This study is a systematic review and meta-analysis following the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) 2020 guidelines [33], as shown in Multimedia Appendix 1, and the Cochrane Handbook for Systematic Reviews of Interventions [34]. The study protocol was developed by the research team and registered in PROSPERO (registration CRD42021240953). The protocol can be accessed through the PROSPERO website. No ethics approval was required.

Searches were conducted in MEDLINE (through PubMed), Web of Science, and Scopus, including publications between January 2015 and December 2023. The search strategy and query were designed to include all relevant publications considering the established eligibility criteria. The search query is presented in Textbox 1 and Multimedia Appendix 2 and includes keywords based on three categories: (1) CPR, (2) feedback devices, and (3) outcomes. Each category considered several keywords that were searched in the publications’ titles and abstracts. The search strategy was adapted to all included databases. The reference lists of relevant systematic reviews were screened to identify additional studies. Duplicates were removed using Mendeley Reference Manager and, later, verified using CADIMA software (Julius Kühn-Institut) [35].

The PICOS (participants, intervention, control, outcomes, and studies) framework was used to support the identification of the eligibility criteria.

Studies that included participants performing CPR in real patients were excluded. Studies focused on team CPR performance were excluded. Animal studies, case reports, conference abstracts, reviews, trial protocols, and studies with incomplete or missing data were excluded. Publications prior to 2015 were excluded to minimize bias from outdated CPR guidelines and to ensure that the technology in the included studies provides corrective feedback, thus reducing heterogeneity and enhancing the reliability of the findings. No country or language restrictions were considered.

The study selection process was executed using CADIMA software [35], which facilitated effective management and tracking of the selection process as well as analysis of interreviewer agreement.

Initially, a pilot analysis was conducted by 2 reviewers (AN and IJ), who independently assessed 100 abstracts. This preliminary stage aimed to evaluate the strength of agreement between reviewers and to identify and adjust any significant discrepancies in their assessment. The interrater reliability was quantified using the Cohen's κ statistic.

After refining the analysis approach, as informed by the initial pilot analysis, the reviewers transitioned to the screening phase. During this phase, titles and abstracts underwent an independent screening process conducted by the same 2 reviewers (AN and IJ). This step was crucial to ensure a thorough and unbiased selection process, and only relevant studies were considered for full-text review.

The next step was the inclusion phase, where selected papers were subjected to a comprehensive full-text analysis. This phase was expanded to include a third reviewer (CSC) to bring additional perspective and expertise, particularly in cases of initial disagreement. All 3 reviewers (AN, IJ, and CSC) independently analyzed the full texts.

In both the screening and inclusion phases, differences in opinions were resolved through in-depth discussions until a consensus was reached. During the inclusion phase, each reviewer meticulously documented the reasons for excluding papers. This step was critical for ensuring transparency and accountability in the selection process.

Data extraction was performed by 2 reviewers (IJ and AN) using a standardized data extraction form. CPR quality parameters were considered, namely the overall chest compression quality score. Additional parameters were extracted, including compression depth, compression frequency, chest wall recoil, and hand positioning.

Extracted data included authors, year of publication, country, study design, type of participants (laypeople or health care professionals), type of intervention, type of control, feedback device, number of participants in each group, training duration, CPR quality overall score (%), and quality of each parameter of chest compressions (%).

The quality of the selected studies was assessed using the Cochrane risk-of-bias tool for randomized trials (RoB 2) [36] by 2 independent reviewers (AN and IJ). RoB 2 tool evaluates each study using 5 different criteria: randomization process, deviations from intended interventions, missing outcome data, measurement of the outcome, and selection of the reported result. For each item, the risk was marked as low, some concerns, and high. Uncertainties and disagreements were addressed and resolved through consensus. Studies were not blinded regarding authors, institutions, and journals. Publication bias was assessed through a funnel plot.

A meta-analysis and a narrative synthesis were conducted to identify associations and explore heterogeneity in the selected studies. Data were analyzed using R (version 4.3.2; R Foundation for Statistical Computing). In the meta-analysis, the mean and SD of the data were primarily used. In cases where studies presented median and IQR, these were converted into mean and SD under the assumption of normally distributed data [37]. This assumption is based on the statistical principle that, in a normally distributed sample, the mean is an approximate measure of the median, and the SD can be estimated by dividing the IQR by 1.35 [37,38]. If the mean and SD could not be estimated, those were excluded from the meta-analysis. The effect size was calculated using the mean differences with 95% CIs. The random-effects model was used because of the variation in study characteristics. Heterogeneity was studied using I² statistic, with values below 40% representing low heterogeneity [34]. Several studies had a multiple-arm RCT design; in those cases, the relevant groups (intervention and control) were included in the pairwise comparison of intervention groups. Studies with distinct groups of adult and pediatric CPR training were included with multiple intervention groups.

Sensitivity analysis was conducted through subgroup analysis to identify the sources of heterogeneity and measure the effects within each subgroup (eg, acquisition vs retention). This analysis also determined if the performance scores varied across the subgroups. The subgroup analyses explored potential effect modifiers, including skills acquisition (assessed immediately after training), training duration (≤30 vs >30 minutes), and time elapsed after training (3 vs 9-12 months). Other subgroups were considered (eg, target audience—health care professionals vs laypeople) but were not performed due to the limited or absent number of trials available. Subgroups with enough data were used for meta-analysis.

A narrative synthesis was performed for all included studies, including those excluded from meta-analysis, following the SWiM (Synthesis Without Meta-Analysis) reporting guidelines [39]. The studies were grouped by the training-assessment interval, clearly distinguishing between skills acquisition and retention over time. When available, the main outcome selected was chest compression score. The certainty of evidence was evaluated using the GRADE (Grading of Recommendations Assessment, Development and Evaluation) approach [40].

The studies’ selection process is represented in the PRISMA flow diagram [33] (Figure 1). The initial search retrieved 10,095 results, of which 4427 were duplicates. Of the remaining 5668 unique publications, 5619 were excluded after screening the title and abstract. The remaining 49 publications were screened for a full review, resulting in the 20 studies included. All included studies reported approval by an ethical committee.

Cohen's κ coefficient of agreement between the reviewers was 0.49 at the pilot screening stage, representing a fair agreement [34]. Throughout the full screening stage, the Cohen’s κ coefficient improved to 0.62, representing a good agreement. The overall percentage of agreement was 99%.

Table 1 presents the main characteristics of the 20 included studies. Studies data were collected from 15 distinct countries. The study population is diverse, including laypeople [23,41-43], health care professionals [44-46], and lifeguards [47], among others. The sample size is heterogeneous. Although the majority of the included studies used Laerdal QCPR technology in the intervention group, the training methodologies vary substantially, with considerable differences in training duration, elapsed time to assessment, and assessment format (Table 2).

Table 2 details relevant characteristics regarding study design and outcomes. Training methodologies included short single training sessions [23,41,42,44,46,47,49,50,52,53,55,56,58,59], short recurrent practice over time [45], and long single training sessions [48,51,54,57]. The educational adjuncts in all studies were based on corrective feedback devices, although some also included video-based instructions or allied corrective feedback devices with instructor feedback [42,43,46,48,51,55,58,59]. Regarding training-assessment interval, skills acquisition assessment was not comprised in all studies, and assessment intervals for skills retention span from few weeks to several months. Skills acquisition was assessed by 14 studies [41-44,46,48-53,57-59]. In total, 6 studies presented the assessment of skills retention at 3 months [42,44,45,49,57,59], and 3 studies [45,49,59] included an assessment at 9 to 12 months. Assessment duration also varied among the studies, with most considering 2 minutes of CPR, with only 4 studies using a different assessment duration [41,47,53,55]. Assessment exercises ranged from continuous chest compression [23,41,42,45,49,56-59] to a defined number of CPR cycles [44,55]. The outcomes were mostly presented using mean and SD and median and IQR for chest compression–related parameters and overall score. Additional specifications of the included studies can be found in Multimedia Appendix 3.

The risk of bias was assessed in all 20 RCTs. Among the studies included, 40% (n=8) presented a high risk of bias [42,43,46,48,49,52,55,57], and 50% (n=10) presented a low risk of bias [23,41,44,45,50,51,53,56,58,59]. Table 3 illustrates the risk of bias in each domain (D1-D5), based on RoB 2 [36], along with the overall risk of bias.

Across all domains, the randomization process was the primary reason for the risk of bias due to no information or absence of random allocation sequence or no information or absence of allocation sequence concealment. Studies were assessed to have an overall “high risk” of bias if there were at least 1 “high risk” in any of the criteria (D1-D5).

The skills acquisition subgroup included 8 studies [41,46,48,51-53,57,58] involving 1477 participants. Meta-analysis results revealed high heterogeneity (I²>75%, as illustrated in Figure 2). All studies presented a positive mean difference for chest compression score, with an average of 17.37%, favoring the use of corrective feedback devices.

Within the skills acquisition subgroup, further analysis was conducted to examine the duration of training. Five studies, involving a total of 1225 participants, focused on training sessions shorter than 30 minutes [41,46,52,53,58]. These studies uniformly reported a favorable mean difference in chest compression score, with an average improvement of 7.21%, attributed to the use of corrective feedback devices (as shown in Figure 3A). These studies exhibited low heterogeneity (I²=0%), underscoring the consistency of the positive effects observed. Three studies, involving 252 participants, reported on training sessions with more than 30 minutes [48,51,57] with high heterogeneity (I²>75%). Despite the observed variability, all studies consistently presented a substantial mean improvement in chest compression score, averaging 38.91% (as shown in Figure 3B), underscoring the impact of using corrective feedback devices in longer training sessions.

Skills retention at 3 months was examined in 3 studies [45,49,59], involving 366 participants, with low heterogeneity (I²=35%). These studies collectively showed a positive mean difference in depth scores, averaging 17.99%, favoring the use of corrective feedback devices (as illustrated in Figure 4A). For skills retention between 9 and 12 months, 3 studies [45,49,59] were analyzed, revealing low heterogeneity (I²=33%). All studies reported a positive mean difference in depth score, with an average improvement of 7.84%, favoring the use of corrective feedback devices (as shown in Figure 4B).

Other meta-analyses were conducted to investigate potential sources of heterogeneity, including a subgroup analysis based on risk of bias and study population; however, no significant results were observed. This meta-analysis did not assess the quality of chest recoil, hand positioning, and compression frequency due to insufficient data or challenges in extracting mean and SD values. Certainty of evidence was assessed using the Grading of Recommendations Assessment, Development and Evaluation approach, revealing a very low to low certainty (as shown in Table 4). Publication bias was not detected through the analysis of funnel plots (Multimedia Appendix 4).

This paper critically evaluates the effectiveness of corrective feedback devices in CPR training for both laypeople and health care professionals. While several reviews support the use of feedback devices for CPR training [29-32], some studies report inaccuracies related to the use of these devices, namely an overestimation of chest compression depth [60,61]. To the best of our knowledge, there is no recent and thorough systematic review and meta-analysis specifically evaluating the effects of CPR training with corrective feedback devices on the skills acquisition and retention among both laypeople and health care professionals. Existing reviews often rely on outdated guidelines [29,30], or encompassed a broader scope, such as the use of feedback devices in real clinical settings [29-31].

Notably, reviews prior to 2015 often did not differentiate between devices providing guidance and those offering corrective feedback, a distinction introduced in 2015 CPR guidelines [28]. One of the earliest systematic reviews showed evidence favoring the use of CPR feedback devices [29]; however, it included studies based on outdated guidelines and “prompt” devices.

A more recent review also reported that feedback devices promote better chest compressions [30]. However, this review included nonrandomized trials, studies based on outdated guidelines, and studies performed in real settings (human studies), increasing the heterogeneity and potentially reducing the quality of the review. Subsequent reviews [31,32] recommended the use of feedback devices but did not include a meta-analysis [31] and focused on specific groups—health care professionals [31] and laypeople [32]. Additionally, these reviews included studies using devices that lacked corrective feedback.

This systematic review and meta-analysis aim to bridge the identified gaps by examining the impact of corrective feedback devices on the quality of chest compressions during CPR training across diverse groups. This review exclusively focuses on studies using corrective feedback devices for skills acquisition and retention, within simulated or training environments, targeting both laypeople and health care professionals. The rigorous search strategy and selection process ensured a comprehensive and inclusive examination, capturing all pertinent studies within the scope of this review.

The risk of bias assessment revealed a general trend of overestimated effects due to methodological flaws, with few studies adhering to CONSORT (Consolidated Standards of Reporting Trials) reporting guidelines [62]. Poorly designed studies may yield larger effect estimates when compared to those with rigorous methodologies, potentially compromising the robustness and reliability of the findings. Several studies included in the meta-analysis were noted to have some concerns or high risk of bias, which may influence the overall findings. These were not excluded to maintain an adequate number of studies under analysis, considering that all of those are RCT, which, per se, represent a superior level of evidence [63].

The majority of studies included in this review featured intervention groups that conducted training autonomously or through self-training without direct instructor involvement. However, some studies did incorporate a degree of instructor mediation within the intervention groups. In these cases, the corrective feedback device remained the central component of the intervention, with the instructors’ roles primarily confined to presenting the metrics generated by these devices.

The presence or absence of an instructor in the intervention groups was not considered a criterion for inclusion in this review, which may have introduced a degree of heterogeneity. The varied outcomes and measurement approaches revealed a notable absence of standardized research protocols in this field. Specifically, for identical chest compression parameters, multiple approaches for measuring outcomes were identified. Adoption of standardized protocols and assessments in future research could greatly contribute to accurately extract intervention effects, allowing a more reliable and precise comparison of results.

The conducted meta-analysis revealed significant heterogeneity. Several factors may have contributed to this variability: the quality of the studies included; the variable protocol designs and methodologies used across these studies, including variations in training methods and assessment procedures; the diversity of the outcomes measured; the presence of nonreporting bias, where studies showing nonsignificant or minimal effect sizes might be underreported; and intrinsic heterogeneity due to a wide range of variables identified as relevant to the analysis.

Despite significant heterogeneity among some study results, this meta-analysis provides insights into the positive impact of corrective feedback devices on CPR training. As illustrated in Figure 2, there is an overall positive mean difference between the control and intervention groups, suggesting a positive impact of corrective feedback devices in enhancing skills acquisition and improving the quality of chest compressions. Notably, the duration of training plays a critical role in this improvement. Shorter training sessions show a reduced mean difference in chest compression scores between groups, indicating that extended exposure to corrective feedback devices yields better training outcomes. This observation is further supported by the consistency of results across studies with training durations of less than 30 minutes, which exhibit low heterogeneity.

In the analysis of skills retention over time, the depth of chest compressions emerged as the only parameter suitable for inclusion in the meta-analysis. Examining this specific parameter revels that the difference between the control group and the intervention group considerably decreases, when comparing data from the 3-month and 9- to 12-month subgroups, accompanied by an increase in heterogeneity (as illustrated in Figure 4). This trend suggests that, while the immediate benefits of using corrective feedback devices to enhance the quality of chest compression depth are clear, these benefits tend to converge over time, leading to less pronounced differences between groups and greater variability in study findings.

The maximum follow-up time analyzed in this review was 12 months due to the lack of reliable studies with longer follow-up periods. More robust studies with extended follow-up times are needed, especially considering that the recommended interval for recertification is 24 months. The studies with health care professionals [44-46] were too few to allow a comparison with studies focusing on laypeople; however, the results suggest a similar impact of corrective feedback devices in both groups. Future work could focus on exploring the differences in CPR skills acquisition and retention between these groups.

Regarding the corrective feedback devices used, the majority were Laerdal products, including full-body simulators or torsos. It was unclear whether these devices provided real-time feedback or presented a summary of the CPR quality at the end of each training session. While real-time feedback could enable immediate correction of maneuvers, it might also serve as a distraction during training sessions. This analysis, to our knowledge, has not been studied and could be an important topic for future research.

The majority of the studies included in this review revealed that the use of corrective feedback devices during CPR training led to significant enhancements in 1 or more key chest compression parameters. Although the improvements across all parameters were not uniform, the evidence strongly supports that corrective feedback devices contribute to an overall improvement in CPR skills acquisition among both laypeople and health care professionals, especially when exposed to longer training sessions. Longitudinal analysis of skills retention indicates that the long-term benefits become less distinct, especially after 3 months. These results are consistent with the findings from a previous systematic review [32].

The contribution of corrective feedback devices is significant not only for more objective and standardized training but also for enabling more frequent practice, without being dependent on the presence of an instructor. This facilitates continuous improvement and broader preparation for performing CPR.

This review is not without limitations. The high heterogeneity, variations in the study designs, and generally low quality of the included studies challenge the validity and generalization of conclusions. Additionally, this review only included RCTs, which may hinder relevant conclusions resulting from other designs. The exclusion of studies published before 2015 may have also omitted relevant data. Moreover, the use of different tools across studies to evaluate training outcomes introduces further variability, potentially affecting the judgment of findings.

This review supports the use of corrective feedback devices in CPR training as a beneficial complement or surrogate to traditional training methods (instructor-based feedback). Corrective feedback devices offer objective insights that can enhance immediate skills acquisition and may contribute to long-term retention. However, these findings must be interpreted with caution due to the significant variation in study designs and outcomes, along with the overall low quality of existent studies, highlighting the urgent need for well-designed, high-quality studies with standardized protocols.