Modeling the Microsurgical Learning Curve Using a Poisson-Based Statistical Approach for Skill Assessment

The learning curve (LC), a multifaceted concept, plays a pivotal role in evaluating surgical training. This study aimed to define critical inflection points in the microsurgical learning curve, develop a reliable index for skill assessment, and statistically validate this approach using Poisson distribution theory.

A standardized microsurgical training protocol was employed using a biological simulator. Data regarding time to complete the task and error rates were collected over 132 attempts by a single operator. The primary outcome variable, the major mistake average (MMA), was used to generate a learning curve. Its progression was analyzed using autoregressive integrated moving average (ARIMA) modeling and validated with Poisson dispersion theory to determine the randomness of error occurrence at advanced stages of training. The entire trial was conducted by a single operator, a consultant neurosurgeon from our institution, who had been properly instructed on the protocol and the corresponding operator's manual.

Task completion time (TCT) ranged from 860 to 3,054 seconds (mean: 1,472 seconds; R² = 0.561). MMA peaked at the 19th attempt (0.263) and decreased progressively, reaching 0.091 by the 132nd attempt (R² = 0.835). Three distinct phases of learning were identified, culminating in a plateau phase during which major mistakes followed a Poisson distribution (Chi² = 3.841), suggesting random occurrence independent of skill deficits.

The MMA was found to be a robust and objective indicator of microsurgical proficiency. Its statistical validation using Poisson distribution theory supports its utility in skill assessment and training programs. Further studies involving multiple operators are warranted to confirm these findings.