Background: A surgical procedure is a complex behavior that can be constructed from foundation or component behaviors. Both the component and the composite behaviors built from them are much more likely to recur if it they are reinforced (operant learning). Behaviors in humans have been successfully reinforced using the acoustic stimulus from a mechanical clicker, where the clicker serves as a conditioned reinforcer that communicates in a way that is language- and judgment-free; however, to our knowledge, the use of operant-learning principles has not been formally evaluated for acquisition of surgical skills.
Questions/purposes: Two surgical tasks were taught and compared using two teaching strategies: (1) an operant learning methodology using a conditioned, acoustic reinforcer (a clicker) for positive reinforcement; and (2) a more classical approach using demonstration alone. Our goal was to determine whether a group that is taught a surgical skill using an operant learning procedure would more precisely perform that skill than a group that is taught by demonstration alone.
Methods: Two specific behaviors, "tying the locking, sliding knot" and "making a low-angle drill hole," were taught to the 2014 Postgraduate Year (PGY)-1 class and first- and second-year medical students, using an operant learning procedure incorporating precise scripts along with acoustic feedback. The control groups, composed of PGY-1 and -2 nonorthopaedic surgical residents and first- and second-year medical students, were taught using demonstration alone. The precision and speed of each behavior was recorded for each individual by a single experienced surgeon, skilled in operant learning. The groups were then compared.
Results: The operant learning group achieved better precision tying the locking, sliding knot than did the control group. Twelve of the 12 test group learners tied the knot and precisely performed all six component steps, whereas only four of the 12 control group learners tied the knot and correctly performed all six component steps (the test group median was 10 [range, 10-10], the control group median was 0 [range, 0-10], p = 0.004). However, the median "time to tie the first knot" for the test group was longer than for the control group (test group median 271 seconds [range, 184-626 seconds], control group median 163 seconds [range 93-900 seconds], p = 0.017), whereas the "time to tie 10 of the locking, sliding knots" was the same for both groups (test group mean 95 seconds ± SD = 15 [range, 67-120 seconds], control group mean 95 seconds ± SD = 28 [range, 62-139 seconds], p = 0.996). For the low-angle drill hole test, the test group more consistently achieved the ideal six-step behavior for precisely drilling the low-angle hole compared with the control group (p = 0.006 for the median number of technique success comparison with an odds ratio [at the 95% confidence interval] of 82.3 [29.1-232.8]). The mean time to drill 10 low-angle holes was not different between the test group (mean 193 seconds ± SD = 26 [range, 153-222 seconds]) and the control group (mean 146 seconds ± SD = 63 [range, 114-294 seconds]) (p = 0.084).
Conclusions: Operant learning occurs as the behavior is constructed and is highly reinforced with the result measured, not in the time saved, but in the ultimate outcome of an accurately built complex behavior.
Level of evidence: Level II, therapeutic study.