Virtual Reality for Cardiopulmonary Resuscitation Healthcare Professionals Training: A Systematic Review

Introduction Virtual reality (VR) is becoming increasingly popular to train health-care professionals (HCPs) to acquire and/or maintain cardiopulmonary resuscitation (CPR) basic or advanced skills. Aim To understand whether VR in CPR training or retraining courses can have benefits for patients (neonatal, pediatric, and adult), HCPs and health-care organizations as compared to traditional CPR training. Methods A systematic review (PROSPERO: CRD42023431768) following the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) guidelines. In June 2023, the PubMed, Cochrane Library, Scopus and Cumulative Index to Nursing and Allied Health Literature (CINAHL) databases were searched and included studies evaluated in their methodological quality with Joanna Briggs Institute checklists. Data were narratively summarized. Results Fifteen studies published between 2013 and 2023 with overall fair quality were included. No studies investigated patients’ outcomes. At the HCP level, the virtual learning environment was perceived to be engaging, realistic and facilitated the memorization of the procedures; however, limited decision-making, team building, psychological pressure and frenetic environment were underlined as disadvantages. Moreover, a general improvement in performance was reported in the use of the defibrillator and carrying out the chest compressions. At the organizational level, one study performed a cost/benefit evaluation in favor of VR as compared to traditional CPR training. Conclusions The use of VR for CPR training and retraining is in an early stage of development. Some benefits at the HCP level are promising. However, more research is needed with standardized approaches to ensure a progressive accumulation of the evidence and inform decisions regarding the best training methodology in this field. Supplementary Information The online version contains supplementary material available at 10.1007/s10916-024-02063-1.


Introduction
New digital technologies have involved every aspect of life with virtual reality (VR) as an innovative tool in training purposes [1].Although more than 100 definitions have been established to date [2], the term VR identifies various ways of simulating real situations using computers and the aid of specifically developed interfaces [3].Recently, VR has been included in an overarching term named 'extended reality' which includes augmented reality, mixed reality, and other immersive technologies [4].All these digital technologies have been introduced and tested in different training fields, cardiopulmonary resuscitation (CPR) included, underlining their potentialities in improving and/or maintaining skills [5].The traditional CPR-standardized face-to-face courses are still offered to provide skills that need complex retraining due to their decreasing over time; however, these traditional courses require physical, human and time resources not easily ensured by the care systems and health-care professionals (HCPs) who are immersed in complex clinical environments, requiring a continuing development of diversified competences and skills [6].Moreover, when the initial and retraining HCPs learning needs are not provided effectively, there is an increased risk of suboptimal care resulting in poor survival outcomes from cardiac arrest [5].Considering these

Search methods
The search was conducted in PubMed, Cochrane Library, Scopus, and Cumulative Index to Nursing and Allied Health Literature (CINAHL) databases, in June 2023.The keywords (Medical Subject Headings or free textterms) used were: "Virtual Reality"; "Healthcare Professional"; "Cardiopulmonary Resuscitation", in combination with the Boolean operators "AND" or "OR".The search strings are reported in the Supplementary Table 1.

Study selection
A total of 1042 records were identified, 335 from the databases and 707 from citation searching.First, studies from the databases were screened.Duplicates (n = 78) were eliminated.During the screening process at the title and abstract levels, 213 studies were excluded as not satisfying the inclusion criteria.Subsequently, full texts were read and 31 were not eligible.Four articles did not clearly define the population (e.g., HCPs, lay people or students); therefore, the corresponding authors were contacted and only two answered allowing to exclude the articles according to the inclusion/ exclusion criteria; the other two authors did not provide any answer, thus giving the impossibility to assess eligibility of the studies, these were excluded.The inclusion of one study was debated among researchers because it was not made explicit the extended reality used [14]; for this reason, it was excluded.
Articles from citation searching were also screened.From a total of 707 studies, 31 were excluded because duplicates.Of the remained, 676 were excluded because they did not satisfy the inclusion criteria.At the end, two articles were retrieved and included.
Overall, a total of 15 studies were considered (Fig. 1).The entire selection process involved two researchers (RT, AG) who worked independently; in case of disagreement a third researcher (SC) was consulted to reach consensus.

Quality assessment
The Joanna Briggs Institute checklists for randomized controlled trials [15], quasi-experimental [16], and cross-sectional studies [17], respectively were used.Some studies did not clearly indicate the design [18][19][20]: in the case of pre-post measures (even with a randomized sample), studies were considered quasi-experimental, while in the case of a single point evaluation, these were considered crosssectional.The assessment was performed independently by two researchers (RT, SC); in the case of disagreement a third researcher (AG) was consulted to reach consensus.

Data extraction and synthesis
A data extraction table was firstly piloted in three studies and no modifications were required.Then, the data extraction process was conducted by two researchers (RT, AG) populating the grid with the following data: author(s), publication year, country where the study has been conducted; study objective and design; type of CPR (basic or advanced), training or retraining; data collection year and duration; sampling, sample, setting; virtual simulation modality, devices; main features of the CPR training offered as comparator; tools for outcome measurement; and main findings.A third researcher (AP) verified this process and resolved disagreements.Then, studies were grouped according to the CPR target population.A narrative synthesis of their main characteristics and quality was first provided [21]; then, the VR CPR training and comparison modes were described in their main features; a summary of the outcomes measured, and their metrics were then provided.

Adults
Brzozowski et al. [22] 2022 ND The process of all the simulations required less than 25 min Sessions were held for a period of 1 week each month for 3 consecutive months.Between 48 and 79 time slots were available weekly 20-s presentation video followed by the scenario.This is followed by a 1-min video, a mutual feedback debriefing with the researchers and again a scenario Single group pre-post test Prior to the baseline virtual simulation, demographic and selfreport confidence surveys were completed online by participants The baseline assessment checklist was completed by the project team during each virtual simulation.During the repeated simulation, the project team completed the same checklist, and participants completed the self-confidence survey Self-confidence was measured with a tool minimally adapted from a previous study, evaluating 16 activities to be done in the first five minutes waiting for the ALS team, Likert scale from 1 (strongly disagree) to 4 (strongly agree) Initiation times for various CPR procedures (baseline virtual simulation vs second virtual simulation) Participants' confidence scores improved from baseline after VR; they were confident in: identifying an unresponsive and/or deteriorating patient (check for a pulse, assess for breathing) + 14.3% managing a patient in crisis prior to the rapid response team or code team arrival + 14.3% -using the defibrillator on their unit (applying pads, charging, shocking) + 28.6% participating in a group simulation + 28.6% giving SBAR report to the rapid response/ code team + 14.3% Participants maintained their confidence in the remaining six items Average resuscitation response times were all reduced from baseline to immediate post VR simulation The average time decrease was rhythm analysis 90.3 s defibrillation 87.The difference in performance between control and persuasive groups was not statistically significant; differences in performance between control and minimally persuasive groups was significant (p = 0.05 for PEA and p = 0.02 for VF/ VT) The pre-post comparison of performances of the groups showed that control (p = 0.017 for PEA, p = 0.01 for VF/ VT) and persuasive (p = 0.02 for PEA, p = 0.048 for VF/VT) groups improved their performances significantly, whereas the minimally persuasive group did not (p = 0.45 for PEA, p = 0.46 for VF/VT) (SD 4.4) years ND The XR-BLS simulator was created by employing a half-torso manikin in a virtual reality environment and using BLS guidelines A head-mounted display and hand-tracking device were used to perform chest compressions and ventilation and to enable the use of an AED in a virtual environment The primary input for learners was via the use of their own hands, head and voice Hand-tracking technology (Leap Motion tracker) allowed learners to be able to see 3D graphics of their real hands in action and to manipulate 3D objects (i.e., the AED) The manikin (Nurugo B100) was equipped with five sensors for vibration detection, depth and rate of the chest compression, chest compression location, opening of the airway and volume of the rescue breath The AI instructor was programmed to provide real-time feedback (e.g., quality of ventilations or chest compressions) The virtual characters had real human features Participants performed a 20-min learning mode with an AI instructor.Then, the training was performed in the practice mode, in accordance with real-life usage, followed by test mode

Not performed
After the usability test, an expert group survey was conducted with 8 items: 3 about the ease of use of XR-BLS, 3 about the delivery of training, and 2 about the AI instructor The evaluation of the XR-BLS simulator (joining the options strongly agree and agree) was: learning to operate the system is easy (100%) the possibility to see hands and manikin well in VR (100%) the personal interaction with the system clear and understandable (100%) to become proficient on BLS using this system (100%) the information was effective to complete the task and the scenario (100%) the instruction in VR tutorial were organized, clear and easy to understand (93.8%) it was easy to understand the AI instructor's explanations and instructions (100%) the interaction with the AI instructor is clear and understandable (87.5%) (2024) 48:50 50 Page 10 of 23 The participants were initially asked to view a 20-min tutorial video specific to each role This was followed by a 30-min session where they were asked to run through four to five virtual mock codes each approximately 5 min in length After the session, the participants were required to fill out a questionnaire about the ease of use and system usability For salivary cortisol levels, the attending/ fellow group had a 0.12 (IQR 0.09-0.17)μg/dL, whereas residents 0.19 (IQR 0.12-0.33)μg/dL The higher levels in residents were also statistically significant (+ 0.07 μg/dL, p = 0.001) A demographic survey was also completed Following the course, participants were encouraged to use their assigned digital intervention weekly and to engage in standard bag-and-mask skills practice with a manikin at the HBB practice corner set up at their facility.
At the overall level, the data collection duration lasted from two years with an additional six-month follow-up to a single day [27,33].The remaining studies collected data over two and over three months [18,22,29]; however, in 11 studies the duration was not specified.
The methodological quality of the studies (Supplementary tables 2, 3 and 4) was overall fair; in the six crosssectional studies confounding factors and strategies to deal with them were not identified [19,20,24,26,28,30]; in five RCTs the allocation to treatment groups and the outcome assessors were not blinded [25,27,29,31,32]; in two quasi-experimental studies the control group was not identified [18,22].

Intervention group
Almost all studies carried out the VR training using virtual 3D scenarios with the aid of additional accessories, such as a viewer [19].Some studies also employed a high-fidelity manikin connected to a computer [19,20,23,26,29,32], and compared the use of a very persuasive scenario with many components (visual aids, live instructions, and text messages) and a less persuasive one (only task completion messages at the end of these) [20].In some studies, instructors were allowed to independently modify scenarios in real time [19,24,28].
Five studies showed an initial tutorial regarding the VR simulation [18,22,26,27,31]; which varied from a 20-s presentation to 20 min with an AI instructor [22,26].The total duration of the training session was usually one hour [18,27], but the VR exposure time ranged from three or five minutes per scenario to approximately 30 min [20,32].

Comparison group
The training methods considered as comparators were mainly the traditional CPR using a full-body or half-torso low-fidelity manikin [25,27,29], or a high-fidelity manikin [24]; or two different versions of the VR simulator (with and without a persuasive features) [20].In the other two studies the comparators were not specified [23,30].Moreover, a pre-post comparison was performed with the same group [18,22]; while two groups, with different qualifications and experiences, were compared within the same VR simulation [28,31].

HCP experience
At the HCP experience level, VR scenarios were considered realistic (e.g., 80% [18]; 97% [19]).Moreover, around 72% of participants perceived the environment as engaging and facilitating the memorization of the procedures; 86% considered the VR simulations a useful training and agreed to recommend it to other colleagues [27].On the other hand, the high-fidelity simulation group reported similar satisfaction and utilization scores, but significantly better scores regarding the provided feedback (99%, Interquartile Range [IQR] 89-100) as compared to the VR group (79%, IQR 71-88, p < 0.01) [24].Using VR has led to greater commitment and increased knowledge acquired; furthermore, the gaming approach was perceived positively [30].However, some disadvantages emerged in using VR scenarios, including limited decision-making freedom, the lack of the possibility to build a team, as well as the absence of a real psychological pressure and a frenetic environment, as lived in the reality [28].In this regard, the level of stress during VR simulations among residents showed higher salivary cortisol levels than consultants (+ 0.07 μg/dL, 95%, CI = 0.03-0.12,p = 0.001) [31].During the same VR simulations, heart rates also differed between residents and consultants, with higher values for the former (85.9 hearth rate per minute, range 78-93.7)compared to the latter (78.9 hearth rate per minute, range 71.2-87.4)[31].

HCP knowledge
At the HCP knowledge level, studies comparing the number of correct questions through a pre-post test showed a statistically significant improvement [18,22,33].For example, the number of correct answers per HCP participants increased on average by 4.8 (95%, CI 3.4-6.2,p < 0.001) after the virtual simulation [18].

HCP performance
Performance measured as the degree of confidence with the automated external defibrillator (e.g.improvement in the time of positioning the pads, in rhythm analysis and defibrillation) and participation in a group simulation, was higher (+ 28.6% both) after virtual simulation [22].In contrast, 43% of participants after the traditional training were able to administer stacked shocks within one minute, whereas none of the participants in the VR group could so; moreover, the target sternotomy time was achieved in 100% of procedures in the control group and in 83% in the VR group [27].However, in the VR group the total number of errors was lower compared to the control group (11 vs 15) [27].The more previous training in CPR possessed by a participant the greater the percentage of correct compressions obtained in a VR CPR session (F-Ratio 14.95; p < 0.001) [23].A total of 88% of participants using VR achieved a correct compression rate and 65% a positive result in the depth of compression applied to the manikin [19].Better retention of skills in bag and mask ventilation at six months (-15% VR group, p = 0.10; -21% video group, p < 0.01; -27% control group, p = 0.001) was documented [33].
An improvement was also found when comparing the pre-post performances of the low-fidelity manikin group (p = 0.017 for pulseless electrical activity, PEA, p = 0.01 for ventricular fibrillation/ventricular tachycardia, VF/VT) and the persuasive VR group (p = 0.02 for PEA, p = 0.048 for VF/VT) but not for the minimally persuasive VR group (p = 0.45 for PEA, p = 0.46 for VF/VT) [25].Different, participants in the group with the less persuasive scenario reported a perception of greater usability (p = 0.0944) and ease in interfacing with the simulation compared to the very persuasive scenario (p = 0.0813) [20]; however, the results were not statistically significant.The group with live feedback on the quality of chest compressions achieved better performance in both frequency and depth compared to the non-feedback group (35.78% vs 7.27%, p < 0.001) [29].It was also seen that the chest compressions rate of the more experts was not influenced by the feedback, while the performances of the novices improved with the use of the feedback (112.07,SD 3.70 vs 98.39, SD 29.23, p = 0.001) [29].

Economic Impact
At the organizational level, four studies mentioned the cost-saving opportunities that VR could have in the future, despite the initial expense for the purchase of software (as licenses) and materials (as computers, sensors) [18,24,28,31].However, only one calculated the economic impact of using VR compared to the traditional method, highlighting a significant saving of 83% for a single learner for four training sessions (132.29 dollars for VR versus 772.00 dollars for the traditional method) in terms of human resources, classrooms as well as time spent in travelling [24].

Study characteristics
The research in the field of VR CPR has been established in the last ten years, with 15 studies, including 1,035 total participants and conducted in almost all continents, suggesting that these educational strategies can be ubiquitarian and considered also by lower-income countries.As documented in a recent systematic review, VR CPR training among adult laypersons has instead been investigated by only six studies involving a total of 731 participants between 2017 and 2021 [10].
Measuring the effectiveness, the feasibility and the utility of VR CPR has been the main intents of studies available; therefore, this research field seems to be focused around two main lines.The first is aimed at developing and introducing VR into CPR training, and the second in measuring its outcomes, both in basic and advanced CPR skills, mainly with regards to adult patients, involving HCPs in hospital settings.This substantial differences in research approaches, divided between educational impact [31], and those of digital devices design, development and piloting [20], may explain some missing elements that have emerged in the study methodology quality assessment as: study duration, mainly short or not reported; limited sample size, from a few HCPs to a maximum of 274; lack of description of the control group, or no control group established (four studies); and the poor description of the professional background, baseline competences and/or participants' experiences.The lack of description of the HCPs expertise, should be considered as an important lack in assessing whether the outcomes are purely associated to the VR CPR or also to prior experience.Overall, studies available seem to express a research field in its early stage.Guidelines based on broad consensus regarding the elements to document in future studies to harmonize their reporting and ensure that some essential data are described (e.g., VR CPR as a refresher or first training), by differentiating those aimed at piloting the digital technologies and those at measuring their effectiveness, may increase transparency, comparability and the accumulation of the evidence produced.

Intervention and comparison group training methods
Five main considerations may be derived from VR devices and training methods documented to date in the available literature.
Second, while some studies have been based on relatively simple technologies [20,29], others have implied several and complex technologies, from software created for 3D gaming to sensors connected to a computer [18,24,30].On the one hand, this great heterogeneity suggests an ample creativity and innovation; on the other, it prevents comparison and conclusions regarding the outcomes.Third, interactivity as the possible way to provide (or not) real-time feedback on participants' performance (e.g., depth and frequency of chest compressions) and the likelihood of changing the scenario suggest that some studies embraced the main characteristics of VR, i.e. that it includes a totally immersive environment in which the person is completely isolated from the external setting and transported into a parallel reality, reproduced with the use of additional devices, such as visors or sensory gloves [19,31].
Fourth, in all studies, the sections are short in duration, sometimes providing brief tutorials to familiarize participants with the tasks; in all cases, from the initiation to the end of the section, the CPR training seems to save time, and this may benefit the HCPs, the health-care organization and the continuing education agencies to address other learning needs.Regarding the comparisons, a wide range of training methods have been used: alongside efforts in standardizing the description of the comparator used in the studies, comparing the effectiveness of VR CPR with the most traditional training method used in daily practice may help in accumulating evidence in this field.

VR CPR outcomes
Different outcomes have been measured to date, employing several measures both subjective (e.g., self-reported participants' satisfaction [24]) and objective (e.g., sensor for automatic detection of participants' performance [19]), preventing the production of a cumulative synthesis of the evidence available.First, the substantial infancy of this research field may have prevented any evaluation on patient outcomes, requiring complex study designs and long followups.Also, in the review concerning lay people, a few studies have investigated the association of the educational interventions with patients' outcomes after cardiac arrest, both in the short and long term [10].However, given that the main aims of the CPR training, regardless of the methods, is to improve health outcomes, this lack should be addressed in the future to assess the VR role in this training field.
Second, outcomes most evaluated are those at the HCP level.With this regards, Kirkpatrick's framework has established the evaluation system of HCP educational programs as categorized into four levels, from the first and easier to assess (reactions) to the following and more complex (learning, behavioral and results) [34].In the context of VR CPR training, the first have been documented in 13 studies [18-20, 22-27, 29, 31-33], reporting a good level of usability, appreciation, and satisfaction.One study underlined that the simulation failed to reproduce the characteristics of the real emergency scenario [28].
Learning outcomes, the second level of the Kirkpatrick's framework [34], have been measured in all studies.Knowledge has improved, along with some evidence regarding better memory retainment of the acquired practices [18].With the traditional CPR education, the decline in theoretical notions and practical skills occurs just two months after the end of the course, reaching a peak at six months [35]; in contrast, VR training has shown that knowledge and skills at one, three, and six months remain higher [33].Moreover, some better performances have also been reported in terms of the decreased time needed to assess the heart rhythm and use of the defibrillator, which can be particularly useful in rare conditions, such as cardiac arrest in children [36].Notably, two out of three studies concerning neonatal emergencies were conducted in Africa [32,33], where the neonatal mortality rate is extremely high suggesting that VR methods of training may have a potential role in some conditions characterized by the rare occurrence of the phenomenon in the real world, issues regarding geographic accessibility, and underdeveloped economic conditions [37].
No studies investigated behavioral modifications [34], such as the extent to which the trainees apply the learning and change their behavior in real practice, immediately or months after the training.Conducting prospective studies monitoring the behavioral modification of participant HCPs after VR CPR training over time may be challenging, time-consuming and influenced by several confounding factors (e.g., having attended other training initiatives [38]).Most training interventions are evaluated only in terms of theoretical-practical knowledge achieved immediately at the end, and not in terms of changes in the medium or long term [39,40].In addition, only a few health-care facilities and universities use VR, and this may limit the likelihood of conducting such evaluations [41].
Alongside the lacks in behaviors modification, VR CPR training has been subjected to a little scrutiny regarding its capacity to develop decision making and team-building skills (e.g., team training and interaction with the team) [28].Both skills have been described as fundamental in the CPR practice [41], and this limitation might affect the real-world applicability of VR in this field.Involving more participants within the same VR scenario and promoting their interaction over that between the HCPs and the VR software may be useful [28]; reproducing the chaotic environment typical lived during the emergency situations, may also increase the reality of the VR scenarios and promote the transferability of the competences gained in the real world.
Concerning the impact of the results on the health-care organization, no studies have assessed this association with VR CPR training also in this case likely due to the complexity of these investigations [42,43].Despite some studies reported cost-saving opportunities [18,28,31], only one calculated the economic impact [24].Understanding whether VR can have benefits in terms of efficiency is essential to promote and convince of the need of an initial investment that may have important long-term cost-saving effects [44].Producing more evidence may also inform the long-term sustainability of investments in these new technologies.

Limitations
This systematic review has several limitations.Despite the attempt to conduct an accurate systematic review by registering in advance the protocol, and strictly following the PRISMA guidelines [11], selection biases may have occurred.Moreover, to provide a broader picture on all studies available all retrieved articles have been included despite their methodological quality.In addition, the background of researchers involved in the analysis of the studies, although experts in CPR and research methodologies, may have influenced the data extraction by missing some aspects regarding the technologies used.

Conclusions
The research in the field of VR CPR training is recent and conducted in several countries, mainly focused around two aims: developing the VR technologies and methods of training; and measuring the VR CPR effectiveness or practical implementation.Studies available concern mainly adult CPR training in hospital settings, with mixed voluntary HCPs; they are not-always specified as training or retraining and limited in the sample size.Outcomes have been assessed mainly at the HCP level, investigating reactions, and effects on knowledge and performance.
At the overall level, the evidence available regarding VR CPR training is promising: however, robust studies are required to overcome the methodological gaps in this research field.Guidelines establishing the essential data that should be communicated while reporting studies are recommended.Moreover, the effectiveness of VR CPR training research on clinical outcomes (by considering also pediatric patients) should be investigated; in addition, describing changes in the performances and behavior as manifested in the management of real-world cases after having received the VR CPR training is a priority.In this context, designing more complex studies by involving several HCPs in the same scenario, connected remotely, and stimulating interactions and strategies to also investigate their decision-making processes, may be important.All these evaluations may expand the evidence available and inform decision-makers regarding long-term investments and sustainability of the VR CPR training.

Fig. 1
Fig. 1 Flow diagram of the study selection process -Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) [11] Legend: CINAHL, Cumulative Index to Nursing and

Table 1
Characteristics of the included studies

Table 1 (continued) Author Publication year Country Objective Design Basic/advanced CPR Training/retraining Data collection year, duration Sampling Sample Setting Virtual simulation modality Devices Comparison Tools for outcome measurement Main results
3 s

Table 1 (
continued) levels (role, interface, feedback and server), six diversified roles and three interface components (visual, auditory and touch) using systems called Team Speak and Noving Falcon Each team was randomly assigned to one of the three treatment groups: control (traditional ACLS training -low-fidelity manikin), persuasive (VR ACLS training with comprehensive feedback components), or minimally persuasive (VR ACLS training with limited feedback components) Participants were randomly assigned to one of the six ACLS roles: compressor, medicator, defibrillator, airway manager, respirator and leader.Each role was associated with performing a specific set of tasks The study phases were 1) administration of a demographic questionnaire 2) pre-test to assess baseline 3) didactic training (25-min presentation of ACLS protocols with audio recording) 4) intervention groups 5) post-test evaluation 6) final questionnaire about the experience For VR experience, the feedback module involves the task of providing visual (including textual) and auditory feedback to the users during and after a training session