Evaluation of the InterFACE-AR System for Cardiopulmonary Arrest

Introduction Cardiopulmonary resuscitation (CPR) is provided for thousands of patients suffering from cardiopulmonary arrests (CPA) each year in North America. The provision of guideline-compliant basic life support (BLS) and advanced life support (ALS) improves patient outcomes following cardiac arrest. Unfortunately, healthcare providers struggle to consistently perform guideline-compliant BLS4, 5 and ALSuring in-hospital cardiac arrest. Deviations from American Heart Association (AHA) resuscitation guidelines are associated with decreased survival from in-hospital cardiac arrest, signaling a pressing need to design, evaluate, and implement novel strategies to improve the delivery of BLS and ALS care.

Delays in epinephrine administration, delays in defibrillation, medication dosing errors, and errors in identification and treatment of the underlying condition, represent common deviations from resuscitation guidelines associated with poor patient outcomes from cardiac arrest. Prior studies point to high mental workload amongst resuscitation team members as a major contributor to protocol deviations. The dynamic nature of resuscitation presents a significant challenge for team leaders, who engage in complex cognitive process that result in extremely high mental workload, leading to poor decision-making and errors. Reducing mental workload of team members by providing decision support during resuscitation can potentially improve quality of care during cardiac arrest. The investigators aim to address these issues by designing a novel system that integrates augmented reality (AR) and screen-based technology to provide decision support for resuscitation teams during cardiopulmonary arrest.

Cognitive aids are decision support tools that present prompts to encourage recall of information, thus freeing up mental resources to increase the likelihood of desired behavior. Cognitive aids have been trialed in different forms for use during resuscitation, including pocket reference cards and digital apps. Simulation-based studies of cognitive aid used during cardiac arrest events have shown improved adherence to guidelines, improved time to completing critical tasks, and reduced rate of critical errors. Unfortunately, existing pocket reference cards and mobile apps have significant flaws - they all require providers to search through content to identify relevant information. This presents a distraction and increases mental workload - resulting in impaired communication and poor situational awareness - both of which cause delays and errors in treatment.

Augmented reality (AR) and display screens have the potential of addressing these gaps. Augmented reality can present individualized, role-specific guidance to potentially reduce mental workload; while a display screen in the clinical environment can offer information to the entire team to improve communication and enhance situational awareness.

Augmented reality using a head-mounted device allows for participant interaction with the real environment while displaying three-dimensional interactive images in a user's field of view without disturbing normal vision. AR-based cognitive aids enable fast decision making by providing clinically relevant prompts and supporting anticipatory behaviors by listing upcoming tasks. Studies completed by our research team have shown that an AR device providing clinical guidance for the team leader improves compliance with pediatric and neonatal resuscitation guidelines, but fell short on improving time to epinephrine or defibrillation. In these studies, the AR systems were entirely reliant upon data collected from the team leader through the AR device to determine the clinical guidance. This amplified team leader workload which contributed to delays in task completionIn designing a new system, our overarching goal is to reduce mental workload by providing role-specific and timely information to key team resuscitation team members. To achieve this goal, the investigators developed a mobile app (i.e. Guiding Pad app), used by the charting nurse, which provides guidance on pending tasks as the charting nurse enters completed tasks. The Guiding Pad app interfaces with a large liquid crystal display (LCD) screen in the resuscitation room, providing a collaborative platform used to support healthcare providers during management of cardiac arrest - called InterFACE (Interconnected and Focused Mobile Applications in the Patient Care Environment). Integration of the LCD screen allows for display of the cardiac arrest algorithm and a running list of tasks completed, which we anticipate will promote communication and shared situational awareness. When connected to AR devices, the Guiding Pad app allows for role-specific decision support to be shared with the provider wearing the AR device.

In the proposed study, the investigators will evaluate the impact of an enhanced system, InterFACE-AR, which provides role-specific decision support to the team leader and medication nurse through AR devices, while concurrently optimizing team situational awareness by displaying a roadmap for patient care on the LCD screen (Figure 1). Clinical data will be collected from the mobile app on a tablet used by the charting nurse. The trial aims to assess the individual and combined effectiveness of InterFACE-AR components (i.e. AR devices and LCD screen) on adherence to AHA resuscitation guidelines during simulated cardiac arrest by conducting a randomized controlled trial with a factorial design.

The primary aim is to determine, amongst pediatric healthcare teams, if the use of AR devices alone (for the team leader and medication nurse), the LCD screen alone, or AR devices and LCD screen combined (i.e. InterFACE-AR), compared to groups using the AHA pocket reference card (control), improves adherence to AHA resuscitation guidelines. Our secondary aims are to determine the impact of the AR devices, LCD screen and the full InterFACE-AR system on provider workload, cognitive load, leadership and teamwork behaviors, types and patterns of language used during resuscitation, clinical performance (e.g. number and types of pauses in CPR), and to describe user experiences with the components of InterFACE-AR.

METHODS The investigators will conduct a prospective, simulation-based randomized controlled trial with a factorial study design. Simulation-based research confers the advantage of answering research questions without risk of harm to patients. Ethics approval will be obtained at all sites. Participants will be recruited to the study in teams of five, comprised of a team leader, charting nurse, medication nurse, and two bedside clinicians who will perform CPR. Two research actors, playing the scripted roles of airway provider and CPR Coach, will join each group to comprise resuscitation team of seven healthcare providers. Actors will be trained with methodology successfully used in prior multicenter trials. Participant teams will manage one cardiopulmonary arrest simulation scenario, using the specific trial intervention assigned during randomization.

The factorial study design will assess the individual and synergistic impact of two interventions designed to provide decision support: (a) Augmented reality devices for the team leader and medication nurse; and (b) LCD Screen.

Intervention A: Augmented Reality Devices - data collected from the Guiding-Pad app will be fed to two AR devices (i.e. team leader and medication nurse), which will provide role-specific decision support for these two providers.

Intervention B: LCD Screen - data collected from the Guiding-Pad app will be fed to the LCD screen, which will be mounted on a wall, in clear view for the entire resuscitation team to see.

Intervention C: InterFACE-AR System - data collected from the Guiding-Pad app will be fed to the two AR devices (i.e. team leader and medication nurse) to provide role-specific decision support, and to the LCD screen, in clear view for the entire resuscitation team to see.

Control Group: AHA Pocket Reference Card - all participants will have access to the AHA pocket reference card, which is the mostly commonly used cognitive aid during resuscitation.

Following randomization, all participants will view a standardized orientation video describing the clinical environment, equipment, manikin functionality, and participant roles. Video content will be customized to include an orientation to the intervention assigned to each group, followed by a 15-minute table-top practice simulation where participants have opportunity to use the assigned intervention in their provider roles by running through a cardiac arrest scenario while seated at a table. For example, participants in the InterFACE-AR group will do a table-top simulation with the medication nurse and team leader using the AR device, charting nurse using the Guiding Pad app, and the LCD screen visible to the entire team. Participants assigned to the control group will participate in a table-top simulation using the AHA pocket reference card.

An 18-minute cardiopulmonary arrest simulation scenario (ventricular tachycardia to ventricular fibrillation to pulseless electrical activity) will be run for all participants, with CPR quality data collected by the Zoll R Series feedback defibrillator. All recruitment sites will utilize the identical pediatric manikin (SimJuniorTM, Laerdal Corporation), specifically designed and calibrated for CPR training. All simulation scenarios will be tightly standardized by using a scenario template with highly scripted actor roles and patient progression. Following completion of the scenario, participants will complete the NASA-TLX survey, User Experience Questionnaire, System Usability Survey, Technology Acceptance Model survey, and STAI survey. Scenarios will be videotaped from a birds-eye view angle at the foot of the bed and by high-definition action cameras worn hand-free by the leader and medication nurse. After the session, all participants will receive an educational debriefing to address performance issues using a blended-method approach to debriefing.

Randomization will occur at the level of the team, stratified by study site and sex of the team leader (to ensure equal distribution of sex in both arms), and conducted in blocks of 4 to ensure an even distribution of teams across study arms. Randomization packages will be prepared at a central study site using a web-based random number generator (e.g. www.sealedenvelope.com). Sequentially numbered recruitment packages provided for each site will contain sealed opaque envelopes (i.e. one envelope per team) with study arm assignments and unique identifier codes for participants.

Sample size estimation is based on the primary outcome measure. The sample size estimation assumes that the main effects of the 2 different interventions on the time to first dose of epinephrine would comprise the 2 primary comparisons. Allowing for Bonferroni adjustments, p < 0.025 was considered statistically significant. Time to the first dose of epinephrine was previously reported to be approximately 165 sec with a standard deviation of 60 sec. We aim to improve the time to first dose of epinephrine to 120 sec, which is associated with improved clinical outcomes from pediatric cardiac arrest10. Assuming similar variability amongst teams receiving either intervention, to detect a difference of 45 sec in time to administer first dose of epinephrine, with a power of 0.8 and a significance level of 0.025 for each of the comparison, the required sample size will be 68 teams (17 teams per study arm). Based on our prior experiences with simulation-based research, we estimated 15% of missing data due to technical issues. Accounting for missing data, we aim to recruit a total of 80 teams (20 teams per study arm).

STATISTICAL ANALYSIS Demographic characteristics of participants including sex, gender, profession, and experience will be reported using descriptive statistics across all study arms. The analyses will be conducted in the unit of teams. The investigators will use multivariable linear regression analysis to assess the effect of both interventions on the primary and secondary outcomes and include the interaction term in the model. As the effects of the two interventions may be non-additive, the investigators will also report the effect of each intervention at each level of the other intervention. All estimates of intervention effects will be reported as mean differences with 95% confidence intervals.