AI

A pediatric emergency department tested whether AI could help families tell their story before the visit

Boston Children’s tested AI-assisted histories from the waiting room. The pilot examined usability and summary quality.

Author
HDM Staff
Save
Share
Photo credit: HLTH USA 2026.

    LAS VEGAS — Boston Children’s Hospital researchers asked 31 patients or caregivers to use an AI tool to put their account of an illness into words while they waited in a pediatric emergency department. Families generally found it easy to use, and physicians reviewing the resulting summaries rated them favorably, though some lacked information about earlier visits. The team collected the summaries alongside usual care and kept them out of clinical decision-making.

    Alessio Morley-Fletcher, a pediatric hospitalist and co-chair of the hospital’s Department of Pediatrics AI committee, described the study at HLTH USA in a discussion about how patients move between care teams. He framed the pilot around a question healthcare systems rarely ask about the time before an encounter: “When should care begin?” The study examined whether patients and caregivers could use the waiting period to prepare their history, without establishing whether the approach changes treatment or gets patients seen sooner.

    Christina Cipriano, a professor at UMass Amherst and founder of the Education Collaboratory, moderated the Velocity Stage panel. Morley-Fletcher was joined by Dr. An Nguyen, chief clinical officer at CareQuest Institute for Oral Health, and Dr. Stephen Traub, chief medical officer for Optum Care Delivery. The speakers approached the same problem from different points in care: the time a family spends preparing to talk, the moment a referral is considered and the next appointment a patient has to arrange.

    What the emergency department study measured

    In the published pilot, patients or caregivers used a version of ChatGPT-4 that the investigators described as compliant with federal health privacy requirements. The researchers recruited people with Emergency Severity Index scores of 3 to 5 and asked them to use the tool to create a history of the current illness. Pediatric emergency physicians then reviewed the histories for qualities such as completeness, accuracy and readability.

    Participants gave usability a median rating of 10 out of 10, with an interquartile range of 8 to 10. Physician reviewers gave readability a median score of 9, with an interquartile range of 7 to 9. Clinicians also reported missing features of some prior visits. That shortcoming matters because what happened in an earlier encounter may change how a clinician interprets a symptom described today.

    The study was prospective and used ratings and open-ended responses to examine acceptability. It enrolled a small group in one pediatric emergency department, and its summaries did not influence clinical decisions. It therefore provides evidence that participating families could use the tool and that physicians could review the result. Questions such as whether it reduces repeat questioning, changes wait times or improves decisions would require a study that follows its use in care.

    Morley-Fletcher said the work started with a problem both families and clinicians know. A caregiver may recount pieces of a history more than once, while the clinician has limited time to assemble those pieces and ask what is missing. He also called for treating “trust as a clinical end point,” meaning a quality teams would measure and try to improve. That was a proposal he made onstage, outside the scope of what the pilot measured.

    Make the reason for the referral visible

    Traub described a different effort at Optum. Primary care clinicians and specialists jointly develop pathways intended to help a referring clinician decide when specialty care would add value. He called the work “referral process optimization” and said the pathway can help a clinician check whether appropriate steps have been tried and whether the patient is at the point where a specialist is needed.

    He used knee pain as an example of a situation in which the primary care clinician can consult a pathway before deciding what to do next. He said that when the clinician judges a referral is needed, it should go ahead. The example was an account of Optum’s approach, rather than advice on how to treat an individual patient or evidence that any particular pathway reduces delays.

    Traub also emphasized “transitions of care,” giving two practical tests for a handoff. The patient should understand what the clinician has said, and the next person to see them should understand how the new encounter fits the work already underway. The approach puts the reason for a decision alongside the decision itself. It also gives a patient a clearer way to tell whether the referral advanced or whether someone needs to follow up.

    Oral health exposes the gaps between care teams

    Nguyen said oral health belongs in an effective primary care system, where teams can ask whether patients need preventive or dental care and help them reach a dentist. When the two systems operate separately, she said, they “leave patients to navigate the gaps alone.” She described people arranging their own appointments, working out where to go and repeating information about their health as they cross from one practice to another.

    Nguyen said medical teams can identify oral health needs and help patients navigate to dental care, while dental professionals can support chronic disease management within their role. The point was to design the work so that patients can see which member of the team is responsible for the next step. Nguyen called for “data systems that connect to each other” alongside workflows and payment arrangements that make continuing coordination possible.

    Cipriano drew the discussion back to patients and families who can see a record but may still need someone to explain what it means. A report that is available in a portal can help prepare a family for a conversation. The family also needs language it can use to ask questions about the result, particularly if different clinicians see the patient at different stages.

    Across the examples, the practical test for a new intake tool, a referral pathway or a shared record is what the next person can do with it. A family can prepare its story sooner, and a clinician can record why specialty care is needed. The handoff works when the patient understands the next step and the receiving team has enough information to act on it.