Abstract
Background Children's Healthcare of Atlanta (CHOA) provides care for over 2000 pediatric sickle cell disease (SCD) patients annually. Despite the availability of specialized care, families face persistent barriers related to transportation, insurance, financial aid, and health literacy, due to a lack of streamlined resource access. To bridge these gaps, we designed Siby, a conversational artificial intelligence (AI) agent powered by GPT-4o and built through a user-centered design (UCD) framework to provide quick and easy access to resources and plain-language explanations to help caregivers navigate SCD care.
MethodologySiby was developed by AppHatchery, an interdisciplinary team at Emory and CHOA that builds digital health tools, utilizing a five-phase UCD framework: empathize, define, ideate, prototype, test.
In the empathize phase, we identified key challenges faced by families of children with SCD through interviews with 8 caregivers, 5 CHOA providers (newborn screening coordinator, child life specialist, social worker, health navigator, and pediatric hematologist), and observations at a summer camp for youth with SCD. We grouped findings into four domains: (1) emotional and mental health (e.g. isolation, confusion), (2) physical manifestations (e.g. pain crisis management), (3) medications and treatments (e.g. limited access to new therapies), and (4) logistical and social barriers (e.g. transportation, financial strain, insurance complexity, low resource awareness).
In the define phase, we focused on logistical and social barriers as these hinder enrollment and continuity of care, making them foundational to the other domains which rely on ongoing engagement with the healthcare system. Design requirements included a tool that: (i) consolidates resources, (ii) provides logistical guidance, (iii) is intuitive for users with low tech literacy, (iv) is accessible on mobile and desktop platforms, and (v) requires minimal time commitment.
During the ideate phase, CHOA stakeholders and designers collaboratively explored solutions to the design needs. Through feasibility-focused brainstorming, we envisioned a chatbot that connects to the CHOA SCD resource library and answers common logistical questions while amplifying existing resources.
Results The resulting prototype, Siby, is a friendly AI chatbot designed using OpenAI's GPT-4o model to provide 24/7 assistance to families for common non-medical and logistical concerns. Designed with patient privacy in mind, Siby does not collect patient data and is used anonymously. While not offering medical advice, Siby is trained on institutional documents, including CHOA's SCD Handbook, insurance/financial aid guidelines, SCD teaching sheets, and CHOA's SCD webpage.
Siby offers a range of assistance, including directions to CHOA sites, public transit and parking options, information on CHOA's financial assistance policy and social worker contacts. It also offers step-by-step Medicaid guidance covering required documents, form help, and social service agency contacts. Beyond logistics, Siby simplifies SCD concepts, provides education, and guides new caregivers through first appointment preparation.
For wider accessibility, Siby isavailable for both mobile and desktop. The interface is simple and familiar with responses at an 8th-grade reading level and a conversational format for quick, on-the-go access in contrast to sifting through printed materials or web pages.
In initial testing, 10 CHOA clinicians interacted with Siby and provided real-time feedback on usability and content, which informed ongoing revisions. Clinicians saw potential in Siby's 24/7 resource access for families. Preliminary testing with 3 caregivers yielded enthusiastic responses, noting its intuitive interface, warmth, and practical utility. Testing with 12 additional caregivers is planned for the coming month.
Conclusion Through a UCD approach, we developed Siby, an engaging and resourceful AI chatbot that connects caregivers with non-medical information related to pediatric SCD care. Early feedback suggests that conversational tools like Siby have strong potential to empower families and enhance care continuity by making essential resources more accessible. Future efforts will include longitudinal testing, risk assessment, and emergency/mental health response planning. Mixed-methods feedback through surveys and usage data will inform iterative refinement of Siby's scope based on caregiver needs.
