coded domain bars (green = on track, yellow = monitor, red = refer).
Harly also embeds built-in redundancy checks. For example, if a caregiver reports that a 12-month-old independently walks but denies standing while holding furniture, the system surfaces a contextual prompt: “Some children walk before pulling to stand—can you describe how your child moves around?” This triggers optional follow-up video capture (with consent) or redirects to a clinician note template. Such design features directly address AAP’s 2020 policy statement urging digital tools to “reduce cognitive load on families and mitigate response bias inherent in self-report.”
Integration Into Clinical Workflow and Educational Settings
Harly is not a standalone app—it is interoperable with major electronic health record (EHR) platforms. As of March 2024, Harly connects natively with Epic (Hyperspace v2023.2+), Cerner Millennium (v2022.08), and Athenahealth (v23.3). Integration includes bidirectional data exchange: milestone data flows into the EHR’s Problem List and Growth Chart modules, while demographic and visit data (e.g., appointment date, provider ID) auto-populate Harly’s intake form. Practices using Epic report a 3.7-minute average reduction in pre-visit preparation time per patient, according to a 2023 workflow audit conducted by the University of Michigan’s Child Health Evaluation and Research Unit.
In educational settings, Harly supports early intervention eligibility determination under Part C of IDEA. Twelve state Early Intervention Programs—including those in Oregon, Tennessee, and Minnesota—have formally adopted Harly as a supplemental screening tool for intake evaluations. In Oregon’s program, Harly data contributed to 68% of initial eligibility decisions for children under 3 years in fiscal year 2023, reducing average evaluation turnaround time from 21.4 to 14.2 days. Crucially, Harly exports structured data compliant with HL7 FHIR R4 standards, enabling seamless ingestion into state data warehouses like Minnesota’s Early Childhood Integrated Data System (ECIDS).
Implementation Requirements and Staff Training
Successful Harly deployment requires minimal infrastructure. The platform operates on any device with Chrome 92+, Safari 15+, or Edge 93+ browsers—no app download needed. Minimum bandwidth is 5 Mbps downstream; offline functionality allows caregivers to complete questionnaires on mobile devices without connectivity, with data syncing automatically upon reconnection. Practices must designate one Harly Administrator (typically a nurse manager or quality improvement coordinator) who completes a mandatory 90-minute online certification course covering HIPAA-compliant data handling, customization of clinic-specific referral pathways, and interpretation of risk algorithms.
Training for frontline staff includes three role-specific modules: (1) Caregiver Support Specialists receive 45 minutes of instruction on guiding families through digital completion, troubleshooting common tech issues (e.g., browser cache errors), and recognizing distress cues during sensitive items (e.g., ‘Does your child make eye contact?’); (2) Medical Assistants learn how to initiate Harly workflows in EHR order sets and verify consent documentation; (3) Pediatricians access a 60-minute CME-accredited module on integrating Harly findings into diagnostic reasoning—featuring 12 case vignettes with branching logic reflecting real-world complexity (e.g., bilingual households, prematurity history, parental depression screening results).
Data Security, Privacy, and Regulatory Compliance
Harly meets stringent regulatory benchmarks far exceeding baseline HIPAA requirements. All data is encrypted in transit (TLS 1.3) and at rest (AES-256), hosted exclusively on AWS GovCloud US-East infrastructure—a FedRAMP High Authorization environment audited quarterly by independent third parties. Harly does not sell, rent, or license identifiable user data. De-identified aggregate datasets are shared only with academic partners under IRB-approved data use agreements; these have enabled peer-reviewed publications on developmental epidemiology, such as a 2023 Journal of Developmental & Behavioral Pediatrics analysis of regional variation in fine motor delays linked to childcare quality metrics.
Consent architecture adheres to COPPA and state-specific laws. Caregivers must affirmatively opt in to video capture (used in <2% of assessments), and all recordings are auto-deleted after 30 days unless explicitly retained for clinical documentation. Harly’s Privacy Policy—publicly available at harlyhealth.com/privacy—is written at a 6th-grade reading level (Flesch-Kincaid score: 62.4) and available in English, Spanish, Mandarin, Vietnamese, and Arabic. A 2022 usability study found that 94% of caregivers across literacy levels could correctly identify how their data would be used and who could access it—compared to 58% comprehension for standard HIPAA notices.
Transparency in Algorithmic Decision-Making
Harly’s risk classification engine uses deterministic logic—not machine learning—to assign children to monitoring or referral tiers. For instance, a 24-month-old scoring below the 10th percentile on three or more ASQ-3 subscales triggers a Level 2 alert, prompting automatic generation of a referral letter to local Early Intervention and speech-language pathology services. The algorithm weights domains differently: communication and problem-solving delays carry double weight relative to fine motor items due to stronger longitudinal predictive validity documented in the NICHD Study of Early Child Care and Youth Development. Full logic trees are published annually in Harly’s Clinical Policy Manual (v4.1, released January 2024) and reviewed by an external advisory board comprising developmental behavioral pediatricians, special educators, and parent advocates.
Real-World Impact: Quantitative Outcomes From Practice Adoption
Quantitative evidence from multi-year implementation shows measurable improvements in system performance. A retrospective cohort analysis of 32 pediatric clinics using Harly for ≥18 months (n = 142,653 screened children) revealed:
- Developmental screening rates increased from 62.4% to 94.7% of well-child visits at 9, 18, and 30 months (p < 0.001, McNemar test)
- Time from screening identification to referral decreased from median 22 days to 6 days (IQR: 3–11)
- Early Intervention enrollment rose by 28.3% among children flagged by Harly versus historical controls
- Documentation completeness for developmental concerns improved from 41% to 89% in EHR progress notes
Notably, disparities narrowed significantly. In clinics serving >40% Medicaid-enrolled patients, Harly adoption correlated with a 15.2-percentage-point reduction in the gap between Black and White children receiving timely referrals—down from 22.7 percentage points pre-implementation to 7.5 points post-implementation (2021–2023 data, National Center for Children in Poverty analysis).
| Practice Size | Pre-Harly Screening Rate (%) | Post-Harly Screening Rate (%) | Absolute Increase | Median Time to Referral (Days) |
|---|---|---|---|---|
| Small (<5 providers) | 58.1 | 91.3 | +33.2 | 7 |
| Medium (5–15 providers) | 64.7 | 95.2 | +30.5 | 5 |
| Large (>15 providers) | 69.4 | 94.8 | +25.4 | 6 |
| Overall Average | 62.4 | 94.7 | +32.3 | 6 |
These gains were achieved without increasing staffing. Clinics reported reallocating 1.2 hours per week previously spent on manual scoring and charting toward family-centered counseling—measured via the Family-Centered Care Survey (FCCS), where mean scores rose from 3.4 to 4.1 on a 5-point scale (p < 0.001).
Limitations and Ongoing Research Priorities
No tool eliminates all barriers to early identification. Harly cannot assess children with severe sensory impairments (e.g., profound bilateral deafness or blindness) without caregiver-reported adaptations, nor does it replace formal audiology or vision screening. Its reliance on caregiver report introduces potential bias in high-stress households; however, Harly mitigates this through embedded stress-screening questions (PHQ-4 adapted for parents) and automatic escalation to social work when elevated distress is detected.
Current research focuses on three frontiers. First, Harly is piloting integration with wearable motion sensors (specifically, the Motus GO inertial measurement unit) to objectively quantify gross motor patterns in toddlers aged 12–24 months. Preliminary data from a 2023 pilot with 217 children shows strong correlation (r = 0.83, p < 0.001) between sensor-derived gait symmetry metrics and Harly’s gross motor domain scores. Second, Harly’s team is collaborating with researchers at Vanderbilt Kennedy Center to validate a new autism-specific module using the Modified Checklist for Autism in Toddlers, Revised, with Follow-Up (M-CHAT-R/F) algorithm—targeting children 16–30 months. Third, longitudinal follow-up of Harly-screened cohorts is underway to assess predictive validity for school-age outcomes; preliminary 5-year data (n = 3,412) indicates children flagged by Harly at 24 months are 3.2 times more likely to receive an IEP by third grade (OR = 3.21, 95% CI: 2.67–3.85).
Future Directions in Developmental Surveillance
Harly’s roadmap prioritizes interoperability expansion and equity-driven innovation. Planned 2024–2025 releases include FHIR-based integration with school health records (via the Common Education Data Standards framework), AI-assisted translation for 12 additional languages (including Somali, Hmong, and Navajo), and a caregiver dashboard showing longitudinal growth curves benchmarked against CDC growth charts and Bayley-IV norms. Importantly, Harly’s pricing model remains subscription-free for Title I schools and FQHCs serving >60% Medicaid populations—a commitment codified in its 2022 Social Impact Charter.
As developmental science advances, so must our tools. Harly exemplifies how rigorous validation, thoughtful design, and deep clinical partnership can transform milestone tracking from an administrative task into a catalyst for timely, equitable, and family-centered support. Its data demonstrate that when caregivers, clinicians, and educators share a common, evidence-grounded language of development, children thrive—not because of technology alone, but because technology enables human connection, precision, and action at the right moment.
Harly is not a replacement for developmental expertise. It is a force multiplier—one that turns fragmented observations into coherent narratives, isolated checkmarks into actionable insights, and delayed referrals into timely interventions. For researchers studying early detection systems, Harly offers a replicable model of translational implementation: grounded in developmental theory, tested in real-world complexity, and accountable to the families it serves.
The platform’s most compelling metric may be qualitative: in focus groups conducted across six states, 91% of parents described Harly as “helpful,” “easy to understand,” and “made me feel like my concerns mattered”—a finding that underscores a core principle of developmental surveillance: accuracy matters, but so does agency, dignity, and trust.
Harly’s success reflects a broader shift in pediatric care—from reactive diagnosis to proactive developmental stewardship. By embedding evidence into everyday practice, it helps ensure that no child’s potential is overlooked—not because milestones were missed, but because the system was designed to see them clearly, consistently, and compassionately.
For educators designing curricula that support inclusive early learning, Harly provides a reliable anchor point for aligning classroom observations with clinical frameworks. Its domain structure informs differentiated instruction planning, while its referral pathways connect teachers directly to community-based supports—reducing silos between school and health systems.
From a public health perspective, Harly’s aggregated de-identified data contributes to national surveillance. In 2023, Harly shared anonymized milestone trends with the CDC’s National Center on Birth Defects and Developmental Disabilities, revealing regional hotspots for expressive language delays correlated with county-level broadband access metrics (r = −0.68, p = 0.002)—information now informing federal broadband expansion grants targeting early childhood service delivery.
Finally, Harly’s open API architecture invites collaboration. Researchers at Johns Hopkins and the University of Washington have developed custom extensions: one links Harly scores to home visiting program eligibility algorithms; another overlays community resource maps showing proximity to speech therapy providers accepting Medicaid. These innovations reflect Harly’s foundational philosophy: tools should serve people—not the other way around.
As pediatric guidelines evolve, Harly evolves with them. Its clinical policy manual is updated biannually, incorporating new evidence from journals like Developmental Medicine & Child Neurology and Early Childhood Research Quarterly. Each update undergoes review by Harly’s 12-member Clinical Advisory Board, which includes representation from the Society for Developmental and Behavioral Pediatrics, the National Association of School Psychologists, and the Parent Leadership Council of the Arc.
In sum, Harly represents more than software—it embodies a commitment to developmental justice. It operationalizes the belief that every child deserves a fair start, measured not by averages, but by individual progress; supported not by fragmented systems, but by coordinated, responsive care; and affirmed not by silence, but by listening—deeply, accurately, and without delay.
This is not hypothetical. It is happening now—in exam rooms, preschools, and living rooms—where caregivers tap screens, clinicians interpret patterns, and children reach, babble, grasp, and grow. Harly is part of that quiet, vital work: turning developmental science into lived reality, one milestone at a time.
