Ambient augmented artificial intelligence in medicine: Proof-of-concept in pediatric resuscitation

ABSTRACT

Intro/Background: Simulation enhances emergency medicine provider's competence, improving patient health outcomes. However, high fidelity simulation is costly, requires in-person training and is logistically challenging due to the COVID-19 pandemic. Screen-based simulation using virtual patients and augmented reality (AR) are exciting technologies, available on-demand, and can directly address these barriers. Furthermore, enhancing user experience by incorporating artificial intelligence and ambient intelligence (ability to integrate information across multiple platforms) will enrich clinician decision making in real-world context. Purpose/Objective: Primary Aim: (Efficacy) To develop a proof-of-concept, ambient, augmented, artificial intelligence (AAAIM) application and demonstrate its efficacy as an innovative cognitive aid to improve patient outcomes in pediatric resuscitation. Secondary Aims: (Acceptability) Evaluate the acceptability of AAAIM as a clinical decision support cognitive aid for pediatric resuscitation in simulated settings and utilize an iterative, rapid prototyping and design cycle to improve the user experience. Methods: We assembled a team of ED clinicians, educators, developers, designers, and a cognitive psychologist. Using web technology and communication protocols to share information across many devices and services simultaneously in real-time, we used open source software to develop an interactive PALS algorithm for cardiac arrest that was available concurrently on an AR headset, laptop and on the web, which provided prompts to the trainee as he/she underwent a simulated pediatric resuscitation scenario. Outcomes (if available) Primary a) Time between residents' recognition of a patient in cardiac arrest to first treatment. Secondary a) Frequencies of successful completion of each anticipated step of the PALS algorithm b) Stress and cognitive load (using objective and subjective measurements by Electrodermal Activity-EDA) c) Participants' acceptability and usefulness ratings of AAAIM as a cognitive aid (measured by a Modified Technology Acceptance Model survey) Summary: High fidelity, manikin-based simulation training effectively improves provider/trainee decision-making skills, procedural competency and patient outcomes. However, it is expensive and logistically challenging especially given the current COVID-19 pandemic. Currently available cognitive aids, such smartphones, VR/AR headsets, etc., cannot be integrated into clinical workflow (real-world), or have not been designed with the end-user in mind. We have used the tenets of ambient user experience (i.e., ability to display and interact with meaningful content simultaneously), augmented reality (i.e., ability to digitally enhance the real world with holographic images) and artificial intelligence (learn from and respond to the clinician) to create AAAIM-a web-based software application. The application is device and operating system agnostic, available for training (on demand) and clinical care (fully integrated within current workflows). Such a system allows the trainee to improve decision-making skills that can also be broadcast, thus allowing trainers and experts to connect remotely with the user in real-time for immediate feedback. As a proof-of-concept, we have developed a highly interactive PALS cardiac arrest algorithm that sequentially suggests next steps and provides tips for safety checks (calculates medication doses, times chest compressions, recommends a change of compressor at 2 minutes, etc.,). This interactive algorithm can be visualized using ambient technology across platforms (AR headset, smartphone, tablet etc.) and viewed by trainers in the simulated setting. Sessions can be recorded and trainees can assess performance and knowledge gaps. Furthermore, since this application can be used in a real-world context and broadcast, remote experts can connect with the user and interact to co-manage patients.