Presentation
Iterative Refinement of an Automated Chatbot Intervention Embedded in an Electronic Medical Record
SessionPoster Session 2
DescriptionBackground: Falls are the leading cause of injury-related emergency department (ED) visits among older adults, yet few patients are notified of their fall risk or connected to prevention resources. This study describes the development, iterative refinement and implementation of a novel digital health intervention that automates fall risk notification and referral using an electronic health record (EHR)-integrated chatbot, Livi. The intervention is designed to reduce clinician burden and improve patient outcomes by leveraging existing clinical workflows and human-computer interaction principles to deliver location-specific fall prevention resources to high-risk patients at ED discharge.
Methods: The intervention was developed through collaboration between clinical, informatics, and industry partners, including Avaamo’s AI conversational platform. Livi guides patients through a structured dialogue to access free or low-cost, evidence-based fall prevention programs. The chatbot is launched via a QR code embedded in the After Visit Summary (AVS) of patients who screen at high fall risk, based on nursing documentation in the EHR. This fully automated workflow minimizes disruption to clinical care.
User-centered design was central to the development process. Usability testing with 93 older adults, caregivers, and community members across three settings (“Research Roadshows” at one religious site, one assisted living facility, and one academic medical campus) informed iterative improvements to Livi’s interface and content.
Results: Enhancements included increased font size, Spanish language support, expanded geographic coverage, fall risk self-assessment tools, and feedback mechanisms. In November 2024, the intervention was deployed across 17 EDs within a single health system using a shared Epic instance, demonstrating scalability and rapid implementation potential.
Discussion: Key human factors considerations included minimizing cognitive load for clinicians, optimizing accessibility for older adults, and designing intuitive chatbot interactions. The intervention aligns with implementation science principles by embedding digital tools into routine care and engaging end users throughout development.
Future directions include expanding Livi’s conversational capabilities using large language models (LLMs), enhancing usability for diverse populations, and conducting a randomized controlled trial to evaluate clinical outcomes such as fall recurrence, resource utilization, and patient engagement. This work contributes a replicable model for integrating digital health tools into emergency care and highlights the role of human factors in designing scalable, patient-centered interventions.
Methods: The intervention was developed through collaboration between clinical, informatics, and industry partners, including Avaamo’s AI conversational platform. Livi guides patients through a structured dialogue to access free or low-cost, evidence-based fall prevention programs. The chatbot is launched via a QR code embedded in the After Visit Summary (AVS) of patients who screen at high fall risk, based on nursing documentation in the EHR. This fully automated workflow minimizes disruption to clinical care.
User-centered design was central to the development process. Usability testing with 93 older adults, caregivers, and community members across three settings (“Research Roadshows” at one religious site, one assisted living facility, and one academic medical campus) informed iterative improvements to Livi’s interface and content.
Results: Enhancements included increased font size, Spanish language support, expanded geographic coverage, fall risk self-assessment tools, and feedback mechanisms. In November 2024, the intervention was deployed across 17 EDs within a single health system using a shared Epic instance, demonstrating scalability and rapid implementation potential.
Discussion: Key human factors considerations included minimizing cognitive load for clinicians, optimizing accessibility for older adults, and designing intuitive chatbot interactions. The intervention aligns with implementation science principles by embedding digital tools into routine care and engaging end users throughout development.
Future directions include expanding Livi’s conversational capabilities using large language models (LLMs), enhancing usability for diverse populations, and conducting a randomized controlled trial to evaluate clinical outcomes such as fall recurrence, resource utilization, and patient engagement. This work contributes a replicable model for integrating digital health tools into emergency care and highlights the role of human factors in designing scalable, patient-centered interventions.
Event Type
Poster Presentation
TimeTuesday, March 244:45pm - 6:15pm EDT
LocationRhinelander Gallery
Digital Health