Augmented Reality–Assisted Video Laryngoscopy and Simulated Neonatal Intubations: A Pilot Study

A small, lightweight adapter was designed to permit rapid attachment of a camera to a disposable Miller laryngoscope blade (Fig 1). A description of the adapter and preliminary results of this study were previously presented at the International Conference on Human-Computer Interaction (Orland FL, 2019).¹³  This camera and adapter unit acquires and transmits a video of the intubation process to a pair of smart glasses worn by the intubator, which permits an unobstructed view of the patient while also displaying video within the visual field (Fig 2). The video is simultaneously viewable by a supervisor via transmission to a nearby laptop and/or tablet.

The study population was comprised of nurses employed in the intensive care nursery at Duke University Medical Center. This selection of subjects was based on the assumption that they have theoretical knowledge of the intubation process and an understanding of the intraoral anatomy but have minimal hands-on experience with intubating a manikin. Baseline characteristics collected on all study participants included the number of previous simulated intubation attempts, whether the participant performed a simulated intubation within the past year, and hand preference. One potential participant was excluded because of previous experience with intubating live patients. The Duke University Institutional Review Board declared this study to be exempt from review and granted a waiver of consent.

Study subjects (n = 45) were assigned via simple randomization (using a random number generator) to use 1 of the following 3 intubation modalities: DL (n = 15), IVL (n = 15), or ARVL (n = 15). After receipt of a standardized teaching script (see Supplemental Information), each participant completed 5 consecutive intubation attempts on a Life/form Basic Infant CRiSis manikin (Nasco, Fort Atkinson, WI) using a Miller size 1 laryngoscope blade with the camera and adapter unit attached. Video was not captured during attempts by participants in the DL group. Participants in the IVL group had access to a live video stream via a local laptop placed on a table to the left of the intubator. Participants in the ARVL group wore smart glasses while performing DL; video was transmitted to the glasses as well as a local tablet accessed by a supervisor. Individualized coaching during all attempts was provided by a supervisor, who was able to view the video stream in real time while assisting those in the IVL and ARVL groups. Verbal coaching for the ARVL group was supplemented by telestration, where marks made on a tablet by the supervisor were transmitted in real time to the smart glasses. Instruction was provided by a single supervisor (a senior neonatology fellow and experienced intubator).

The outcome of each attempt was documented as successful (wherein the endotracheal tube [ETT] was placed in the airway in <30 seconds¹⁴ ), unsuccessful because of a prolonged intubation attempt (wherein the ETT was placed in the airway within 30–60 seconds), unsuccessful because of failure to intubate within 60 seconds, or unsuccessful because of esophageal intubation. Secondary outcomes included: number of successful intubations per participant; percentage of participants who achieved success within the first or second attempt; time to complete 1 intubation attempt (defined as the time from introduction of the laryngoscope blade into the mouth to its removal); time of fastest attempt; time to visually identify the airway (defined as the time from introduction of the laryngoscope blade into the mouth to the time the participant obtained an intubatable view of the airway); and time between airway identification and intubation. Of note, airway identification in the DL group was self-reported when each participant obtained what (s)he believed to be an intubatable view.

We used medians and percentiles to describe continuous variables and counts and percentages for categorical variables. Statistical analyses were performed by using Fisher’s exact test to determine the associations between the following: baseline characteristics of participants and group, attempt outcomes and group, and success in the first or second attempt and group. The Kruskal-Wallis test was used to compare among all groups the time to intubate, time to visually identify the airway, time of fastest attempt, time between airway identification and intubation, and number of successful intubations per participant. For all results, pairwise comparisons were assessed between the DL and IVL groups, between the DL and ARVL groups, and between the IVL and ARVL groups by using the Wilcoxon rank sum test. All reported P values were 2 sided, and P values of <.05 were considered statistically significant. Data were analyzed by using Stata, version 16.0 (Stata Corp, College Station, TX).

There were no statistically significant differences in the groups’ baseline characteristics (Table 1).

Outcomes of each attempt are reported in Table 2. The overall success rate of intubation attempts done with DL was 32% (24/75) compared to 72% (54/75) using IVL and 71% (53/75) using ARVL (P < .001). Esophageal intubations occurred in 27% (20/75) of attempts in the DL group, whereas there were no esophageal intubations in the IVL or ARVL groups (P < .001). The DL group had 16 of 75 (21%) failures to intubate within 60 seconds, whereas the IVL group had 5 of 75 (7%) and the ARVL group had 8 of 75 (11%) failures to intubate within 60 seconds (P = .03). For this specific outcome, the pairwise comparison between the DL and ARVL groups did not indicate a statistically significant difference. There was no difference between groups for failures due to a prolonged intubation time of 30 to 60 seconds.

The median (interquartile range [IQR]) number of successful intubations per participant in the DL group was 1 (0–3) compared to 4 (3–4) for both the IVL and ARVL groups (P = .002). Notably, 47% (7/15) of providers in the DL group had no successful intubations, whereas all (15/15) providers in the IVL group and 93% (14/15) of providers in the ARVL group had at least 1 successful intubation (P = .003).

The percentage of participants who achieved success within the first or second attempt was 27% (4/15), 80% (12/15), and 73% (11/15) in the DL, IVL, and ARVL groups, respectively (P = .008). For participants in the IVL and ARVL groups, the most frequent attempt at which success was first achieved was the second attempt; for participants in the DL group, it was more common to have never achieved a successful intubation (Fig 3).

The median (IQR) time to complete 1 intubation (successful or otherwise) was significantly faster in the IVL and ARVL groups compared to the DL group (P < .001; Table 3). The largest improvement in time to intubate appeared to be between the first and second attempts for participants in the IVL and ARVL groups, while participants in the DL group did not appear to show dramatic improvement until the fourth attempt (Fig 4). The median time of each participant’s fastest attempt (successful or otherwise) was also faster in the IVL and ARVL groups compared to the DL group (P = .02; Table 3). Additionally, airway identification was significantly faster in the IVL and ARVL groups compared to the DL group, despite the exclusion of 6 attempts in the DL group that never had an airway identification (P < .001; Table 3). For successful attempts, the time between airway identification and intubation was not statistically different among the 3 groups (Table 3).

In pairwise comparisons, differences between the DL and IVL groups, as well as the DL and ARVL groups, were congruent with the results of comparing all 3 groups, except for the outcome of failure to intubate within 60 seconds (as stated above). There were no statistically significant differences between the IVL and ARVL groups.

In a simulation environment, intubation success rates of neonatal intensive care nurses were significantly higher when using either IVL or ARVL compared to standard intubation technique. Because neonatal endotracheal intubation remains a critical skill for providers responsible for delivery room management and nursery coverage, this study responds to an urgent need for novel techniques to improve the acquisition of intubation skills by medical trainees.

In the pediatric literature, intubation competency has been defined as provider success intubating on the first or second attempt ≥80% of the time.¹⁵  Current trainees in pediatric residency programs are often unable to achieve this level of competency before completing their residency.^{1–4,15–18}  In a recent study, DeMeo et al⁴  found that an average of 8 or more intubation opportunities may be required to achieve competency in neonatal intubation. Data from the anesthesia literature are even more discouraging, suggesting that over 35 to 40 intubations performed via DL are needed to achieve procedural competency.^19–21  DeMeo et al⁴  also demonstrated that intubation opportunities for pediatric residents have been reduced to an average of 3 per trainee during their 3-year residency, a finding similar to that of a 2018 study⁵  wherein 88% of their cohort of pediatric residents performed ≤3 intubations during the 1-year study period.⁵  This is in stark contrast to the intubation experience of pediatric residents in earlier epochs, with reports from the 1990s indicating that residents performed nearly 40 intubations during their 3-year training period.²²  Contributing factors include the increasing use of noninvasive mechanical ventilation for neonates^23–25  and the change in management of nonvigorous infants born through meconium-stained amniotic fluid.²⁶ 

Video laryngoscopy has been explored as a tool to teach the skill of neonatal intubation, but the body of evidence has been conflicting. A 2018 Cochrane analysis⁷  suggested that video laryngoscopy increases the success of intubation in the first attempt but does not decrease the time to intubation or the number of attempts for intubation. It has been reported in multiple studies that time to intubation may even be longer with video laryngoscopy compared to DL,^27–29  perhaps reflecting the less intuitive nature of hand-eye coordination that is required to use IVL. Additionally, the ETT must be advanced blindly until it enters the visual field of the camera. While IVL allows for improved visualization compared to DL, a different (potentially more difficult) technique is needed to actually insert the ETT.^6,27 

This current study supports the use of IVL as a superior teaching tool compared to DL and further demonstrates that ARVL is similar to IVL when used in a simulation environment. Both ARVL and IVL allow the intubator to switch between the modalities of direct and video laryngoscopy during the procedure. Based on this study alone, it remains unclear whether it is less disruptive to shift one’s attention to smart glasses (as in ARVL) versus a separate video monitor (as in IVL); future studies involving live patients are planned to elucidate this question.

Within this study, the largest contributor to the disparity in intubation success was the number of esophageal intubations. More than one-quarter of attempts done by providers in the DL group were esophageal intubations, whereas there were no esophageal intubations in the IVL or ARVL groups. We attribute this difference to the specific coaching that is afforded by the live video stream: the expert intubator was able to identify the esophagus and airway for the novice providers in the IVL and ARVL groups, thus preventing malposition of the ETT. In the ARVL group, verbal instruction was enhanced by a telestration feature, whereby the instructor could indicate the anatomy by drawing on the tablet, and these digital marks would be transmitted instantaneously to the display appearing on the glasses. Our findings are supported by results of a study done by O’Shea et al,³⁰  in which factors associated with unsuccessful intubations were examined by retrospectively reviewing videos that were captured during the process of DL (by using a modified traditional Miller video laryngoscope). The authors found that the majority of unsuccessful neonatal intubations performed by inexperienced residents were due to esophageal intubation or failure to recognize midline anatomic structures.

Our results suggest that not only does video laryngoscopy improve one’s ability to distinguish between the airway and the esophagus, it also shortens the time to visually identify the airway. Because participants in the DL group self-reported obtaining an intubatable view, there may have been instances in which the airway was in view but not recognized by the intubator. In fact, the majority of residents in O’Shea et al’s³⁰  study were able to achieve an intubatable view, but it often went unrecognized by the cohort that did not receive video-guided coaching during the procedure. Thus, the improvement in overall time required to intubate in the IVL and ARVL groups may be directly related to this improvement in airway identification or even simply airway recognition. Additionally, time between airway identification and intubation was similar across all groups, signifying that this technology does not appear to worsen fine motor skills or interfere with the intubation process itself. Lastly, our findings of a slower time to intubate and to visually identify the airway in the DL group are especially notable, given the likelihood that a novice provider would be able to incorrectly identify and intubate the esophagus more quickly than (s)he would be able to complete a successful intubation. Because more than one-quarter of attempts in the DL group were esophageal, their airway identification and intubation times may have been artificially shortened.

Most participants in the IVL and ARVL groups achieved success within the first or second attempt, suggesting that novice providers are able to adapt quickly to the use of this technology. Whether more experienced intubators also easily adjust will be explored in future projects.

Lastly, our design of a camera and adapter unit to facilitate video laryngoscopy with blades already used in our intensive care nursery is a novel technique in and of itself. Impact on the intubator is minimal because the camera and adapter unit still allows for excellent direct line-of-sight visualization of the airway (Fig 5). In fact, many participants in the IVL group were observed to use the laptop screen when trying to locate the airway, then convert to DL for the actual intubation. This suggests a promising use for this equipment, which is to allow the supervisor to provide video-guided verbal feedback to the learner who is performing direct (rather than indirect) laryngoscopy. Two recent studies revealed that intubation success rates of novice providers using DL were significantly improved when their supervisor provided coaching while sharing their view via a video screen.^5,10  Both studies used video laryngoscope intubation systems that have slightly different shapes and dimensions to the blade and handle than most conventional neonatal laryngoscopes, and follow-up data from one of these studies revealed that >40% of attempts were not successful because of an inability to direct the ETT toward the vocal cords.³⁰  The authors attributed this difficulty to the shape of the blade, a concern that we circumvented by adding a camera to our nursery’s currently used disposable laryngoscope blade. Additionally, the single-use design of the adapter and its position outside the oral cavity eliminate the need for a costly sterilization process. We anticipate that this technology can be easily integrated into intensive care nurseries using their local laryngoscope blades.

There are several limitations to this study. First, the small sample sizes increased the impact of outliers on the data. Specifically, 1 participant in the ARVL group had much lower success rates than the other 14 participants. We chose not to exclude outliers from the analyses because this was not prespecified in our protocol. However, it is worth noting that the results may have been skewed, leading to the possibility that ARVL may have been more efficacious in some parameters than IVL. An additional limitation was on the reliance of self-reporting by the DL group regarding their visual identification of the airway. Perhaps the most important study weakness was the use of an intubation manikin, the limitations of which are underscored by the fact that all 3 groups showed significant improvements in intubation skills throughout their subsequent attempts. Finally, it is unclear whether these results are generalizable, given that the study participants were nurses (a population unlikely to intubate live patients) and that the 5 attempts for each participant were separated by only a few minutes. Whether the improved skills in the technology-assisted groups persist over time will require further study.

We describe ARVL and a novel form of IVL, which improve the intubation proficiency of neonatal intensive care nurses on manikins by shortening the time required to visually identify the airway and eliminating esophageal intubations. As the number of intubation opportunities for pediatric residents continues to decline, this study responds to an urgent need for new techniques to improve the quality of these attempts. Further studies using this technology on live patients are ongoing.

We thank Lenovo Research (Morrisville, NC) colleagues John Nicholson, Erin Wang, Litao Qiu, and Ming Qian for developing the software, designing and producing the camera adaptor, and assisting with data collection. We thank Blaire Rikard for assisting with data collection.

ARVL

augmented reality–assisted video laryngoscopy

DL

direct laryngoscopy

ETT

endotracheal tube

IQR

interquartile range

IVL

indirect video laryngoscopy