Health related social needs (HRSNs), including unstable housing, food insecurity, transportation barriers, and financial strain, play a key role in child health that accounts for over 50% of variable health outcomes. Yet these needs remain underdiagnosed in standard pediatric care. A new pilot randomized controlled trial is assessing an artificial intelligence (AI)-driven chatbot, DAPHNE, that is designed to improve identification of HRSNs and relate families to essential community resources.
The health outcomes in children with unmet social needs were poorer; 18% of them demonstrated more emergency department visits (incidence rate ratio [IRR]: 1.18, 95% confidence interval [CI]: 1.12-1.23) and 36% more hospitalizations (IRR: 1.36, 95% CI: 1.26-1.47) as compared with children without such risks. The highest burden is in children younger than 2 years, with the early exposure to HRSN linked to developmental and behavioral problems at six months of age (adjusted odds ratios [AOR]: 1.64-1.86, 2.16 for children with multiple needs).
Despite the significant burden, testing remains restricted. Fewer than 16% of physician practices and 25% of hospitals usually conduct routine screening, largely because of limited awareness of referral pathways, staffing constraints, and workflow inefficiencies. Caregivers may also hesitate to disclose needs because of uncertainty about accessing services, language barriers, and stigma. These challenges are especially pronounced among Medicaid-insured populations.
The DAPHNE chatbot aims to overcome these barriers by aiding patient-initiated screening beyond traditional clinical settings. The chatbot uses natural language processing to engage caregivers in private conversations, deliver personalized referrals to community resources, and detect social needs in real time. This method can ease provider workload while improving the accuracy and speed of need identification.
This pilot study will recruit 100 caregivers of children as participants (under the guidance of the Phase IIb ORBIT behavioral framework) and will prioritize pediatric patients (children aged ≤2 years) within a large primary care network. The participants will be randomized in a 1:1 ratio to the standard care and chatbot and standard care (50 participants each). Moreover, 10 healthcare providers will be involved in surveys and interviews to assess the integration of workflow and its viability.
The chatbot uses advanced language models with >99% intent accuracy, runs on secure HIPAA-compliant systems, and addresses four key social needs with escalation for high-risk cases. Primary feasibility outcomes include ≥70% recruitment/retention, ≥80% feasibility/acceptability, a System Usability Scale (SUS) ≥68, and performance standards of <3-second response time and F1 ≥0.7.
Secondary outcomes include satisfaction with resources, self-efficacy, caregiver-reported stress, and quality of life along with engagement metrics like login and usage. Exploratory analyses will assess costs and missed appointments and assess healthcare use, including emergency department visits. The study is currently recruiting, with completion expected by Year 2 supported by a 6-month follow-up and results by Year 3. Limitations include a small sample (n=100), English-only participants, a single-site design, and a short follow-up.
If effective, DAPHNE could provide a scalable approach to integrate the social needs screening into pediatric care, minimizing stigma, simplifying the referral process, and improving early identification and informing large trials.
Reference: Sezgin E, Clarkson E, Logan F, et al. An AI-based chatbot to support health-related social needs among pediatric primary care population: protocol for a pilot randomized controlled trial. PLoS One. 2026;21(4):e0337868. doi:10.1371/journal.pone.0337868
