Effect of in situ coaching on cardiopulmonary resuscitation quality: a multicenter randomized simulation study

Graphical abstract

Highlights

• •CPR coaching achieved superior quality of chest compressions compared with defibrillator monitors and control groups.
• •Improvements were maintained even under muscular fatigue, with significantly better compression rate and chest recoil.
• •Defibrillator monitor feedback did not demonstrate significant benefit compared with unassisted controls.

Abstract

Background

A previous manikin-based simulation study among healthcare professionals developed a cardiopulmonary resuscitation (CPR) coaching method that significantly improved CPR quality compared with real-time audiovisual feedback provided by a defibrillator monitor. This study aimed to determine whether these findings could be replicated in a larger, multicenter trial involving first responders.

Methods

We conducted a multicenter, block-randomized trial among French firefighters. Participants were randomized into three groups (n = 36 each): CPR coaching, defibrillator monitor or control. Each participant performed three 2-min external chest compression sequences on an adult cardiac arrest simulator, which measured CPR quality, compression rate, chest recoil, and depth.

Results

A total of 108 participants were enrolled across two centers; one participant withdrew. In the second external chest compression sequence, the CPR coaching group achieved significantly higher CPR quality than the defibrillator monitor group (median 99% [interquartile range (IQR) 99–100] vs 99% [78–99]; p = 0.0022). This difference persisted in the third sequence (p = 0.0017). Compared with the control group, CPR quality in the CPR coaching group was significantly improved in both sequences (p < 0.001), whereas the defibrillator monitor group did not show a significant difference versus control.

Conclusion

The study confirmed the findings of our previous study with healthcare professionals, demonstrating the superiority of CPR coaching over defibrillator monitor feedback for improving CPR quality in a simulation-based setting with lay rescuers.

Introduction

Each year in France, nearly 50,000 individuals experience an out-of-hospital cardiac arrest (OHCA).¹ Across Europe, the 30-day survival rate ranges from 6% to 18%.2, 3 Early delivery of high-quality CPR is a key determinant of survival after OHCA, driving the search for strategies to optimize compression quality.4, 5 Real-time cardiopulmonary resuscitation (CPR) feedback delivered by automated external defibrillators or defibrillator monitors (DM) is among the most widely used adjuncts for supporting external chest compressions. International Liaison Committee on Resuscitation guidelines suggests using real-time audiovisual feedback and prompt devices during CPR in clinical practice as part of a comprehensive quality improvement program. Results from studies on real-time CPR feedback have been mixed in the clinical setting. The use of real-time audiovisual feedback and prompt devices in isolation is not recommended.⁶

On-scene, real-time guidance during resuscitation by an external professional, commonly referred to as CPR coaching, was the focus of a 2025 systematic review that identified only seven studies, with just one involving adult participants.⁷ The authors concluded that CPR coaching yielded promising improvements in CPR quality but warranted further investigation. The AMACE study⁸ strengthened this evidence by comparing an adult CPR coaching algorithm for external chest compressions with a defibrillator monitor in a manikin-based simulation involving emergency medical service healthcare professionals. Significant improvements in CPR quality and compression rate were observed in the CPR coaching group compared with the DM group. To build on these promising findings and assess the algorithm’s applicability among first responders trained in basic life support, this study, termed CADENCE (Comparison of Aids and Devices in External Cardiac Massage Enhancement), was conducted.

The primary objective was to compare CPR quality when using real-time feedback from a DM versus CPR guided by the CPR coaching method in a simulation-based setting with firefighters. Secondary objectives were to compare the two interventions under conditions of muscular fatigue and to benchmark their performance against an unassisted control group, thereby providing an estimate of each method’s crude efficacy.

Methods

Study design and participant population

An open-label, multicenter, randomized controlled superiority trial was conducted between September 2024 and February 2025 in two French fire and rescue centers located in Normandy. Participants were allocated to one of three parallel groups: (i) an unassisted control group, (ii) a group assisted by a defibrillator monitor (DM), and (iii) a group assisted by a human coach using the CPR coaching method developed in the AMACE study.⁸ Each participant performed three 2-min external chest compression (ECC) sessions on a manikin. The 2-min compression sessions were selected pragmatically to reflect standard field practice, where rescuer rotation typically occurs every two minutes. The first session (S1) was performed without assistance for all groups at study entry. This was followed by a 2-min rest period to allow partial recovery of rescuer performance and to simulate a team changeover during ECC. The second session (S2) was then performed with the assigned assistance device or method. Following S2, a brief 30-s rest period was introduced to limit recovery and maintain muscular fatigue while reflecting real-world CPR rotation practices before performing the third ECC session (S3) under the same conditions as S2. No changes were made to the original protocol after the trial commenced.

Ethics

The study was approved by the Local Committee for Ethics and Health Research of the University of Caen Normandy (ID# 5094). All participants received written and oral information on the study’s objectives and their rights before enrollment. Written informed consent was obtained prior to participation.

Inclusion criteria

Professional firefighters aged ≥18 years who regularly perform ECC were eligible. Participants had to be free of any physical or mental condition that could impair performance of the task. All participants were certified in basic life support in accordance with French civil security training guidelines. Participants were not involved in the design, conduct, or reporting of the study.

Randomization

Participants were manually randomized in a 1:1:1 ratio to one of three groups: the DM group, the CPR coaching group, and a control group (no assistance). Randomization was stratified by center and conducted in blocks of 12 participants. In each center, sets of 12 identical, sealed, opaque envelopes were prepared in advance, with four envelopes assigned to each study group. For each inclusion, participants randomly drew one envelope from the set. Investigators remained unaware of group allocation until the envelope was opened, ensuring allocation concealment.

Intervention

In the DM group, participants used the X Series® device (Zoll®, Chelmsford, Massachusetts, USA). This device employs an accelerometer placed under the rescuer’s hand to provide real-time feedback on CPR quality through visual display of key parameters (depth, rate, recoil) and delivers auditory prompts along with a metronome to guide compressions. The DM-assisted session was supervised by a study investigator. After completing S1, the firefighters received a demonstration of the X series device and its use. The investigator answered any questions before stepping back, and the rescuer was guided solely by the device during S2 and S3.

In the CPR coaching group, participants were coached using the algorithm developed in the AMACE study (Fig. 1). This algorithm, based on European Resuscitation Council guidelines,⁹ was designed by two study investigators and independently validated by four emergency physicians not involved in the trial. In collaboration with a university social psychology laboratory, the algorithm incorporated key elements (rescuer self-efficacy, sense of control, and perceived social support) to optimize the coach–rescuer relationship. The role of coach, defined according to Hunt et al.,¹⁰ was assigned to emergency physicians experienced in OHCA management. Coaching was supported by a metronome set at 110/min. If performance was suboptimal, the coach provided corrective feedback using the algorithm and predefined communication strategies. If performance was adequate, the coach offered regular encouragement while maintaining algorithm-guided monitoring without verbalizing unnecessary steps to avoid distracting the rescuer.

Fig. 1.

Coaching algorithm developed in the AMACE study.

(cm: centimeters, ECC: external chest compressions, min: minutes).

In the control group, participants received no assistance; the investigator simply indicated the start and end of each session.

Outcome measures

The study used Little Anne manikins (Laerdal®, Stavanger, Norway), widely employed in OHCA simulation research.11, 12, 13 CPR quality was assessed using the integrated Laerdal® QCPR software, which has been widely used in prior CPR simulation and resuscitation studies to provide both individual performance metrics and a composite overall quality score based on compression rate, depth, and recoil.11, 13, 14, 15, 16, 17, 18, 19, 20

The primary outcome was CPR quality, assessed using the QCPR composite score (QECC), a proprietary percentage score (100% = optimal) calculated in accordance with European Resuscitation Council recommendations, combining depth, rate, and recoil parameters. This score, while frequently used in practice, has not undergone formal validation studies. Secondary outcomes included the individual scores for depth, rate, and recoil, expressed as percentages. For each parameter (compression depth, rate, and recoil), the QCPR score represents the percentage of total compressions performed within the target ranges defined by current resuscitation guidelines. In accordance with European Resuscitation Council guidelines, target ranges were defined as 100–120 compressions/min for rate, ≥5 cm for depth, and complete chest recoil to baseline position; outcomes were expressed as the percentage of compressions within these ranges.

The primary objective was to compare QECC between the CPR coaching and the DM group in S2. The first secondary objective was to compare depth, rate, and recoil scores between CPR coaching and DM in S2. Additional secondary objectives included five intergroup comparisons: CPR coaching vs. DM in S3, CPR coaching vs. control in S2 and S3, and DM vs. control in S2 and S3.

Statistical analysis

No formal sample size calculation was performed a priori.

Categorical variables are reported as counts and percentages; continuous variables as medians with interquartile range (IQR: 1st–3rd quartiles). The primary comparison of QECC between CPR coaching and DM in S2 was performed using a Mann-Whitney test. 95% confidence intervals for the difference between several specific percentiles of the QECC across the two groups (25th, 15th and 10th percentile) were estimated using a bias-corrected and accelerated bootstrap resampling. A Mann-Whitney test was also used for each of the secondary intergroup comparisons, with p-values corrected for multiple comparisons using the Bonferroni–Holm method according to the number of outcomes tested. Analyses were conducted on an intention-to-treat basis. No interim analysis was planned. For one DM participant who withdrew during S1, a score of 100% was assigned to all four quality measures to apply the most conservative approach relative to the study hypotheses. All analyses were performed at the Biostatistics and Clinical Research Unit, University Hospital of Caen, using IBM SPSS Statistics v23. A two-sided p-value < 0.05 was considered statistically significant. Anonymized statistical datasets are available upon a reasonable request from the corresponding author.

Results

During the recruitment period, 108 participants were eligible, and all agreed to participate. They were evenly allocated to three groups of 36. One participant in the DM group was unable to complete the study due to work-related obligations. All other participants completed the three CPR sequences. The study flowchart is shown in Fig. 2.

Fig. 2.

Participant allocation and session flow of the CADENCE study.

(CPR coaching: cardiopulmonary resuscitation coaching, ECC: external chest compressions, DM: defibrillator monitor, min: minutes, s: seconds, S1: first ECC session, S2: second ECC session, S3: third ECC session. Withdrawal due to service requirements).

Population characteristics

The median age was 38.5 years [IQR 29.5–44.5] in the CPR coaching group, 38.0 years [IQR 29.0–47.0] in the DM group, and 40.5 years [IQR 36–46.0] in the control group. Men accounted for 88.9% of participants in the CPR coaching and the DM groups and 100% in the control group. The median duration of professional experience was 20 years [IQR 10–25] for CPR coaching, 19 years [IQR 12–25] for DM, and 21 years [IQR 14–27] for the control group. CPR quality scores in S1 were close to each other across the three study groups, except for compression rate, with a median much lower in the CPR coaching group (47% [IQR 4–87]) than in the DM group (90% [IQR 49–100]). Detailed baseline characteristics are provided in Table 1.

Table 1.

Participant characteristics.

	Coaching (n = 36)	DM (n = 36)	Control (n = 36)
Age (years)	38.5 [29.5–44.5]	38.0 [29.0–47.0]	40.5 [36.0–46.0]
Sex
Male	32 (88.9%)	32 (88.9%)	36 (100%)
Female	4 (11.1%)	4 (11.4%)	0 (0%)
Professional experience (Years)	20 [10–25]	19 [12–25]	21 [14–27]
ECC quality S1
QECC (%)	95 [88–99]	98 [88–99]	97 [69–99]
Depth (%)	100 [100–100]	100 [100–100]	100 [100–100]
Rate (%)	47 [4–87]	90 [49–100]	70 [4–95]
Full chest recoil (%)	94 [35–100]	87 [22–100]	100 [40–100]

Quantitative variables are expressed as medians with interquartile ranges. DM: defibrillator monitor; QECC: QCPR composite score; S1: first session without assistance).

Primary outcome

In S2, CPR quality, assessed using the QCPR composite score (QECC) was significantly higher in the CPR coaching group (median [IQR]: 99% [99–100] vs. 98.5% [77.5–99]; p = 0.0022). The median scores were very close to each other, but the first quartile value was markedly higher in the CPR coaching group (25th percentile difference: 21.5%, 95% confidence interval [CI]: 1–52%). The difference increased for lower percentiles: 47% (95% CI, 10–60%) for the 15th percentile and 55% (95% CI, 23–94%) for the 10th percentile, indicating a much narrower distribution, toward higher scores, in the CPR coaching group. Results for the primary outcome are summarized in Table 2, and Fig. 3 shows QECC performance across S1 and S2.

Table 2.

Primary and secondary outcome results of the CADENCE study.

	(1)	(2)	(3)	P-values*
	Coaching (n = 36)	DM (n = 36)	Control (n = 36)	(1)–(2)	(1)–(3)	(2)–(3)
CPR quality S2
QECC (%)	99 [99–100]	99 [78–99]	95 [66–99]	0.0022	<0.001	0.45
Depth (%)	100 [100–100]	100 [100–100]	100 [100–100]	0.32	0.62	0.67
Rate (%)	99 [97–100]	95 [58–100]	51 [1–98]	0.044	<0.001	0.054
Full chest recoil (%)	100 [95–100]	86 [25–100]	100 [26–100]	0.0048	0.32	0.52
CPR quality S3
QECC (%)	99 [99–100]	97 [78–99]	94 [60–99]	0.0017	<0.001	0.34
Depth (%)	100 [100–100]	100 [100–100]	100 [100–100]	0.36	0.24	0.11
Rate (%)	100 [99–100]	94 [48–100]	61 [1–99]	0.0047	<0.001	0.11
Full chest recoil (%)	100 [100–100]	57 [12–100]	100 [63–100]	<0.001	0.24	0.11

^*P-values for secondary outcomes were adjusted for multiple comparisons (CPR coaching: cardiopulmonary resuscitation coaching; DM: defibrillator monitor; ECC: external chest compressions; QECC: QCPR composite score; S2: second session with the assigned intervention; S3: third session with the assigned intervention).

Fig. 3.

Median CPR quality (QECC) values with interquartile ranges are shown for each group across the three sessions, illustrating sustained superiority of CPR coaching over DM and control.

CPR: cardiopulmonary resuscitation; ECC: external chest compressions; DM: defibrillator monitor; QECC: QCPR composite score; S1: session before intervention; S2: first session with intervention; S3: second session with intervention).

Secondary outcomes

In S2, the compression rate was significantly higher in the CPR coaching group compared with the DM group (median [IQR]: 99% [97–100] vs. 95% [58–100]; p = 0.044), as was complete chest recoil (100% [95–100] vs. 86% [25–100]; p = 0.0048). No significant difference was observed for depth (both medians 100% [100–100]; p = 0.32). In S3, the difference in QECC between the CPR coaching and the DM groups became more pronounced (CPR coaching: 99% [99–100] vs. DM: 97% [78–99]; p = 0.0017). Similarly, the compression rate (p = 0.005) and complete chest recoil (p < 0.001) remained significantly better in the CPR coaching group. Compared with the control group, CPR coaching demonstrated significant superiority in QECC (p < 0.001) and compression rate (p < 0.001) in both S2 and S3. Complete chest recoil was higher but not statistically different in CPR coaching versus control in S2 (p = 0.32) and S3 (p = 0.24). No differences were observed for depth in either session. No significant differences were found between the DM and control groups for QECC in S2 (p = 0.45) or S3 (p = 0.34), or for any other parameters in either session. Results for the secondary outcomes are summarized in Table 2. An overview of QECC performance across S1, S2, and S3 for the three groups is presented in Fig. 3.

Discussion

This study, conducted among firefighters in a simulation setting, demonstrated significantly better CPR quality, assessed using the QCPR composite score (QECC), with the AMACE-developed CPR coaching algorithm compared with a DM.⁸ Although statistically significant differences in the QECC composite score were observed between the CPR coaching and device monitoring groups in both S2 and S3, the clinical significance of these differences should be interpreted cautiously. The proprietary nature of the QECC algorithm limits direct clinical translation of small absolute score differences. Importantly, the observed effect was not driven by changes in median performance but rather by improvements among the lower percentiles of the CPR quality distribution, indicating that coaching primarily enhanced performance among participants with lower baseline CPR quality. This interpretation is supported by improvements observed in individual objective CPR metrics, particularly compression rate and complete chest recoil. These findings suggest that CPR coaching may be particularly valuable in preventing poor-quality compressions during prolonged resuscitation efforts, although further studies in real-world cardiac arrest settings are required. These effects became even more pronounced during CPR performed under muscular fatigue, as observed in S3. Compression rate and complete chest recoil are recognized as key determinants of CPR quality, as they enhance cardiac output, venous return, and coronary perfusion, ultimately improving patient survival.21, 22 Compared with an unassisted control group, CPR coaching significantly improved CPR quality (QECC) and compression rate in both S2 and S3, whereas DM use showed no significant benefit over control.

These findings are consistent with the AMACE pilot study,⁸ which compared the same CPR coaching algorithm with a DM in a manikin-based simulation involving 39 emergency medical service healthcare professionals. That study demonstrated significantly higher QECC (p = 0.010) and compression rate (p = 0.0012) in the CPR coaching group, while complete chest recoil was improved but not significantly (p = 0.080). The optimization of human factors is likely a major contributor to the positive outcomes observed with CPR coaching. Shortly before AMACE was published, Lauridsen et al. performed a systematic review on CPR coaching during CPR, concluding that the method was promising but required further investigation.⁷ Together, AMACE and CADENCE support those conclusions and position CPR coaching as a valuable method for optimizing OHCA care. Existing studies evaluating CPR quality optimization have primarily focused on training settings, automated audiovisual feedback, or dispatcher-assisted guidance, frequently using QCPR-derived performance metrics among students, laypersons, and healthcare professionals.11, 13, 18, 23 These investigations typically involved short standardized simulation scenarios and did not incorporate structured human coaching delivered on scene. Evidence specifically addressing real-time on-scene CPR coaching in adult resuscitation therefore remains extremely limited.

The CADENCE study extends prior work by evaluating a standardized human CPR coaching algorithm among adult first responders under progressive muscular fatigue conditions, using both composite and component CPR quality outcomes. Differences in participant profiles, exposure duration, guidance modality, and outcome metrics across studies likely contribute to variability in reported effects and may explain the performance distribution patterns observed in the present trial.

Currently, coaching is most commonly delivered by telephone. Multiple studies have validated its effectiveness in this format. However, the lack of visual control over the rescuer’s actions may hinder feedback and supervision. Telephone coaching may also be limited by technical factors (equipment availability, network coverage, device failure) and by environmental factors (noise pollution, distractions) that reduce its benefits.24, 25, 26, 27, 28 Finally, it does not exploit the full spectrum of human communication channels, including non-verbal cues. In contrast, the CPR coaching method proposed in AMACE addresses these limitations but requires an on-site coach, who is rarely present in the initial phase of resuscitation. CPR coaching could therefore complement telephone coaching, taking over from remote guidance initiated at OHCA onset, providing a continuum of supervision that could improve performance and outcomes.

In both AMACE and CADENCE, CPR coaching was delivered by emergency physicians. In France, medical teams may take up to 20 min to reach the scene,²⁹ compared with an average of 13 min for rescue crews.30, 31 Implementing CPR coaching by trained rescuers before medical team arrival may represent a key enhancement of the AMACE protocol.

Several limitations should be acknowledged. First, CPR sessions were limited to two minutes to reflect standard rotation practice in France; while this enhances ecological validity, it may limit extrapolation to prolonged continuous resuscitation scenarios and may have reduced detectable between-group differences. Second, muscular fatigue during S3 was not objectively quantified, and the short rest interval may not fully reproduce fatigue levels observed during prolonged continuous CPR. Third, the primary outcome relied on the QCPR composite score (QECC), which is based on a proprietary algorithm with limited transparency and no direct clinical correlate. However, its individual components (compression rate, depth, and recoil) were analyzed as secondary outcomes to enhance interpretability. In addition, the QCPR system reported only the percentage of compressions within predefined target ranges, thereby preventing analysis of mean absolute compression depth, rate, and recoil values. To further validate these findings, future studies using more transparent global metrics, such as the “overall excellent CPR” score,32, 33 would be of interest.

Finally, although simulation is a standard method for evaluating CPR quality improvement strategies, real-life resuscitation dynamics may differ, and the study protocol was not prospectively registered.

Conclusion

In this simulation-based trial, CPR coaching using the AMACE-developed algorithm demonstrated superior CPR quality among firefighters compared with device-based feedback. The CPR coaching approach appears cost-effective, simple, and adaptable across rescuers and operational contexts. These findings support further research to refine CPR coaching implementation, integrate it with dispatcher-assisted coaching, and evaluate its effectiveness in real-world out-of-hospital cardiac arrest outcomes.

Funding

This study did not receive any public or private funding.

Declaration of competing interest

The authors declare no conflicts of interest.