The learning curve for the Shouldice Repair: a pilot analysis of post-training specialized surgeons at the Shouldice Hospital

This study was approved by the Lakeridge Health Research Ethics Board.

This was a retrospective chart review to collect information on the learning curve of a Shouldice primary inguinal hernia repair (referred to as Shouldice Repair in this study), at the Shouldice Hospital.

The learning curve was analyzed by using operating time which has also been done for Lichtenstein and similar research in TAPP repairs [8, 9, 10].

The parameter of 300 cases (considered the early learning block) was used as a parameter because the 2018 HerniaSurge guidelines state, “Although no studies exist on a comparison of the learning curves of the different non-mesh techniques, the HerniaSurge group agrees that the Shouldice technique is not easy to learn. In The Shouldice Hospital, surgeons are only considered qualified after 300 cases!” [3]. The late learning block, 900-1,000th Shouldice Repair for primary inguinal hernia, was chosen based on a previous publication from the Shouldice Hospital, “These individuals were deemed as experts and invited to participate if they had each performed at least 1000 repairs.” [2].

Postoperative complications included recurrences, hematomas, seromas and wound infections. This information was found during chart review, through follow-up notes in the patient’s chart. As there was no systematic follow-up, we did not measure time lapse to complications. KEY WORDS: Hernia, Learning curve, Shouldice repair, Shouldice hospital

The study population consisted of four surgeons who worked at the hospital in 2023, were hired within the last 10 years, and performed ≥ 1,000 primary inguinal hernia repairs. The operations where the surgeon was not the primary operating surgeon, i.e. the surgeon was labelled as the assistant or records showed the operation as done by both the primary and assistant surgeons (in training or training another) were excluded.

The patient data included a population of male and female patients accepted for surgery at the Shouldice Hospital, who had a Shouldice Repair. Patients were excluded if they underwent a mesh repair, repair of a scrotal hernia, non-Shouldice or modified Shouldice tissue repair, a hernia that was not an inguinal type, recurrent groin hernias, repairs with bowel resection, and/or and surgery was deemed an emergency.

All surgeons undergo a training period, these cases were not included in the study. During this training period the surgeons had highly experienced senior surgeons acting as the assistant or teacher. Surgeons are finished training upon the chief surgeon’s approval. After this approval the newly independent surgeon operates without surgical assistance from senior surgeons. Once a surgeon obtains this approval they are considered as an independent post-training specialized surgeons.

Data was collected from the hospital’s database, records, as well as from a manual review of operative notes and charts. No follow up took place, postoperative outcomes were collected through recheck notes in the patient charts. Data was collected as close as possible to gather the approximate first consecutive 300 cases (early learning block operations) and 900-1000th cases (late learning block operations) where the surgeon was the primary operating surgeon. Hospital management was contacted for employment records, which were used to gather information on the surgeons training period, when surgeons are in “training” (assistant surgeons) versus operating independently. Additionally, a short questionnaire was emailed to the surgeons to gather data on previous surgical experience. The surgeons were asked about their experience prior to starting at the Shouldice Hospital (how many years of surgical experience they had, roughly how many inguinal hernia operations completed, and how many Shouldice Repairs had been done). Data was also collected on the number of operations each surgeon had done prior to their first operation as the primary operating surgeon (Shouldice Experience on primary inguinal hernias).

Data analysis included descriptive statistics for frequency/percentage of the population as well as mean and standard deviations. The patient characteristics were computed for each trainee and compared univariably. Additional analysis included a multivariate analysis to determine if hernia size is a risk factor for longer operating time, as well as analysis to determine homogenous populations (comparing BMI, age, and gender). Sample size was not calculated as this was a pilot project.

Based on the initial 300 surgeries, the individual learning rates of each surgeon were determined. Descriptions of differences between individual learning curves were provided. In order to obtain a reliable estimate, an overall learning curve (pooled estimate) was determined, based on the geometric average of individual performance. The learning curve was examined using cumulative sum analysis (CUSUM). CUSUM analysis identifies differences between the raw data from each individual case and the mean value of the group. These deviations are then accumulated sequentially, allowing for the detection of trends, shifts, and turning points (known as trend transitions). Therefore, the curves move upward for longer operating times and increasing cases of postoperative complications. A plateau in the learning curve occurs when the CUSUM values level off or show minimal change.

Due to the nature of observational data, there is a possibility of an uneven distribution of patients with high complication risks among consecutive series or between surgeons. This can lead to bias, resulting in overestimated operative time and complication rates unrelated to the surgeon’s level of surgical experience during periods when high-risk patients are densely distributed. To address this possible bias, a risk-adjusted cumulative sum analysis (RA-CUSUM) was performed. Therefore, predictive factors for surgical duration and occurrence of postoperative complication were identified through a multivariate analysis. Linear regression models were employed for operating times and logistic regression models for complication occurrences. For multivariate regression modelling adjustment, the following variables were incorporated: age, BMI, sex, ASA score, hernia size, hernia location, number of comorbidities (diabetes, dyslipidemia, hypertension, coronary heart disease, deep vein thrombosis, stroke, transient ischemic attach, myocardial infarction, perivascular disease), and smoking. The multivariable models were employed to estimate the impact of the variables on the outcome. Accordingly, for each case with its respective value combinations, an expected operating time and a probability of the occurrence of a complication were calculated (expected value) using the multivariable regression models. The observed operating time or presence of a complication (coded as 1 for yes and 0 for no) was divided by the expected value and ratios above 1 indicate inferior performance while values below 1 indicate superior performance. Further details of the RA-CUSUM procedure, along with the mathematical details, are described elsewhere [10]. In principle, the CUSUM and RA-CUSUM line rises for longer operating times and more complications but falls for shorter operating times and fewer complications. High and low peaks within the RA-CUSUM curve are utilized to identify and differentiate possible learning phases. The shape of the learning curve in combination with the location of turning points may give hints for an optimal mentorship period. Data analysis was performed with R 4.2.3 (R Foundation, Vienna, Austria). Significance threshold was set to 0.05 for all analyses.

The descriptive results comparing patient, operation, and outcomes by surgeon are presented in the supplemental material (Table S1). The results comparing the characteristics collected from the early vs. late learning block cases are presented in Table 1. The complication rates, stratified by each learning block, are depicted in Fig. 1.

The average operative length for Surgeon A was 63.6 (12.4) minutes for cases 1-100, 59.4 (9.5) minutes for cases 101–200, 58.4 (9.5) minutes for the cases 201–300, and 54.7 (12.0) minutes for cases 900-1,000. The average length of operation for Surgeon B was 66.5 (14.1) minutes for cases 1-100, 60.2 (10) minutes for cases 101–200, 60.7 (10) minutes for cases 201–300, and 58 (8) minutes for cases 900-1,000. The average length of operation for Surgeon C was 61.9 (11.3) minutes for cases 1-100, 57.7 (8.8) minutes for cases 101–200, 55.5 (9.5) minutes for cases 201–300, and 47.1 (7.5) minutes for cases 900-1,000. The average length of operation for Surgeon D was 57.6. (8.6) minutes for cases 1-100, 54.9 (12.5) minutes for cases 101–200, 54.1 (10.7) minutes for cases 201–300, and 54.0 (10.8) minutes for cases 900–1000 (Figure S1-S2). On average, it took 1014 days (2.8 years, 2.5–3.4 years) to achieve 1000 Shouldice repairs for primary inguinal hernias.

The slope of the CUSUM curve is positive from cases 1–86, reaching a plateau between cases 87–165, and negative from case 166 onwards. These sections of the CUSUM graph correspond to the learning, competency, and mastery phases. All four surgeons reached a plateau in operating time after 78 to 88 operations.

The univariable effects of the covariates on operating time, as well as the results of the multivariable regression model are summarized in supplementary Table S2. Factors that substantially and significantly increased operating time in multivariable analysis were repair of femoral hernia (tissue repair) (+ 21.0 min, p = 0.003), large hernia size (+ 8.3 min, p < 0.001), male sex (+ 7.9 min, p < 0.001), and use of wire as suture material (+ 5.2 min, p < 0.001). Based on the regression model, the pooled risk-adjusted CUSUM curve was calculated (Fig. 2).

The learning phase thresholds are almost the same, as obtained in the adjusted CUSUM curve. The individual learning curves of each surgeon are depicted in supplementary Figure S1-3.

The 27 postoperative complications (Recurrence: n = 16, Infection: n = 5, Hematoma: n = 4, Seroma: n = 2) were almost equally distributed among the first 300 operations. The logistic regression analysis identified only smoking as a significant patient risk factor (Odds ratio = 14.9, p < 0.001, Table S3).

For the pooled data, the performances of only 4 male surgeons with different levels of experience were used. A sufficient gender distribution, an equal level of experience, and a higher number of surgeons would have led to more clear results. On the other hand, the pattern of the learning curve is similar for all 4 surgeons. The results therefore appear to apply to both experienced and inexperienced surgeons.

As there was no systematic follow-up, surgical site infections and recurrences were only reported based on patients who reported complications. No exact time periods can be given. It can be assumed that some complications were treated in other hospitals without informing the Shouldice Hospital. For that reason, these variables were not used as a primary endpoint.