Learning curve patterns generated by a training method for laparoscopic small bowel anastomosis

The characteristics of the ID of a simulation-based training course may influence the performance level that the participants can achieve as well as the time needed to attain that level [29]. Our results showed that residents had to perform an average of 23.8 procedures to attain proficiency in GJA and 24.2 for JJA, while 85 % of them met the standards after 30 GJA and 34 JJA. An individual analysis of anastomotic time enabled description of four types of learning curves (exponential, rapid, slow, and no tendency) that could be predicted after procedure number 8.

In another study that used a standardized technique to teach laparoscopic bowel anastomoses, the learning process required 40 procedures with a simulator [31], instead of the 23.8 and 24.2 in our study. We think the difference may lie in the fact that the study aimed to standardize the suturing technique used for the anastomosis and not to define a comprehensive ID. In a series registered in our simulation lab between 2004 and 2007, the average training time to complete GJA and JJA was 87.2 and 72.7 min, respectively [23]. The reduction of training time found in the present study (73.7 min for GJA and 57.4 min for JJA) followed the redesign and standardization of the ID using the educational concepts and theories described above. To correlate the ID with learning outcomes, it is important to define the methodology used [13]. A critical issue when defining the ID is to clearly differentiate the tools used for learning from the actual educational methods [35]. In surgical simulations, this has been a challenge, as the simulator itself has frequently been regarded as the educational method. On the other hand, the same simulators (e.g., endotrainers, virtual simulators, mannequins) can be used in widely different educational methods. Therefore, the ID represents the specific techniques used for learning.

In our study, time was the principal criterion for determining the learning curve. In a systematic review of minimally invasive abdominal surgery, the procedural time was also the most commonly used variable (86 %). Other outcomes frequently measured included intraoperative outcomes (56 %), postoperative outcomes (54 %), intraoperative technical skills (17 %), and patient-oriented outcomes (49, 8 %). In our study, leakage, an intra- and postoperative outcome, was the main variable assessed, as observed in the majority of the articles revised [36]. The overall results show a typical learning curve as described in the Dreyfus Model of Skill Acquisition, with an initial phase of rapid qualification followed by another phase of slow development [37].

Despite the fact that different surgeons are likely to learn at different rates, most studies compare mean duration of the operation between groups [36]. In our study, however, four types of learning curves were identified for both types of anastomosis. The trainees with an exponential pattern (type 1) showed a higher starting point than those showing a rapid curve (type 2), but both reached the standards after a comparable number of procedures. The different starting point reveals that each person has individual experiences and backgrounds outside and inside the operating room that can lead to a different initial level of expertise. Some participants may have played video games before the study, and video game users seem to learn endoscopic techniques more quickly [38]. On the other hand, participants practiced laparoscopic appendectomy and cholecystectomy during the study, and the differences in previous experience may have influenced the starting point. Interestingly, the slope of the curve (how fast a person learns a new task) was similar for both groups, which might demonstrate that the generic skills in laparoscopy (i.e., innate psychomotor abilities) were similar among the participants in the two groups. The trainees with a slow learning curve (type 3) did not need longer time to complete the initial procedures, which might also be correlated with previous experience. However, the rate of learning was slower than that of the participants in the type 1 and 2 groups, which probably indicates lower innate psychomotor abilities. Interestingly, this group was also capable of obtaining proficiency with deliberate and repetitive practice with feedback. The type 4 trainees showed no clear tendency. This correlates with previous observations that indicate certain individuals cannot attain proficiency despite extensive training. This is a controversial issue. Do these subjects lack the abilities to develop laparoscopic technical skills, or are the ID and time allotment inadequate? This question poses a challenge for the professional bodies responsible for training and certification. If a type 4 (or 3) learning pattern is identified, instructors can use the technical factors described in the Methods section to guide formative assessment, identify common errors, and prescribe repetitive and deliberate practice until performance improves. When type 1 and 2 participants attain proficiency, they either continue to practice to reach expert or master level, operate on patients with supervision, or learn another procedure.

Figures 1 and 2 provide examples of each type of learning pattern. The percentage of individuals with a type 2 learning curve was higher for JJA (45 vs. 20 %). This result, coupled with the lesser time required to complete an anastomosis during the entire learning curve, confirms a faster rate of learning with this procedure.

These findings have potential implications for designing training programs for residents or for experienced surgeons aiming to learn new procedures.

This study has implications for how simulation training is implemented in educational curricula. To date, many simulation-based educational programs have been designed to include as many competencies as possible within the time frame available for training, without taking into consideration learning outcomes [39]. This approach may not ensure that all trainees reach proficiency (or mastery) for each competency. Knowing the average training time needed with a specific ID will help estimate the number of sessions required to reach the performance goal. Using this approach may limit the number of procedures taught using simulation to those that are more prevalent, complex, and associated with a higher risk of patient morbidity and mortality.

Another implication of this study is that it suggests a more effective use of resources available for training. The deep understanding of the different learning patterns generated by a specific ID enables early identification of individual learning needs. This can be detected early in the developmental process (after eight procedures in our results). This helps to plan the probable number of procedures required by a particular individual. Trainees who reach the desired level early can move on to another module, and trainees who need more practice may seek specific advice and feedback early on [40].

Another finding of this performance-oriented individual approach is the need to identify reliable, objective, long-term outcomes, and also develop valid and reliable tools to assess performance [41]. This is especially important for institutions that wish to (1) certify surgeons based on objective, valid, and transparent criteria and (2) implement a skills-assessment curriculum to identify individuals that may not attain proficiency despite extensive training (type 4). In the latter case, we believe assessment should be done before trainees formally enroll in a residency, preferably in medical school, to help the subjects who lack these innate abilities choose an alternative professional field [42].

The precise definition of proficiency in terms of learners’ achievement will determine a trainee’s readiness to proceed with patients. This also may impact patient safety, as it sets a definable milestone for transitioning to clinical practice.

Finally, creating opportunities for individualized instructor-guided training and reflection may help motivated individuals to become reflective and self-regulated learners. They can potentially have the tools to improve their own and other team members’ performance throughout their career [43].

There are several limitations of this study. The learning curve patterns resulted from an individual analysis of the participants and not from integration of all results into groups. This was due to a limited number of trainees, and additional cases are needed to validate the patterns we identified. The 2-year duration of the study was based on the timing of the actual resident training program. Future research should be planned during a shorter time period to prevent the influence of confounding effects while practicing other minimally invasive techniques. Participant characteristics such as previous experience with laparoscopic techniques, video game use, concurrent surgical activity during the study, or innate psychomotor ability testing were not evaluated and might have correlated to the different learning curves. Finally, this study evaluated one ID—there was no comparison group—and the results were compared to other ID studies described in the literature.

These findings suggest several paths for future research. Once learning curve patterns have been identified and correlated with a specific ID, studies can search for evidence of the most effective strategies within a design. Findings can be related to the sequence and elements of the ID used to build the training activity or to the educational methods within each element so that they can better support the learning needs of trainees while accelerating learning.

We propose several strategies to analyze the impact on the rate of learning. One is “part-task” training, wherein tasks are deconstructed into parts to be learned separately before practicing the procedure as a whole [44]. Another is cognitive task analysis, which identifies the knowledge, thought processes, and goal structures experts rely on during task performance [45]. Another strategy is guided experiential learning, wherein learners receive strong, early guidance through a script and storyboarded video demonstrations, procedural checklists, practice with increasingly difficult problems, and evaluations [46]. There are numerous debriefing methods that can be compared (good judgment [28], video-assisted [47], in-simulation [48], technical and cognitive [49], within-team [50], scripted [51], and a blended approach [52]). In our study, we used an ex vivo animal model, but live tissue, cadavers, or virtual simulators can also be compared [53, 54]. Other strategies can include activation and assessment of prior knowledge [55], spatial analysis and video gaming skills [56], and early detection, classification, and correction of consequential errors [57].