Stressful Environments Negatively Impact the Resuscitation Skills of Doctors of Different Experience. A Randomized Controlled Cross-Over Trial

Abstract

Background

Cardio-pulmonary resuscitation (CPR) is an essential skill that all health providers must master. Simulation training on traditional manikins in a normal classroom does not prepare the trainees for the stress that could reduce the resuscitation performance of especially inexperienced providers during a real cardiac arrest. The aim of this randomized cross-over trial was to explore whether training using augmented reality (AR) was realistic enough to induce stress and impact the resuscitation skills of doctors of different experience levels.

Methods

Novices, intermediates, and experienced physicians participated in this single-center, randomized, cross-over, simulation-based study. All participants had to perform CPR on a high-fidelity manikin with and without AR. In the immersive AR environment, sounds from multiple sources and visuals of bustling crowds were used to simulate a stressful environment. The State-Trait Anxiety Inventory (STAI) were used to measure the participants’ perceived stress levels during the simulation sessions. Frequency and depth of chest compressions, and the ratio of chest compression time to the entire recovery cycle, i.e. the Chest Compression Fraction (CCF) were used to objectively measure the quality of the resuscitation. Trial Registration: not applicable.

Results

This study included 80 participants, novices=28, intermediates=27, and experienced=25. The novices reported significantly more stress during the AR simulation while participants with a higher level of experience were less affected by the stressful environment induced by the AR system. All three groups performed significantly worse regarding compression depth when they were in the stressful AR environment. Overall, 66 (83%) participants used the correct compression rate in the standard environment vs. 44 (55%) in the AR induced stressful environment, p<0.001; only the experienced did not perform significantly worse, p=0.36. The CCF increased significantly from 56% to 60% when performing resuscitation while using AR, indicating slightly better adherence to international recommendations.

Conclusions

AR successfully added stress to performance of resuscitation in a simulated environment. It made it difficult for participants to achieve accurate depth of compression under pressure and the inexperienced also performed worse regarding compression rate and reported feeling more stressed. Training resuscitation using AR might prepare health care providers for the stressful situations of a clinical cardiac arrest.

Introduction

Cardiac arrest is a leading cause of death in developed countries with high incidence rates and mortality rates affecting more than 500,000 people every year.^1,2 Prompt provision of high-quality CPR can significantly improve survival rates ³ making CPR a fundamental skill that all healthcare personnel must master.

Standard textbooks and written guidelines are the main learning method for acquiring necessary theoretical knowledge, but hands-on training is required to acquire practical skills. When CPR is required in clinical settings, it is often in urgent and stressful situations that are not well suited for clinical training and improvement in skills. Hands-on simulation-based training using manikins is widely used and can be combined with devices providing real-time feedback that helps ensure the quality of the CPR training. ⁴ Chest compression quality has been shown to be affected by anxiety level. ⁵ Therefore, staying calm and performing high-quality CPR in a stressful environment is a critical skill that doctors of all experience levels need to manage and improvement in skills. However, CPR is often taught in traditional classroom settings, which do not replicate the stress associated with CPR in clinical settings. If the simulation of stressful scenarios can be incorporated into training, the transfer of CPR skills acquired through training to CPR in clinical settings might be improved.

AR is a popular and emerging technology. Unlike fully immersive VR, AR maintains a connection between users and the real world by design, which is a synthesis of virtuality and reality. ⁶ AR attracts users’ visual and auditory senses by creating illusions of the surrounding environment^7,8 and has the potential to supplement traditional learning methods. The technological advances allow us to use AR and video-monitors to simulate stressful environments ⁶ but it is unknown whether training in this more realistic and more stressful environment will result in better learning and better preparation for practice in clinical settings. The first step in this line of research would be to explore whether the AR simulation is realistic enough to induce stress for both novices and more experienced doctors to a level that changes their performance levels.

This study aimed to investigate whether AR immersion in a simulated stressful environment affected the perceived stress and the resuscitation skills of doctors with different levels of experience.

Methods

Setting

The single-center, randomized, cross-over, simulation-based study was conducted from January 2024 to February 2024 at the Clinical Skills Center of the Affiliated Zhuzhou Hospital Xiangya Medical College CSU, Zhuzhou, China. The study was approved by the ethics committee of the Affiliated Zhuzhou Hospital Xiangya Medical College CSU (No:KY2023059-01), who also waived the need for consent to participate according to national legislation.

Participants

Novices, intermediates, and experienced physicians were recruited by invitation or through WeChat groups and included in the study. Novices were medical students in the 4^th or 5^th year of their undergraduate medical degree program at the Xiangya Medical College of CSU and Hunan University of Chinese Medicine, without any real-life CPR experience on patients. Intermediates were residents in the 2^nd or 3^rd year of their residency programs at the Affiliated Zhuzhou Hospital Xiangya Medical College CSU, who had taken a CPR training course using a model with built-in feedback device during the last two years and had performed 1 to 10 CPRs on patients. Experienced physicians were doctors from the emergency department and other departments such as cardiology, neurology, intensive care medicine, and anesthesiology. They had received American Heart Association’s (AHA) Basic Life Support (BLS) course according to the AHA guidelines ⁹ and all had performed more than 30 CPRs on patients. All participants volunteered to participate in the study and were not financially compensated. Exclusion criteria were: (1) Participants who were unable to perform chest compressions due to physical concerns or illnesses, (2) Participants who have experienced mental stress or anxiety disorder recently, (3) Participants who had engaged in vigorous exercise 30 minutes prior to the recording or had consumed any drinks containing caffeine or alcohol 24 hours before performing the procedure.

We explained the nature of the experiment to the participants before the experiment. All included participants completed a demographics questionnaire online and underwent a standardized 15-minute qualifying training session to ensure similar levels of familiarity with the training equipment. The main investigator was available to answer questions and resolve any technical issues but did not offer any guidance or help during the simulated resuscitations. All participates completed the trial in their free time.

Intervention and Control

All participants were given the same case of a patient drowning and experiencing sudden out of hospital cardiac arrest. They had to perform CPR on a high-fidelity manikin with and without AR. We utilized the Laerdal Resusci Anne QCPR, which provides quantitative feedback on key factors influencing CPR quality.

In the immersive AR environment, the trial assistant helped participants wearing the Google glasses and explained the points for attention. Participants manually triggered the start button after seeing the scenario description in the pop-up window through Google glasses. At this time, the sound from multiple angles and the visual effect of the bustling crowd are used to truly simulate the pressure in the whole rescue process (see Figure 1). In contrast, the traditional training environment does not include the vision or hearing of the AR environment, and the same scenario description is only carried out through the prompt card. After reading, the participants started the first aid operation by themselves. Participants were randomly assigned to perform with AR first (group A) or without the use of AR tools first (group B). Both modes performed 5 cycles of data collection. Participants were informed of successful resuscitation when assessing their pulse about 2 minutes later, and the trial was terminated.

Figure 1.

Visual effects seen by participants through AR glasses in an immersive AR environment

Outcome Measures and Timing

Participants’ perceived stress levels during the simulation sessions were evaluated through self-report instruments collected separately after each CPR procedure. The State Trait Anxiety Inventory (STAI) is a measurement tool used to assess the psychological load and the changes in participants caused by high fidelity stress simulation environments. ¹⁰ It includes the State Anxiety Inventory (S-AI) and the Trait Anxiety Inventory (T-AI) and employs a 4-point Likert scale to assess anxiety. S-AI mainly reflects subjective and transient feelings of nervousness, worry and fear, especially anxiety levels in stressful situations, which is not stable in terms of time and intensity. T-AI mainly reflects a personality trait with individual differences in anxiety tendencies. It reflects the individual’s frequent emotions with a relatively high level of consistency. ¹¹ The S-AI and T-AI each have 20 questions, with a minimum score of 20 points and a maximum score of 80 points. The higher the total score on a rating scale, the higher the anxiety level of the participant.

The outcome measures were selected according to the AHA guidelines and included frequency and depth of chest compressions (the recommended rate is 100 to 120 per minute and the correct depth is 50 - 60 mm), and the ratio of chest compression time to the entire recovery cycle, i.e. the Chest Compression Fraction (CCF) that is recommended to be over 60% to limit interruptions and maintain coronary perfusion during resuscitation. A sensor placed on the chest of the mannequin measured both compression rate (in milliseconds) and depth (in millimeters) at initial and peak levels.

Sample Size Calculation

Our primary outcome measure was the proportion of chest compressions meeting AHA standards for depth. This number has been reported to be approximately 60% among health care professionals deprived of CPR feedback training. ¹² We expected that our participants would achieve 70% correct compressions in a quiet learning environment (due to their training) but only 60% in the stressful environment. With a power of 80%, P=0.05, and assuming comparable variability of 15% in both environments, a sample size of 72 would be required.

Statistical Analysis

Descriptive statistics was used to elucidate the demographic characteristics of two groups of participants. Independent sample t-test was used when quantitative data was normally distributed, and Mann Whitney U test was used when it was non-normally distributed. Fisher’s exact test was used for categorical data. Paired sample t-test was used to compare the STAI score, S-AI score, T-AI score, and CCF value with and without AR. Chi square test was used to analyze the difference in the accuracy of pressing frequency between two groups. A non-parametric test (Wilcoxon signed rank sum test) was used to analyze the differences in accuracy of compression depth.

All data analyses were performed using a statistical software (SPSS Statistics 27; IBM Corporation). P values < 0.05 were considered to indicate statistical significance.

Results

This study included 80 participants, 42.5% (n=34) males and 57.5% (n=46) females, aged 28.9 ± 7.2 years. Twenty-eight were in the novice group, 27 in the intermediate group, and 25 in the experienced group. None of the participants had any experience using AR applications. The participants were randomly divided into group A (37 participants) and group B (43 participants). There were no statistical differences between group A and B in terms of gender, professional title, education, major, experience performing CPR, CPR training certificates acquired, identity, and rotation experience in specific departments (see Table 1). The pre-study variables of the two groups passed the homogeneity test.

Table 1.

The Demographic Characteristics Between the Two Groups

​		Group		p value
		Group A (n = 37)	Group B (n = 43)
Age (years)		28±7	29±8	0.71
Gender	Male	17 (50%)	17 (50%)	0.65
	Female	20 (44%)	26 (57%)
Professional title	None	17 (45%)	21 (55%)	0.67
	Junior	10 (59%)	7 (41%)
	Intermediate	4 (44%)	5 (56%)
	Senior	6 (38%)	10 (63%)
Education level	Bachelor	30 (51%)	29 (49%)	0.21
	Master	7 (33%)	14 (67%)
Status	Intern	12 (43%)	16 (57%)	0.52
	Resident	15 (56%)	12 (44%)
	Staff	10 (40%)	15 (60%)
Discipline	Internal medicine	11 (46%)	13 (54%)	0.63
	Surgery medicine	6 (60%)	4 (40%)
	Gynecology	1 (100.0%)	0 (0%)
	Stomatology	0 (0%)	1 (100%)
	Emergency & Intensive care medicine	7 (54%)	6 (46%)
	Clinical medicine	12 (39%)	19 (61%)
Experience of CPR	0	12 (43%)	16 (57%)	0.52
	1∼10	15 (56%)	12 (44%)
	>30	10 (40%)	15 (60%)
Certificate of CPR	None	8 (36%)	14 (64%)	0.66
	First Responder	8 (53%)	7 (47%)
	BLS	6 (46%)	7 (54%)
	ACLS	1 (100%)	0 (0%)
	First Responder & BLS	11 (55%)	9 (45%)
	First Responder & ACLS	3 (33%)	6 (67%)
Worked in ICU, Emergency or Cardiology Departments	No	5 (42%)	7 (58%)	0.77
	Yes	32 (47%)	36 (53%)

As shown in Table 2, the STAI, S-AI, and T-AI scores of the AR group were significantly increased indicating that the AR simulation was sufficiently realistic to induce self-perceived stress. Further analysis revealed that participants with a higher level of experience, were less affected by the stressful environment induced by the AR system. The STAI, S-AI, and T-AI scores of the novices group showed statistical differences (p < 0.05), while only the S-AI scores of the intermediate group showed statistical differences (p < 0.05). There was no statistical difference in the experienced group.

Table 2.

The Results of State Trait Anxiety Inventory and the Two Sub-Scales After Performing CPR in a Standard Training Environment and in an AR Environment Including Stressors

​		AR induced stressful environment (n = 80)	Standard environment group (n = 80)	p
S-AI	Total (n=80)	39.8±11.2	35.2±10.3	<0.001
	Novices (n=28)	44.8±8.6	38.6±8.5	0.001
	Intermediates (n=27)	40.4±10.3	35.3±10.4	0.027
	Experienced (n=25)	33.6±12.0	31.2±11.0	0.18
T-AI	Total (n=80)	39.3±10.7	36.4±9.6	0.007
	Novices (n=28)	44.1±8.1	38.7±8.0	0.003
	Intermediates (n=27)	38.9±11.5	37.4±10.0	0.49
	Experienced (n=25)	34.3±10.2	32.6±10.0	0.32
STAI	Total (n=80)	79.0±21.4	71.5±19.6	<0.001
	Novices (n=28)	88.8±16.2	77.4±16.3	0.001
	Intermediates (n=27)	79.2±21.3	72.6±20.1	0.13
	Experienced (n=25)	67.9±22.0	63.8±20.6	0.22

As shown in Table 3, the AR group showed a significant decrease in the quality of compression frequency and depth, while the CCF increased significantly. Further analysis revealed that doctors with different experiences also showed different outcomes in chest compression quality. There were statistically significant differences (p < 0.001) in the accuracy of compression frequency, compression depth, and CCF value in both the novice and intermediate groups. The experienced group only showed statistical differences in accuracy of compression depth (p < 0.001), while there were no statistically significant differences in compression frequency and CCF value (p > 0.05). It was difficult for all three groups to achieve accurate depth of compression under pressure.

Table 3.

Chest Compression Quality Among Participants of Different Levels in a Standard Training Environment and in an AR Environment Including Stressors

​		AR induced stressful environment	Standard environment	p
Used correct compression rate (%)	Total (n=80)	44 (55%)	66 (83%)	<0.001
	Novices (n=28)	14 (50%)	22 (79%)	0.026
	Intermediates (n=27)	14 (52%)	25 (93%)	<0.001
	Experienced (n=25)	16 (64%)	19 (76%)	0.36
Accuracy of depth	Total (n=80)	33%±36%	98%±6%	<0.001
	Novices (n=28)	45%±34%	96%±10%	<0.001
	Intermediates (n=27)	26%±37%	99%± 3%	<0.001
	Experienced (n=25)	27%±37%	99%± 4%	<0.001
Chest Compression Fraction	Total (n=80)	60%±8%	56%±5%	<0.001
	Novices (n=28)	58%±9%	54%±6%	0.003
	Intermediates (n=27)	61%±7%	56%±4%	0.001
	Experienced (n=25)	60%±9%	57%±5%	0.082

Discussion

Simulation-based learning has significant benefits for training. AR technology combines reality and the virtual world through the use of several programs on available devices such as Google Glass and Microsoft Hololens 2, allowing learners to experience an immersive clinical environment. ¹³ This technology has been successfully implemented in the field of CPR and has demonstrated its feasibility and usability.^14-16 The three key components of fidelity associated with simulation are device fidelity, environmental fidelity, and psychological fidelity. ¹⁷ All participants in this study performed out-of-hospital CPR on a teenager who had suffered a cardiac arrest after drowning in a simulated park lakeside scene. This AR was purchased (commercially available), but my team collaborated with engineers to provide experimental scenario scripts and live sound sources, which were maintained and implemented by the engineers. This study on the basis of device fidelity, further improving environmental fidelity and psychological fidelity helped recreate the circumstances of real clinical practice.

In this study, participants conducted self-assessment after each simulation and the AR group experienced increased psychological pressure due to fidelity and unfamiliarity compared to the traditional group performing in a quiet and calm environment. In addition, the level of anxiety experienced at the end of the simulation session was negatively correlated to the level of experience a participant. Although simulation is an educational method that minimizes risks, it potentially creates psychological risks for trainees. Research has found that the main causes of stress are real-world reproduction, external intervention, patients’ critical situations, and the possibility of adverse outcomes.^18,19 In addition, sources of stress also include the lack of experience, lack of self-confidence, behavioral errors, and the presence of teachers.^20,21 Existing literature has found that sudden exposure to stressful environments during simulation-based training is negatively correlated with the acquisition of knowledge, skills, and attitudes. ²² Elevated stress affects cognitive and memory changes.^23,24

Guideline-based CPR has been proven to be crucial for survival in cardiac arrest. It is usually recommended to repeat CPR training every 1 or 2 years, to maintain the continuity of knowledge and skills. ²⁵ Even experienced medical professionals need continuous practice to meet high-quality requirements.

There is evidence to suggest that indicators closely related to patient prognosis are appropriate compression speed and depth. ²⁶ Studies have found that in hospitals, 36% -87% of CPRs are not performed with high-quality chest compressions.^27,28 Specifically, the quality of CPR (CPR) performed by healthcare providers is lower than recommended guidelines in terms of speed and depth. ²⁹ Due to the influence of environment and conditions, the quality of CPR provided by out of hospital healthcare providers is even less satisfactory. In critical situations, emotional control of the rescuer plays a crucial role. In this experiment, the accuracy of compression frequency and depth in the AR group were significantly lower than that in the control group, indicating that the accuracy of compression frequency and depth significantly decreases under stress, and psychological and physiological stress has a negative impact on clinical manifestations. ³⁰ On the other hand, it is worth acknowledging that due to the rescuer prioritizing cardiac compressions during CPR, the CCF value increased. Our study showed that clinical experience can counteract self-perceived stress and partly compensate for the drop in performance caused by a stressful environment. The experienced group reported significantly lower levels of stress and anxiety compared to their less experienced colleagues and showed stability in compression frequency accuracy. Their CCF was also unaffected by pressure, but the compression depth accuracy dropped significant while using AR group.

It is unknown whether training using our AR system will better prepare healthcare providers for the stressful situations encountered during a clinical heart arrest. In a randomized trial, an AR CPR application did not improve the quality of CPR performed compared to standard training. ³¹ However, simulating critical situations or complex scenarios with various stressors multiple times during training may promote coping strategies to reduce trainees’ anxiety and improve the quality of their CPR performance. Some believe that the closer the stimulus measures are to the actual situation, the more they can promote learning outcome transfer. This study allowed participants to run simulated scenarios that were suitable for their abilities, providing a new possibility for future research on high-quality CPR outside the hospital, while also providing evidence for the development of AR emergency training to improve the quality of CPR. Simulation of urgent scenarios in a safe and controlled environment helps enhance trainees’ critical thinking, confidence, and clinical judgment. ³² Research shows that AR immersive training can provide feedback and enhance the trainee’s emergency muscle memory, prolong the retention time of CPR operation skills, and extend to simulate an inclusive recovery experience based on physiology in high-pressure environments, ¹⁴ thereby improving the ability to transform skills into real emergency scenarios.

This research has several strengths. First, it is a randomized controlled cross-over trial, so that all interventions are compared with the participants themselves as the control group, so that it is not affected by the characteristics of the participants. Secondly, we recruited 80 participants, including doctors at different levels, avoiding the bias caused by a single group. Several limitations also should be acknowledged. Firstly, the trial relies on only one simulation scenario, lacking different scenarios to capture a wider range of conditions and reduce bias. Secondly, skill decline has not been assessed. In addition, equipment and maintenance costs, as well as public acceptance towards AR, are also issues worth considering. In the future, recruitment should be conducted in multiple centers for further research to evaluate the long-term effectiveness and applicability of interventions in emergency situations.

In summary, stress is an important variable that affects the outcomes of simulation-based medical education. Immersion in a stressful environment can affect the CPR skills of doctors with different experiences. The incorporation of AR in CPR training may be an effective educational modality, in which high-stress environments can be included in trainees’ experience with CPR during training to better simulate real clinical care.

Appendix.

Abbreviations

CPR

Cardio-pulmonary resuscitation

AR

Augmented Reality

VR

Virtual Reality

STAI

The State-Trait Anxiety Inventory

S-AI

The State Anxiety Inventory

T-AI

The Trait Anxiety Inventory

CCF

Chest Compression Fraction

AHA

American Heart Association

BLS

Basic Life Support

ORCID iD

Manhui Huang https://orcid.org/0009-0007-7119-0090

Ethical Considerations

The study was approved by the Ethics Committee of the Affiliated Zhuzhou Hospital Xiangya Medical College CSU (No: KY2023059-01).

Consent to Participate

The need for consent to participate was waived according to national legislation.

Data Availability Statement

The data is available on request by emailing the corresponding author.*