Learning curves in orthopedic and trauma surgery: a systematic review

Abstract

Background The process of teaching, learning and evaluating technical skills represents a challenge for surgeons. In the past, companionship has allowed the transmission of knowledge, but without evaluating its performance. Techniques in development were performed most often with limited or without any control. However, recently, a need for a change in that domain has emerged. In order to answer this important question, the number of studies interested in learning curves in medicine, particularly in orthopedic and trauma surgery, is increasing. Questions/purposes - What are the different methods applied in order to model the learning processes for orthopedic or trauma surgical procedures? - Does a test exist to compare the learning curves between them? Methods A systematic literature search of MEDLINE, CENTRAL and Embase was carried out on October 01, 2024. From each article, data regarding the design of the study, the specialty, the primary objective, the primary endpoint, the series sizes, the methods applied to model the learning curves, the use of graphs and the use of tests were retrieved. Results In total, 216 studies were found relevant. Most trials were single center, retrospective and publicly funded. We found seven different methods applied to evaluate learning curves: CUSUM (cumulative summation) methods, LC-CUSUM (learning curve cumulative summation), comparisons of experienced vs beginner participants, comparisons of first procedures vs others, general linear regression models, development of a score to rate the difficulty of the procedure and survival models. We did not find any application of a method that directly compares two learning curves. Conclusions In the orthopedic and trauma literature, seven different methods are used to model and evaluate learning curves. There is not enough evidence to conclude which approach is most appropriate to describe learning curves in orthopedics. However, the LC-CUSUM and regression models appear to be the most robust approaches. No method was found in order to compare the learning curves between them. Level of evidence Level II, systematic review.