We performed this study in accordance with guidelines outlined in the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) (PROSPERO ID: 341439). We conducted a comprehensive search for relevant studies (without language limitations) from major online databases, such as PubMed, Embase, and the Cochrane Library, from inception to 31 March 2022. Two independent reviewers scanned the literature and included the eligible studies by common consensus after multiple rounds of screening.

The search process included (i) reading the reference section of all relevant research carefully; and (ii) manually searching abstracts of key journals and papers published at major annual conferences. The search terms used were a mix of (“learning transfer” [All Fields] OR “boundary crossing training” [All Fields]) AND (“clinical performance” [All Fields] OR “academic theoretical knowledge” [All Fields] OR “professional practice experience” [All Fields]). We also checked the reference lists of the screened studies to identify other similar studies. The search strategy is shown in Table 1. We included experimental studies that examined the influence or effectiveness of learning transfer on the clinical performance of medical staff. The PICOS criteria are used to select the eligible studies. Studies were included if they satisfied the following inclusion criteria: (1) The study was limited to RCTs and humans; (2) All participants are medical staff; (3) The study included both an experimental group and a control group. The experimental group was subjected to a learning lesson while the control group was performed without learning a lesson. (4) The study reported the effect of learning transfer on the performance of medical staff in each group. The exclusion criteria were as follows: (1) unqualified study design, such as non-RCT design, single-arm extension study, observational studies; (2) case reports, editorials, or reviews; (3) duplicated reports.

For all articles included, we extracted the following information from the original articles: first author, publication year, country, database, study duration, study design, study subjects, mean age of study subjects, gender of the study subjects, and outcomes. Two reviewers independently performed an analysis of methodological quality. The quality assessment included the following items: allocation generation and concealment, blinding, follow-up duration, loss follow-up (%), and data-analysis method (intention-to-treat or per protocol). Divergences were resolved through discussion and consensus. Further, we used version 2 of the Cochrane risk-of-bias tool for randomized trials (RoB 2) to assess the risk of bias for Systematic Reviews of Interventions. Any disagreement was resolved through discussion with a third author. All analyses were performed by Review Manager version 5.4.1.

The results of the systematic review are presented in Figure 1. We identified a total of 14 studies related to the transfer of learning after a thorough review of all papers. The characteristics of the studies are listed in Table 2. Among the studies considered in this paper, seven were conducted in Europe, three were in Canada, two were in the United States, and one each was conducted in Australia and South Korea.

Table 3 shows the results of a methodological quality assessment of all included studies. We considered inadequate allocation concealment and sequence generation the most common sources of potential bias. Due to the few studies included and the degree of heterogeneity observed in the study design, interventions, and outcome indices, meta-analysis was considered impractical.

Figure 2 presents a summary assessment of bias risk. Butler et al. (21), Bjurström et al. (22), Yang et al. (27), and Beattie et al. (29) did not clearly describe how the research populations are selected. Yang et al. (27) and Beattie et al. (29) did not clearly explain whether the participants were blinded. Yang et al. (27), Tolsgaard et al. (24), Rutherford-Hemming et al. (25), Kulamakan et al. (26), and Anacleto et al. (30) lost a number of research objects to follow-up, and therefore we must assume a high risk of bias. Setting the issue of uncleared blinded participants aside, all but Rutherford-Hemming et al. (25), Genç and Öner (28), and Lee et al. (31) did not blind participants; thus, their assessment of outcomes must be regarded as questionable.

Jensen et al. (19) reported a significant transfer effect when performing the second task, as there was a learning session in the first task. Butler et al. (21) reported that medical students were trained to perform diagnostic arthroscopy of the knee on anatomic dry models before training on cadaver specimens. Their average number of minimum proficiency tests and the average time of proficiency were significantly less than for those who only trained on cadaveric specimens. Kulamakan et al. (26) performed two sequential experiments. In Experiment 1, increasing context variation and conceptual analogies both significantly led to higher performance for far transfer. Experiment 2 demonstrated that even though there was a superficial similarity to previous examples, learners' shifts to using structural characteristics to classify new problems were caused by such analogies and context variation.

Yang et al. (27) reported that the participants performed appendectomy training in the virtual reality simulator before the tutorial procedural tasks of laparoscopic cholecystectomy, and needed significantly fewer movements and shorter path lengths than those who started with the tutorial procedural tasks of laparoscopic cholecystectomy directly. Anacleto et al. (30), after watching the video-mentored tutorial (VMT), observed a decrease in the total number of errors in peg-transfer (PT) and needle-guidance (NG) exercises in the participants who had watched the European Basic Laparoscopic Urologic Skills (E-BLUS) video before. Compared with the group in which students participated in a conventional demonstration of a Papanicolaou smear, Lee et al. (31) reported that self-confidence, learner satisfaction, and critical thinking were significantly higher in the simulation problem-based learning (S-PBL) group. After nine task repetitions over 3 weeks on one model, Ferguson et al. (23) reported that when the participants performing the knee and shoulder tasks swapped models, there was no immediate evidence of skill transfer.

Anastakis et al. (18) reported that bench and cadaver training were both superior to text learning and were equivalent. Tolsgaard et al. (24) reported that compared to the single-student composition, dyads illustrated higher training efficiency in terms of simulator score per number of attempts. Compared with nurses in the online self-study module, Rutherford-Hemming et al. (25) reported that the simulation group showed higher levels in both short-term and long-term skill performance.

Genç and Öner (28) reported that participants receiving an integrated instructional video performed significantly better on transfer than the procedural-only video. Beattie et al. (29) reported that groups trained in 3D and tested in 3D demonstrated superior training performance and took fewer repetitions to reach proficiency than the group in 2D.

The last two papers addressed in this study discuss the relationship between teacher guidance and learning transfer. While conducting research with 61 clinical nurse specialists as participants, Heaven et al. (20) assessed the three time points and observed that only those who experienced supervision showed any demonstration of transfer. Although neither group promoted more revelation of clues or concerns, people in the experimental group responded more effectively to the revealing clues. Bjurström et al. (22) reported that experienced surgeons performed significantly better than the control group and the self-guided training group. Between the educator-guided training group and the experienced surgeons, there was no significant difference. Nevertheless, having an educator present during training seemed to have a beneficial effect.

Our study integrated the current findings of 14 studies and illustrated the correlation between the learning transfer and clinical performance of medical staff. However, the experimental items in each article based on learning transfer are not all the same. Therefore, we cannot compare them to determine which methods of learning transfer ensure significant clinical performance for medical staff. In short, we can only understand the transfer of learning in different fields based on a synthesis of current findings.

Previous research showed that learning transfer is considered a major influence on clinical performance (3). Learning in hospital settings encompassed both formal and informal activities. Formal learning means formalized and standardized education, including career staff training, preceptorships, maintenance education, and job training. Nonetheless, owing to the nature of shift work and the organizational complexity and diversity, it is not possible to make sure that medical staff can improve their clinical performance only through formal learning. Informal learning, which consists of communication, interaction with others, role modeling, and team-based learning, is more flexible and plays an important role in developing the medical staff's clinical performance, especially professional practice experience (32, 33). Informal learning is not simply passive inputting of information but involves constructing the meaning of information actively by recording accumulated long-term memory and existing experience. Therefore, it is crucial to decide which approaches should be used in informal learning, supporting continual self-directed learning.

In this study, we found better task performance after medical students were trained in the virtual simulator, which indicates that when knowledge and skills had common basic principles, the learning transfer would be apparent (27). Learners apply the acquired knowledge to the new learning or work, illustrating the importance of learning transfer to the clinical performance of medical staff and cultivating their practical ability and creative spirit. With an effective transfer, learners can learn faster and better in a limited time and transfer more accurately in the appropriate environment. It seems that learning transfer also depends on the direct instruction of teachers (3, 20), as instruction can more effectively promote learning transfer. The report showed no significant difference between the training group guided by educators and experienced surgeons (22). Nevertheless, strengthening guidance during learning can help students to improve their specific knowledge to general principles as early as possible.

The systematic review found that despite the lack of consistency in the duration of the intervention, practice time, assessment, and outcome measure, there is a significant learning transfer in the final task after interventions such as different training methods and guidance. According to the results, in order to make an effective learning transfer, professionals should focus on the following three points: (1) Look for the similarities between concepts and principles; (2) notice the summary of learning methods, that is, master the method of solving problems in the learning process, and (3) accumulate a wide range of learning experience in all aspects.

Owing to the developments of new technologies and the shift in the medical paradigm, E-Learning within the CBBL framework is seen as a very promising tool to prompt the advancement of learning transfer. Turk et al. explored that E-Learning within CBBL framework not only facilitates the creation of up-to-date teaching content but also addresses difficulties in transforming declarative knowledge into procedural knowledge and skill (34). Lütgendorf-Caucig highlighted that the integration of different teaching modalities is beneficial for the knowledge acquisition for clinical decision making in a multidisciplinary environment like oncology (35). It has been also emphasized in other studies that CBBL is an effective way in gaining improvements in performance among medical staff and is essential for associative and procedural learning that is necessary for clinical reasoning processes (36–39). Hence, we would recommend that the concept of interactive CBBL methods should be developed further and applied in other medical fields. To guarantee the high quality and employ correct didactic dimensions in terms of constructing the interactive questions, it may be helpful to create a guideline for question generation with the collaboration of medical education experts from their research field. Meanwhile, in order to provide a diversity of CBBL materials, further scientific, methodological, theoretical, and practice-based breakthroughs must be achieved.

We individually evaluated these studies using assessment tools and covered most of the articles related to learning transfer. However, several limitations need to be addressed. First, few studies included questions on the reliability of the research results and the strength of the conclusions. Second, due to the considerable heterogeneity of research design and outcome variables, it was impossible to perform an effective meta-analysis. Third, the research objects participating in our review may have differed in analysis and generalizability. Finally, there may be interference from other related factors. Although the participants in some studies are similar in age, their sex ratios are quite different; moreover, nearly half of the studies did not clearly indicate this. In addition, there is currently a lack of studies that provide the quantitative results to meet the condition of performing an effective meta-analysis. We recommend that researchers conduct randomized controlled trials to further evaluate this correlation. We also recommend a study comparing the transfers of different interventions to provide more comprehensive and general findings.