Retention of CPR skills learnt in a brief educational video with short-time self-training versus mastery learning among lay responders in Thailand: an assessment of a training system Authors: Kiattichai Daorattanachai, Intanon Imsuwan, Winchana Srivilaithon, Kumpol Kornthatchapong, Pornpawee Wongpophon Published: 2025 ABSTRACT AbstractABSTRACTObjectivesTo evaluate the effectiveness of short-duration self-learning (SSL) and mastery learning (ML) strategies on the acquisition and 6-month retention of cardiopulmonary resuscitation (CPR) skills and basic life support (BLS) knowledge among hospital administrative staff in Thailand.DesignComparative assessment of two CPR training strategies.SettingTertiary-level academic institution in Thailand.ParticipantsA total of 163 hospital administrative staff aged ≥18 years without prior BLS certification were enrolled and randomly allocated to either the SSL group (n=82) or the ML group (n=81). Participants with previous CPR certification, medical limitations preventing chest compressions or those unwilling to attend the 6-month retention test were excluded. 71 participants in each group completed the 6-month follow-up.InterventionsAll participants viewed an 8 min self-directed instructional video on BLS and subsequently performed CPR practice using real-time feedback manikins. The SSL group practised independently for up to five 1 min sessions, while the ML group practised until achieving a predefined mastery threshold (QCPR (Quality Cardiopulmonary Resuscitation) score >90%).Primary and secondary outcome measuresThe primary outcome was CPR skill retention at 6 months, measured by QCPR overall performance score. Secondary outcomes included observational BLS performance (scene safety, responsiveness, emergency activation, breathing assessment and correct hand placement), knowledge retention scores and the number of attempts required to achieve mastery in the ML group.ResultsBoth SSL and ML groups demonstrated significant improvement in CPR performance and BLS knowledge immediately after training (p<0.001). At 6 months, no statistically significant differences were found between groups in CPR skill retention (mean difference 0.7%, 95% CI −1.9% to 3.3%, p=0.62) or BLS knowledge retention (mean difference 0.2, 95% CI −0.4 to 0.8, p=0.51). Female participants required more attempts to achieve mastery compared with males (mean 3.0 vs 2.1, p=0.007).ConclusionsBoth SSL and ML effectively enhanced CPR performance and knowledge retention among non-medical hospital staff. SSL offers a practical, resource-efficient and scalable approach for CPR training in large-scale community or institutional settings. Future studies should investigate the cost-effectiveness and optimal refresher frequency for sustaining CPR competency.Trial registration numberTCTR20210521003.Keywords: Cardiopulmonary Resuscitation, Out-of-Hospital Cardiac Arrest, MEDICAL EDUCATION & TRAINING Strengths and limitations of this study Comparative assessment design allowed evaluation of two training strategies. Comparative assessment design allowed evaluation of two training strategies. Standardised video-based instruction ensured consistent content delivery. Standardised video-based instruction ensured consistent content delivery. Real-time feedback manikins provided objective performance data. Real-time feedback manikins provided objective performance data. Six-month follow-up enabled evaluation of long-term retention. Six-month follow-up enabled evaluation of long-term retention. Single-centre design with hospital staff participants may limit generalisability to the general public. Single-centre design with hospital staff participants may limit generalisability to the general public. Introduction Out-of-hospital cardiac arrest is a major global healthcare challenge, contributing to numerous preventable deaths. Early initiation of basic life support (BLS) by lay bystanders significantly improves survival rates, potentially tripling the chances of survival.1 To promote public participation, the 2020 American Heart Association (AHA) cardiopulmonary resuscitation (CPR) guidelines recommend chest compression-only CPR, a simplified and accessible approach for lay rescuers.2 Video-based instruction has emerged as an effective BLS training method, offering advantages such as efficiency, emphasis on key concepts and reduced time constraints while fostering a positive learning attitude.3,6 However, while online learning effectively disseminates theoretical knowledge, it remains insufficient for teaching psychomotor skills.7 A brief video-based BLS course, supplemented by supervised hands-only compression practice, enhances lay rescuer competency, yet achieving proficiency may still require face-to-face training.8 BLS skills tend to deteriorate over time,9,11 and real-time visual feedback has been shown to improve CPR performance among non-healthcare providers.12 13 Despite these advantages, studies consistently report a significant decline in CPR skills and BLS knowledge after initial training.14 15 Mastery learning (ML) has been widely applied in medical training and proven effective for skill retention.16 17 However, its impact on maintaining chest compression skills among laypersons remains inconclusive.18 This study aimed to compare CPR skill retention and BLS knowledge between ML and short-duration self-training methods among laypersons trained through a video-based approach. Methods Study design and setting This study was designed as a comparative assessment of two CPR training methods conducted at the Faculty of Medicine, Thammasat University, from February to October 2023. Selection of participants Participants were required to be laypersons aged 18 years or older. Exclusion criteria included: (1) prior BLS certification, (2) self-reported inability to perform chest compressions or a medical condition exacerbated by physical exertion and (3) unwillingness to participate in the 6-month retention test. Protocol and study intervention Following the acquisition of informed consent, demographic information—including sex, age, body mass index (BMI), education level and prior experience with BLS—was collected. Participants subsequently engaged in a structured, three-step BLS learning protocol. First, participants viewed an 8 min self-directed learning video that introduced key BLS concepts, including the chain of survival, assessment of responsiveness, step-by-step BLS procedures, proper chest compression techniques and tempo guidance using songs paced at 100–120 beats per minute. Participants were subsequently introduced to and practised using real-time feedback manikins to develop familiarity and basic CPR skills (online supplemental file 1). This was followed by a pretest CPR assessment to evaluate baseline psychomotor skills. Participants were then randomised into either the short-duration self-learning (SSL) group or the ML group using permuted block randomisation (block size=4), generated via an online tool (http://www.randomization.com). The SSL group practised independently for a total of 5 min, divided into 1 min sessions with rest intervals, for up to five attempts. In contrast, the ML group practised with real-time feedback until they achieved a predefined mastery threshold, defined as a QCPR score exceeding 90%.19 To assist in maintaining an appropriate chest compression rhythm, both groups used songs paced at 100–120 beats per minute during practice. CPR performance was assessed 30 min after training to minimise fatigue-related effects, and skill retention was re-evaluated at least 6 months later. BLS knowledge and emergency service awareness were evaluated using a 10-item questionnaire (online supplemental file 2), administered at three time points: pretest (prior to viewing the instructional video), post-test (following manikin-based training) and a retention test conducted 6 months after the intervention.15 Outcomes and measurement The primary outcome of this study was CPR skill retention, evaluated by comparing overall compression performance scores between the immediate post-test and the 6-month retention test in the ML and SSL groups. Secondary outcomes included observational CPR performance, which assessed critical BLS steps such as scene safety evaluation, responsiveness assessment, activation of emergency services, breathing assessment (within 10 s) and correct hand placement for chest compressions. Quantitative CPR performance metrics were collected using the QCPR application, which measured average compression depth (centimetre), the percentage of compressions within the correct depth range (5–6 cm), compression rate (beats per minute), the percentage of compressions within the optimal rate (100–120/min), percentage of complete chest recoil and the overall compression performance score (0–100%). These scores were calculated using Laerdal Medical’s proprietary algorithm, aligned with the AHA consensus on CPR quality (online supplemental file 3). Mastery-level achievement was defined as a QCPR score exceeding 90%, and the number of attempts required for ML participants to reach this threshold was also recorded. BLS knowledge retention was assessed using pretest, post-test and 6-month retention test scores derived from a 10-item questionnaire. Compression performance was measured using a Laerdal Little Anne wireless skill reporter manikin connected to the QCPR application, which provided real-time feedback on compression depth, rate and chest recoil. At the end of each 1 min compression-only training session, participants received summary feedback to support skill improvement. To ensure objective evaluation, all CPR assessments were conducted as 2 min compression-only tests without real-time feedback. Sample size estimation As no previous studies have directly compared 6-month CPR skill retention between short-duration self-training and ML groups, the study by Srivilaithon et al was used as a reference for sample size estimation. Based on an OR of 0.46 (95% CI 0.09) and a p value of 0.35 in comparing trained versus untrained participants, the required sample size was calculated using STATA V.14. Assuming 80% power to detect a statistically significant difference with a two-sample comparison of proportions at a significance level of 0.05, the estimated sample size was 86 participants per group. Statistical analysis Data were analysed using STATA V.14. Categorical variables were summarised as frequencies and percentages and compared using the χ2 or Fisher’s exact test. Continuous variables were presented as mean±SD or median (IQR) and analysed using the independent t-test or Mann-Whitney U test, as appropriate. Paired t-tests assessed within-group changes over time, while repeated measures analysis of variance was used for comparisons across multiple time points. Risk ratios (RR) with 95% CIs were calculated to compare observational CPR performance between groups. Generalised linear models were applied for binary outcomes across time points. A multilevel regression analysis was used to compare prelearning and postlearning skills, as well as postlearning and retention skills, within the same participants. A p value <0.05 was considered statistically significant. Analyses were conducted using a per-protocol approach. Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research. Results A total of 188 administrative officers aged 18 years or older from Thammasat Hospital were assessed for eligibility. Of these, 163 participants were randomised into either the SSL group (n=82) or the ML group (n=81). 21 participants withdrew from the study due to various reasons, including physical disability and pregnancy, resulting in 71 participants remaining in each group for the 6-month retention assessment. The mean age of participants was 36.7±8.7 years, and the majority were female (86.5%). Most held a bachelor’s degree or higher (74.5%), and only a small proportion had prior CPR training or real-life CPR experience. Figure 1 illustrates the study design and participant flow across each study phase. Consolidated Standards of Reporting Trials (CONSORT) diagram and participant flow throughout the study. BLS, basic life support; CPR, cardiopulmonary resuscitation. All participants were hospital administrative officers working in non-clinical departments, encompassing diverse occupational roles such as finance, accounting, supplies management, procurement and general administration. The detailed occupational distribution of participants is presented in table 1. Table 1Occupational distribution of study participantsOccupationNumber of participants (n=188)Revenue and accounting officer45Inventory registry clerk39Finance officer26Supplies officer21Procurement officer20Administrative officer19Accountant18Total188 Occupational distribution of study participants Baseline characteristics between the SSL and ML groups showed no statistically significant differences in age, sex, education level, prior CPR training or real-life CPR experience. The only significant difference observed was in BMI, which was lower in the ML group compared with the SSL group (22.9±4.9 vs 25.9±7.3, p=0.003). Although a greater proportion of participants in both groups were female, this difference was not statistically significant. Full baseline characteristics are provided in table 2. Table 2Baseline characteristics of participantsCharacteristicTotal(n=163)Short-duration self-learning(n=82)Mastery learning(n=81)Age, mean (±SD), years36.7 (±8.7)37.7 (±8.9)35.7 (±8.4)Female, n (%)141 (86.5)67 (81.7)74 (91.4)Body mass index, mean (±SD), kg/m224.4 (±6.3)25.9 (±7.3)22.9 (±4.9)Education level, n (%) Primary school1 (0.6)1 (1.3)0 High school diploma16 (10.2)8 (10.2)8 (10.1) Diploma graduate23 (14.7)12 (15.4)11 (13.9) Bachelor’s degree103 (65.6)51 (65.4)52 (65.8) Master’s degree or higher14 (8.9)6 (7.7)8 (10.1)Previously trained in CPR, n (%)3 (1.8)2 (2.4)1 (1.2)Trained in CPR within the past 12 months, n (%)1 (0.6)01 (1.2)Real-life CPR experience, n (%)2 (1.2)1 (1.2)1 (1.2)CPR, cardiopulmonary resuscitation. Baseline characteristics of participants CPR, cardiopulmonary resuscitation. Observational and quantitative CPR performances were assessed in both the SSL and ML groups. Observational performance included critical BLS steps such as scene safety, responsiveness assessment, activation of emergency services, breathing check and correct hand placement. Quantitative performance was measured using a QCPR device, which provided objective data on compression depth, compression rate, percentage of correct depth and rate, complete chest recoil and overall compression performance score. Both groups demonstrated high overall compression performance scores, with no statistically significant differences observed at post-test and retention. These results are presented in tables3 4. The completeness of adherence to the initial resuscitation steps in both groups is illustrated in online supplemental file 4. Table 3Comparison of observational CPR performance among short-duration self-learning and mastery learning groupsSSL (n=82)ML (n=81)Risk ratio (95% CI)P valueChecking scene safety, n (%) Pretest0 (0)1 (1.2)NANA Post-test82 (100)81 (100)NANA Retention test; SSL (n=71), ML (n=71)71 (100)71 (100)NANAChecking unresponsiveness, n (%) Pretest10 (12.2)14 (17.3)1.4 (0.67 to 3.00)0.363 Post-test82 (100)81 (100)NANA Retention test; SSL (n=71), ML (n=71)71 (100)71 (100)NANACalling emergency services, n (%) Pretest0 (0)1 (1.23)1NA Post-test82 (100)81 (100)NANA Retention test; SSL (n=71), ML (n=71)71 (100)71 (100)NANAChecking breathing for no more than 10 s, n (%) Pretest9 (11.0)14 (17.3)1.57 (0.72 to 3.43)0.253 Post-test82 (100)81 (100)1NA Retention test; SSL (n=71), ML (n=71)71 (100)71 (100)1NACompression position, n (%) Pretest77 (93.9)71 (87.7)0.93 (0.85 to 1.03)0.171 Post-test81 (98.8)81 (100)1NA Retention test; SSL (n=71), ML (n=71)70 (98.6)71 (100)1NACPR, cardiopulmonary resuscitation; ML, mastery learning; SSL, short-duration self-learning. Comparison of observational CPR performance among short-duration self-learning and mastery learning groups CPR, cardiopulmonary resuscitation; ML, mastery learning; SSL, short-duration self-learning. Table 4Comparison of quantitative CPR performance among short-duration self-learning and mastery learning groupsSSL (n=82)ML (n=81)Mean difference (95% CI)P valueAverage compression depth (cm) Pretest5.1±0.35.1±0.30.02 (−0.07 to 0.10)0.691 Post-test5.1±0.25.1±0.20.03 (−0.4 to 0.10)0.415 Retention test; SSL (n=71), ML (n=71)5.6±0.85.4±0.70.14 (−0.12 to 0.40)0.285Percentage of compressions within the correct depth range (5–6 cm) Pretest94.9±11.892.1±17.02.82 (−1.70 to 7.35)0.219 Post-test96.3±12.094.7±15.71.61 (−2.71 to 5.94)0.463 Retention test; SSL (n=71), ML (n=71)94.9±15.094.1±14.40.77 (−4.11 to 5.66)0.754Chest compression rate (beats per minute) Pretest110.4±8.5112.2±12.61.78 (−1.53 to 5.10)0.290 Post-test112.8±5.1114.1±5.51.37 (−0.27 to 3.01)0.102 Retention test; SSL (n=71), ML (n=71)111.3±4.7112.6±4.91.31 (−0.28 to 2.90)0.105Percentage of compressions within the optimal rate (100–120/min) Pretest81.1±24.980.0±25.21.13 (−6.61 to 8.88)0.773 Post-test86.6±20.083.0±22.23.63 (−2.90 to 10.17)0.274 Retention test; SSL (n=71), ML (n=71)87.6±15.688.7±18.40.87 (−4.80 to 6.54)0.761Percentage of complete chest recoil Pretest95.0±12.493.0±15.62.01 (−2.34 to 6.37)0.363 Post-test91.1±18.292.1±16.11.01 (−4.30 to 6.33)0.707 Retention test; SSL (n=71), ML (n=71)93.5±14.794.9±13.51.43 (−3.25 to 6.13)0.546Overall compression performance score Pretest93.3±15.393.9±10.30.66 (−3.39 to 4.71)0.749 Post-test97.4±4.297.1±5.30.24 (−1.24 to 1.72)0.747 Retention test; SSL (n=71), ML (n=71)96.6±5.096.8±5.70.20 (−1.58 to 1.97)0.826Plus-minus values are means±SD.CPR, cardiopulmonary resuscitation; ML, mastery learning; SSL, short-duration self-learning. Comparison of quantitative CPR performance among short-duration self-learning and mastery learning groups Plus-minus values are means±SD. CPR, cardiopulmonary resuscitation; ML, mastery learning; SSL, short-duration self-learning. When comparing prelearning and postlearning, it was found that both methods significantly improved the learners’ practical skills, particularly in terms of scene safety assessment (SSL; RR=81.99, p<0.001), checking unresponsiveness (SSL; RR=5.47, p<0.001, ML; RR=8.99, p<0.001), activation of emergency services (SSL; RR=81.99, p<0.001), checking breathing (SSL; RR=5.47, p<0.001, ML; RR=10.13, p<0.001) and correct hand positioning for chest compressions (SSL; RR=1.13, p<0.006, ML; RR=1.07, p<0.026). At the 6-month follow-up, no statistically significant differences were observed between groups in most BLS skills. These findings suggest that both methods result in similar long-term retention for most CPR skills. Detailed results are provided in table 5. Table 5The performance of basic life support skills at prelearning compared with immediate postlearning, and immediate postlearning compared with 6 months postlearning in the short-duration self-learning group and the mastery learning groupThe assessed skillPrelearning versus immediate postlearningImmediate postlearning versus 6 months postlearningShort-duration self-learning(n=82)Mastery learning(n=81)Ratio of differenceP valueShort-duration self-learning(n=71)Mastery learning(n=71)Ratio of differenceP valueChecking scene safety Risk ratio (95% CI)81.99 (11.62 to 578.6)N/A1.01 (0.82 to 1.26)1.00 (0.73 to 1.38)1.00 (0.73 to 1.38)N/A P value<0.0010.91211Checking unresponsiveness Risk ratio (95% CI)5.47 (3.46 to 8.65)8.99 (4.85 to 16.69)1.06 (0.88 to 1.28)1.00 (0.73 to 1.38)1.00 (0.73 to 1.38)N/A P value<0.001<0.0010.51311Calling emergency services Risk ratio (95% CI)81.99 (11.62 to 578.6)N/A1.01 (0.82 to 1.26)1.00 (0.73 to 1.38)1.00 (0.73 to 1.38)N/A P value<0.0010.91211Checking breathing Risk ratio (95% CI)5.47 (3.46 to 8.65)10.13 (5.23 to 19.59)1.08 (0.89 to 1.30)1.00 (0.73 to 1.38)1.00 (0.73 to 1.38)N/A P value<0.001<0.0010.44311Compression position Risk ratio (95% CI)1.13 (1.03 to 1.22)1.07 (1.01 to 1.13)0.96 (0.92 to 1.01)0.99 (0.72 to 1.37)1.00 (0.73 to 1.38)N/A P value0.0060.0260.1310.9911Chest compression depth Risk ratio (95% CI)0.98 (0.93 to 1.02)0.99 (0.92 to 1.05)1.02 (0.98 to 1.06)0.80 (0.57 to 1.13)0.92 (0.66 to 1.28)0.96 (0.89 to 1.03) P value0.3140.70.3480.2140.6190.271Chest compression rate Risk ratio (95% CI)1.09 (0.97 to 1.21)1.01 (0.91 to 1.12)0.99 (0.92 to 1.07)1.03 (0.75 to 1.43)1.06 (0.76 to 1.48)1.01 (0.96 to 1.08) P value0.1380.7980.8880.8440.7180.633Complete chest recoil Risk ratio (95% CI)1.04 (0.81 to 1.35)1.02 (0.79 to 1.31)0.98 (0.82 to 1.17)1.08 (0.73 to 1.62)1.02 (0.68 to 1.54)1.01 (0.85 to 1.21) P value0.7520.8730.8060.6940.9070.871 The performance of basic life support skills at prelearning compared with immediate postlearning, and immediate postlearning compared with 6 months postlearning in the short-duration self-learning group and the mastery learning group Both groups exhibited significant improvements in BLS knowledge scores immediately following training, followed by a decline at the 6-month follow-up. Nevertheless, retention scores in both groups remained higher than pretest levels. In the SSL group, the mean post-test score increased by 4.9±2.3 (95% CI 4.33 to 5.42), and the retention score increased by 2.0±2.7 (95% CI 1.33 to 2.59) compared with the pretest. Similarly, the ML group showed a mean post-test increase of 4.4±2.1 (95% CI 3.87 to 4.86) and a retention gain of 1.8±2.5 (95% CI 1.29 to 2.35); all changes were statistically significant (p<0.001). The decline in scores from post-test to retention test was −2.9±1.8 (95% CI −3.33 to −2.47) in the SSL group and −2.55±1.8 (95% CI −2.29 to −2.12) in the ML group, with no significant difference in long-term knowledge retention between groups. Detailed findings are presented in table 6. Table 6Comparison of general knowledge of basic life support and emergency medical service among short-duration self-learning and mastery learning groups, pretest, post-test and retention test quiz (total score 0–10)SSL (n=82)ML (n=81)Mean difference (95% CI)P valuePretest5.1±2.25.6±2.00.5 (−0.17 to 1.17)0.145Post-test10±0.09.9±0.30.07 (−0.04 to 0.18)0.222Retention test7.1±1.87.4±1.80.28 (−0.31 to 0.88)0.350Plus-minus values are means±SD.ML, mastery learning ; SSL, short-duration self-learning. Comparison of general knowledge of basic life support and emergency medical service among short-duration self-learning and mastery learning groups, pretest, post-test and retention test quiz (total score 0–10) Plus-minus values are means±SD. ML, mastery learning ; SSL, short-duration self-learning. Among participants in the ML group, the mean number of attempts required to achieve mastery was 2.9 (95% CI 2.7 to 3.1). Female participants required significantly more attempts than their male counterparts (3.0 (95% CI 2.8 to 3.2) vs 2.1 (95% CI 1.2 to 3.1), p=0.007), as shown in table 7. Table 7Number of attempts for mastery learning observed among mastery learning groupNumber of attemptsMale (n=7)Female (n=74)Total (n=81)1n (%)3 (42.9)1 (1.4)4 (4.9)2n (%)0 (0)14 (18.9)14 (17.3)3n (%)4 (57.1)47 (63.5)51 (63.0)4n (%)0 (0)8 (10.8)8 (9.9)5n (%)0 (0)4 (5.4)4 (4.9)Mean (95% CI)2.1 (1.2 to 3.1)3.0 (2.8 to 3.2)2.9 (2.7 to 3.1)Percentages may not total 100 because of rounding. Number of attempts for mastery learning observed among mastery learning group Percentages may not total 100 because of rounding. Discussion This study examined the effectiveness of ML and SSL methods in improving and retaining CPR skills and BLS knowledge among laypersons trained via a video-based approach. Both methods significantly enhanced immediate post-training performance in key BLS components, including scene safety, responsiveness checks, emergency activation, breathing assessment and correct hand positioning. At the 6-month follow-up, no statistically significant differences were found between groups in most BLS skills, including compression depth, rate and hand placement. These findings align with previous research by Boet et al and Srivilaithon et al, which reported no long-term advantage of ML over time-fixed methods in BLS skill retention despite an overall decline in compression performance over time.15 18 A possible explanation for the lack of a significant effect of ML is the minimal difference in total practice time between groups. In this study, ML participants required one to five attempts to reach mastery (mean of 3), which was comparable to the SSL group’s maximum practice allowance. Thus, the main distinction was ML’s requirement to meet a 90% performance threshold. However, both groups exceeded this benchmark on the immediate post-test, which may account for the absence of a measurable intervention effect. ML is widely used in medical education to enhance skill retention and reduce knowledge decay. For layperson training, essential elements include recognising cardiac arrest, activating Emergency Medical Services (EMS), performing CPR and using an AED (Automated External Defibrillator).20 This study focused specifically on CPR quality, measured by the QCPR score, which captures key metrics such as compression depth, rate, chest recoil and hand placement. While ML relies on deliberate, repetitive practice often driven by intrinsic motivation in medical professionals, laypersons may be more influenced by external factors such as institutional mandates, workplace policies or personal interest.21 Currently, no established mastery cut score exists for lay rescuers. Our pilot study indicated that five practice attempts with rest intervals represent the maximum feasible workload for non-medical participants. In this study, a 90% mastery threshold was adopted based on prior research using Laerdal QCPR, which defined 60% as minimum competence, 75% as advanced and 90% as high proficiency, along with principles of patient safety and overlearning.19 While this standard may require further validation, setting excessively high benchmarks could unnecessarily prolong training without added benefit.22 23 Yudkowsky et al emphasised that in ML, standards should be challenging but also educationally efficient, warning that overly stringent cut scores may lead to wasted effort without meaningful gains, especially when not linked to improved outcomes at the next stage of training.24 Notably, no decline in CPR compression retention was observed in the ML group, suggesting that this performance criterion may be appropriate for promoting long-term CPR skill retention in laypersons. Our findings partially align with those of Wik et al, supporting the role of repeated practice, real-time feedback and self-directed training in CPR skill acquisition.25 However, we did not observe additional long-term benefit from overtraining beyond a defined mastery threshold, suggesting diminishing returns once competence is achieved. Instead, periodic short refresher sessions may offer a more efficient strategy for maintaining skill retention. Future training models may benefit from hybrid approaches that integrate mastery-based standards with accessible formats and scheduled booster sessions. The high QCPR scores observed in both the post-test and 6-month retention test may be explained by several factors. Real-time feedback manikins likely enhanced psychomotor learning by enabling immediate performance adjustments. Repeated practice, even in the SSL group, may have reinforced core compression skills such as depth, rate and recoil. The use of compression-only CPR simplified the task, reducing cognitive load and allowing focus on key components. Additionally, the QCPR algorithm’s emphasis on mechanical precision may have favoured technical accuracy. As all participants were hospital staff, factors such as baseline health awareness, intrinsic motivation, learning readiness and digital literacy may have further supported high performance and skill retention in this cohort. An 8 min self-directed video was developed to teach key BLS concepts to laypersons, including the chain of survival, responsiveness assessment, EMS activation, compression-only CPR and correct hand placement. Auditory cues paced at 100–120 beats per minute supported rhythm, while simple language, clear visuals and step-by-step demonstrations enhanced accessibility and retention. To promote motivation and align with adult learning principles, the training emphasised the real-life relevance of CPR and simplified steps to support understanding. The concept of ‘time is brain’ was reinforced to encourage prompt action, and theoretical content was linked with practical application to aid cognitive integration. The video deconstructed BLS skills using visual aids to demonstrate posture and body mechanics, beginning with CPR in a standing position to illustrate effective compression. Techniques were shown step by step with multiple camera angles to support mental rehearsal, and participants practised simulated compressions focusing on hip movement to promote proper form and reduce fatigue. The session concluded with compressions synchronised to a rhythmic beat, in line with cognitive load theory to enhance psychomotor learning.26 A follow-up video reinforced BLS steps and guided interpretation of real-time feedback, addressing common errors such as excessive compression rate by encouraging focus on depth and recoil before adjusting speed. This structured approach, combined with immediate hands-on practice using feedback manikins, supported skill acquisition and reflected ML principles.27 Prior studies have shown that real-time feedback can effectively improve CPR performance among laypersons.28 Significant improvement in CPR skills was observed immediately after the learning session, despite participants receiving instruction solely through online materials. All initial CPR steps, including accurate hand placement, improved following practice in both groups, underscoring the critical role of correct hand positioning in both immediate performance and long-term skill retention. However, many participants initially placed their hands incorrectly despite viewing instructional videos, suggesting that video-based learning alone may be insufficient for mastering psychomotor skills. This highlights the value of instructor-guided training, which offers real-time feedback and correction—an essential component missing from online-only approaches. Nevertheless, participants in the ML group demonstrated notable improvement in hand placement accuracy, suggesting that the structured, competency-focused nature of ML may better support the acquisition of this essential skill. These findings reinforce the importance of initial CPR skills practice under instructor supervision to ensure proper technique before transitioning to self-directed or technology-assisted learning. Future research should further examine the comparative effectiveness of instructor-led versus media-based training, particularly in improving hand placement and other critical components of high-quality CPR. Although female participants outnumbered males, they required more attempts to achieve mastery. This may be partially attributed to smaller average body stature among females, potentially limiting the ability to generate adequate force for effective chest compressions. Such physiological differences can affect CPR performance, particularly in tasks demanding upper body strength and precision.29 BLS knowledge retention did not significantly differ between ML and SSL groups; however, both showed marked post-test improvement followed by a moderate decline at 6 months, with retention scores remaining above pretest levels—consistent with prior studies. This may be attributed to the availability of accessible online materials and concise instructional videos that supported ongoing reinforcement. Furthermore, as most participants were hospital staff, their sustained engagement and interest in BLS-related content may have contributed to improved long-term knowledge retention. In accordance with AHA recommendations, ML is particularly well suited for high-stakes behaviours requiring a high level of proficiency, such as chest compressions—a core component of CPR that directly impacts clinical outcomes and is often challenging to perform correctly.30 In this study, real-time feedback was incorporated to support precise skill acquisition. Although CPR proficiency typically declines within 6 months, both ML and SSL approaches demonstrated comparable long-term retention. While ML provided an immediate advantage in skill acquisition, both groups achieved similar CPR skill and BLS knowledge retention at 6 months. From a practical perspective, SSL requires fewer resources, less instructor supervision and shorter training time, making it a cost-efficient and scalable model for large-scale or community implementation—especially in resource-limited or remote settings. ML, on the other hand, ensures consistent proficiency but involves greater time and personnel investment, which may restrict its feasibility outside structured institutional programmes. The sustained retention in both groups underscores the value of simple, accessible video-based tools. These findings highlight the importance of tailoring BLS training strategies to specific populations. SSL offers a pragmatic approach for public and workplace training initiatives, whereas ML remains valuable in clinical or high-stakes environments where performance precision is critical. Future research should aim to validate mastery thresholds for laypersons, investigate booster training to support retention and assess the cost-effectiveness of ML and SSL in large-scale programmes. Integrating real-time feedback and digital tools may further enhance the impact of BLS training in community settings. Overall, this study supports the implementation of both SSL and ML approaches in community-based BLS programmes. ML may be preferable when immediate performance is critical (eg, clinical staff onboarding), whereas SSL may be more practical for large-scale dissemination and long-term retention when combined with periodic refresher opportunities. Limitations This study has several limitations. First, the participant pool consisted exclusively of hospital administrative staff, who may possess higher baseline health awareness, motivation and digital literacy than the general population. As such, the findings may not be generalisable to laypersons without healthcare affiliations. However, although these participants were employed within a hospital environment, they had no formal clinical training or patient care responsibilities. Therefore, their experience and resuscitation skills were comparable to those of typical laypersons in the community. Second, while CPR performance was evaluated using both observational and quantitative metrics, assessments were conducted in a controlled setting, which may not reflect real-life emergency scenarios. Third, the short duration between training and post-test could have contributed to elevated scores in both groups, potentially masking subtle differences in learning effectiveness. Fourth, the ML group required only a small number of attempts to achieve the performance threshold, resulting in minimal time differences between groups and limiting the observable effect of the intervention. Lastly, the use of a specific QCPR scoring algorithm may emphasise mechanical precision, which, while useful for training, might not fully capture the complexity of clinical performance under stress. Conclusion This study demonstrates that both SSL and ML are effective in improving CPR skills and BLS knowledge among laypersons. While ML provided an initial advantage in skill acquisition, both approaches yielded comparable long-term retention at 6 months. The findings highlight the value of accessible, video-based learning tools and the role of real-time feedback in enhancing skill performance. Although ML ensures a higher level of proficiency through structured repetition, SSL offers a scalable and practical alternative, particularly in community and resource-limited settings. Future training models should consider integrating mastery-based elements with digital platforms and periodic refreshers to optimise both initial learning and sustained performance in BLS training. Supplementary material Acknowledgements The authors thank all the participants from Thammasat University Hospital. We also thank Dr Dilok Piyayotai (Dean of Faculty of Medicine, Thammasat University; Associate Professor) for his support of this study. This study is part of the project from the Thammasat University Research Unit in Emergency Medicine and Emergency Critical Care. The funder had no role in study design, data collection, data analysis, manuscript preparation, or the decision to publish this work. Footnotes Funding: The study was supported by Faculty of Medicine, Thammasat University, Thailand (funding number: TP.1-04-2564). Prepublication history and additional supplemental material for this paper are available online. To view these files, please visit the journal online (https://doi.org/10.1136/bmjopen-2025-105678). Provenance and peer review: Not commissioned; externally peer reviewed. Patient consent for publication: Not applicable. Ethics approval: This study involves human participants and was approved by the Human Research Ethics Committee of Thammasat University (Medicine), Thailand (IRB No MTU-EC-EM-1-011/64, COA No 059/2021), in full compliance with international guidelines such as the Declaration of Helsinki and Good Clinical Practice. All confidential data were securely stored and used exclusively for the purposes of the clinical trial, in compliance with ethical standards. Participants gave informed consent to participate in the study before taking part. Data availability free text: Data are available from the corresponding author upon reasonable request. Patient and public involvement: Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research. Data availability statement Data are available upon reasonable request. References Supplementary Materials Data Availability Statement Data are available upon reasonable request.