Bariatric surgery is the fastest growing surgical subspecialty worldwide. It remains the best long-term treatment for morbid obesity, the prevalence of which is rapidly increasing [1, 2]. Morbid obesity continues to pose serious public health concerns in the United States [3]. Obesity shortens life expectancy and is soon projected to overtake smoking as the leading cause of death [4–6]. It is for this reason that training in bariatric surgery has also expanded, primarily through the creation of 1 year fellowships. Controversy exists regarding the potential benefits or drawbacks of having surgery at a teaching institution. While these centers may see more complex procedures and provide more specialized care, the presence of surgical neophytes may either add to or detract from patient care [7, 8]. Differences in outcomes may be more pronounced with more complex procedures such as laparoscopic Roux-en-Y gastric bypass (LRYGB). A technically challenging procedure with potentially serious morbidity, LRYGB is often performed at “Centers of Excellence,” which may involve surgical fellows. The presence of these advanced trainees may affect patient outcomes.

The effect that fellows have on bariatric outcomes might be expected to vary over the training period. A negative “July Effect” on patient outcomes has been variably demonstrated with the introduction of fresh residents and fellows [9, 10]. For LRYGB, the early learning period may be associated with longer operative times and increased complication rates [11]. Fellowship training is expected to get the fellow over the learning curve, estimated to be between 50 and 100 cases for LRYGB [12]. Completing the learning curve is important as it results in better outcomes and lower complication rates in a surgeon’s independent practice following completion of the fellowship [1, 3, 13, 14]. This has been recognized by the American Society for Metabolic and Bariatric Surgery (ASMBS), the Society of American Gastrointestinal Endoscopic Surgeons (SAGES), and the Fellowship Council (FC) in guidelines for granting bariatric surgery privileges [15]. As fellowship training is increasingly becoming a requirement to practice bariatric surgery, knowing the impact of fellows-in-training on patient outcomes is vitally important.

In this study we investigated the impact that surgical fellows have on postoperative morbidity after LRYGB. We hypothesized that the learning curve might affect outcomes but would be surmounted during fellowship training. Using a nationwide surgical outcomes and quality improvement database, we examined perioperative morbidity and mortality of cases involving fellows compared to those performed by attending surgeons independently. By comparing surgeries with fellows to those without throughout the training period, we analyzed the changing effects of fellowship experience on patient outcomes.

Materials and methods

Data source and inclusion criteria

We conducted a retrospective analysis of the data from the National Surgical Quality Improvement Project (NSQIP) database from 2005 to 2010. Briefly, the NSQIP database is a quality improvement and surgical outcomes research effort on behalf of the American College of Surgeons (ACS). On randomly selected surgical patients, the ACS NSQIP collects 135 clinical data points, including of preoperative comorbidities, intraoperative variables, and 30-day postoperative morbidity and mortality outcomes. Information is limited to the index admission. Specially trained nurses record the data which is audited semiannually. The database is deidentified so this study was exempt from Institutional Review Board approval.

All patients who underwent LRYGB as their primary procedure were identified by the Current Procedural Terminology (CPT) code 43644. Surgeries that involved a postgraduate year (PGY) 6 or higher trainee were included as “fellow” cases. Surgeries without a surgical trainee present were deemed “attending” cases. Surgeries with missing trainee data were excluded, as were any revisional surgeries.

Data and outcomes

Standard demographic data were collected. We also examined the following comorbidities for significant associations and potential confounding: (1) diabetes, (2) hypertension, (3) presence of coronary artery disease [CAD, defined as any of the following: angina, myocardial infarction (MI), previous cardiac angioplasty, or bypass surgery], (4) active pulmonary disease (pneumonia and severe chronic obstructive pulmonary disease, COPD), and (5) evidence of hepatic dysfunction (ascites or esophageal varices).

Because mortality after LRYGB is low, the primary outcome was any postoperative morbidity. Secondary outcomes were surgical site infections (SSIs); pulmonary (i.e., reintubation, prolonged intubation, and pneumonia), cardiac (i.e., cardiac arrest and MI), and thromboembolic events [deep vein thrombosis (DVT) or pulmonary embolus (PE)]; urinary tract infections (UTIs), sepsis, shock, reoperation, and mortality. Outcomes were chosen based on likely association with the variable of interest: presence of trainees during surgery.

Statistical analysis

Mean and median values were used to describe continuous data, with discrete variables displayed as frequencies. For bivariable analyses, two-tailed t tests and Mann-Whitney U tests were used to compare continuous data, while Fisher’s exact or χ 2 tests were used for categorical variables. Multivariable regression controlled for age, body mass index (BMI), American Society of Anesthesiologists (ASA) classification, diabetes, and CAD due to their strong correlation with postoperative morbidity. A priori comparisons were made between attending and fellow outcomes regardless of time of year. To further investigate the potential for confounding by time of year and level of experience, subgroup analysis was performed (1) on the first half of training (July–December), (2) on the second half of training (January–June), and (3) by controlling for the time of year in multivariable regression. A predetermined subgroup analysis compared the outcomes of fellow cases during the first half of training to those of the second half of training. All statistical analysis was performed using Stata SE ver. 12 (StataCorp, College Park, TX).

Results

Population characteristics

There were 38,033 LRYGB cases entered in the NSQIP database from 2005 to 2010. Of these, 949 cases had missing trainee information (2.5 %) and 12,214 involved trainees PGY 1–5; all of these cases were excluded from the study. The final study population was 24,870 LRYGB, consisting of 19,000 (76 %) attending cases, and 5,870 (24 %) fellow cases. The study population was mainly female (79 %) and Caucasian (73 %), with a median age of 45 years (range = 16–90+). The median BMI was 45 kg/m2. The most common comorbidities were hypertension and diabetes in 55 and 25 % of the cases, respectively. Only 3.3 % of the patients had an established diagnosis of CAD. In this population, 61 % of patients had one or more comorbidities and 70 % were given an ASA score of 3 or higher. Median operative times were longer for fellow cases (157 min) compared to attending cases (110 min, p < 0.0001).

Medical comorbidities were similar between groups. Rates of active smoking, alcohol abuse, diabetes, hypertension, cardiac disease, and pulmonary disease were comparable (p > 0.05, see Table 1). Patients in the fellow group had a slightly higher median BMI compared to those in the attending group (45.9 vs. 45.3, p < 0.001). More patients in the attending group had an ASA score ≥3 (70 vs. 67 %, p < 0.001).

Table 1 Population description
Full size table

Outcomes

Our primary outcome, overall morbidity, was seen in 5.3 % of patients. Morbidity was marginally higher for fellow cases (6.2 %) compared to attending cases (5.0 %) (p < 0.001) (Table 2). On adjusted analysis using multivariable regression, fellow cases had a 30 % increased likelihood of morbidity compared to cases where an attending surgeon operated without a trainee (p = 0.001) (Table 3). The most common morbidities were superficial surgical site infections (SSSIs, 1.7 %, n = 414) and UTIs (0.9 %, n = 232). SSSIs occurred slightly more frequently in the fellow group (2.2 vs. 1.5 %, p < 0.001), as did UTIs (1.2 vs. 0.9 %, p < 0.001) and the diagnosis of sepsis (0.9 vs. 0.6 %, p = 0.01). There was a persistent increase in the odds of SSSIs (OR = 1.5, p = 0.01), UTIs (OR = 1.4, p = 0.002), and sepsis (OR = 1.4, p = 0.03) on multivariable analysis. Other findings on adjusted analysis included a higher probability of prolonged ventilation (OR = 1.7, p = 0.03) and pneumonia (OR = 1.6, p = 0.02). Each of the other morbidities, namely, pulmonary, cardiac, thromboembolic, and septic complications, occurred in less than 1 % of patients and was comparable between groups. As expected for LRYGB, overall mortality was low, at 0.2 %. There was no difference in mortality between fellow and attending groups on unadjusted or multivariable analysis.

Table 2 Postoperative morbidity
Full size table
Table 3 Adjusted odds of morbidity associated with fellow participation
Full size table

Subgroups

The effect of fellow participation on outcomes was assessed by separate analysis of outcomes in the first and second halves of a standard fellowship year. During the months of July–December, the fellow group experienced increased overall morbidity (OR = 1.4, p < 0.001), SSSIs (OR = 1.65, p < 0.001), UTIs (OR = 1.66, p = 0.008), and rate of sepsis (OR = 1.74, p = 0.01). Pulmonary outcomes were also worse with respect to prolonged mechanical ventilation (OR = 2.82, p = 0.002) and pneumonia (OR = 1.96, p = 0.005) However, when comparing fellow outcomes to attending outcomes for the January–June period, there were no differences (Fig. 1A, B). To check the robustness of this finding, we controlled for time of surgery in a multivariable model and found that outcomes between attending and fellow groups were similar. When comparing fellow outcomes for the first half versus the second half of the training year, median operative times were similar (159 vs. 154 min, p = 0.8). However, there was a reduction in morbidity rate (7.0 vs. 5.3 %, p = 0.02) from early to later in training (Fig. 2). There was a decrease in the rate of every morbidity, including SSSI, UTI, pneumonia, sepsis, and prolonged ventilation; however, the decreases did not reach statistical significance. Additionally, there was a 25 % reduction in odds of morbidity on adjusted analysis (p = 0.01) (Table 4).

Fig. 1
figure 1

A Odds of morbidity with a fellow involved with LRYGB in the first half of training. B Odds of morbidity with a fellow involved with LRYGB in the second half of training

Full size image
Fig. 2
figure 2

Morbidity in cases involving fellows over time

Full size image
Table 4 Fellows only: odds of morbidity in second half compared to first half of academic year
Full size table

Discussion

Attending versus fellows

The challenge of any institution that trains physicians is to provide exceptional patient outcomes while integrating both the experienced and the novice practitioner into their care. Reassuringly, there was no difference in mortality for patients operated on by fellows when compared to the attending surgeons, in our study. We did, however, find an absolute increase of 1.2 % in morbidity when fellows were involved in surgery compared to when attending surgeons operated without a trainee. Based on these data, 1 in 83 patients would be at higher risk for some morbidity, minor or major, at an institution with fellows. This higher morbidity involves primarily infections such as SSSIs and UTIs. Though not insignificant, these complications are easily treated and would be classified as minor morbidities by most scoring systems. These infections and pneumonias often manifested systemically and resulted in more diagnoses of sepsis. Additionally, there was a higher risk of pulmonary insufficiency resulting in an inability to be weaned off the ventilator; unfortunately, the temporal and therefore causal relationship with pneumonia cannot be determined from these data. The similar rates of shock and bleeding, as well as the overall short stays in the hospital, between attending and fellow groups suggest that morbidity was not major or associated with the requirement of intensive care.

Critics of health care occurring at teaching hospitals argue that there are inevitably more errors by the novice residents and that longer operative times compromise outcomes. Our data confirm this for the early experience of postgraduate fellows. There are several possible mechanisms, all of which likely converge, to explain our findings. The obvious reason is a technical error or break in sterile technique associated with a new practitioner. While fellows are not novices or interns and should be comparable to an attending surgeon in regard to sterile technique, the new training environment or the mental workload of learning a new procedure could result in a lapse in such routine practices [16]. An alternative theory might focus on the longer operative times for surgeries involving fellows. Although prolonged operative time is associated with increased risk of infection, the difference of median operative times between fellows and attending surgeons of 47 min may or may not be the explanation for the findings [17, 18].

In the operating room, a fellow is only a part of the care team. The presence of a fellow at surgery may be an indicator for teaching institutions that not only surgical but also nursing, anesthesia, and perioperative trainees are involved. The presence of other inexperienced practitioners might partially explain the observed increased rates of SSSIs and UTIs. The potential for surveillance bias in teaching institutions may also falsely increase apparent morbidity. As trainees are more likely to overuse diagnostic and laboratory testing, minor morbidities such as SSSIs, UTIS, and bacteriuria may be more commonly observed and treated in their patients [19, 20]. In addition, less experienced clinicians may also be prone to misclassify wound erythema or fat necrosis as a SSSI and thereby falsely raise morbidity rates.

Comparing the outcomes of surgery performed with and without trainees, the literature reaches varying conclusions [2, 5, 7–9]. Some studies demonstrate better outcomes in teaching cases. In a multi-institutional study, Lee et al. [8] found not only a higher rate of perforated appendicitis but also better patient outcomes at teaching hospitals compared to those without trainees. Improved outcomes with trainees may arise from earlier detection and management of complications. “Failure to rescue,” often defined as mortality following a complication, is less common in training institutions [21]. Our data are more consistent with studies positing increased morbidity with trainees. For other complex abdominal procedures, Raval et al. [7] asserted that surgical residents increased the odds of morbidity by 6–26 % but did not influence mortality. Training institutions continue to provide excellent patient care but appropriate supervision remains essential to maintain optimal outcomes.

July effect

If patient outcomes are influenced by trainees, this effect should be most pronounced earliest in training when experience is at its nadir. In our study, fellow participation was associated with an increased risk of morbidity, SSIs, UTIs, pneumonia, sepsis, and ventilator dependence from July to December. Throughout this period, mortality was similar between Fellow and attending groups. The higher morbidity but comparable mortality supports the explanation that trainee-associated morbidity is minor. Alternatively, the presence of a fellow may confer a lower risk of “failure to rescue” once a complication has occurred. Similarly, a recent study of the Nationwide Inpatient Sample exposed an 81 % increased mortality and a twofold risk of intraoperative complications during surgery for spinal metastases in July [22].

Conversely, large studies from the Medicare database and Canada refute the existence a July effect in both medical and surgical patients [23, 24]. Highstead [25] demonstrated equivalent morbidity and mortality between early and late periods in an academic year. Settings with close supervision or where protocols are strictly observed, such as trauma centers, likely reduce the variability of care associated with a novice provider. Surgical training today invariably involves an attending surgeon during the technical procedure. Rigorous oversight and continuity provided by experienced surgeons as well as systems-level interventions may mitigate the morbidity associated with fresh trainees.

Changes with time: learning curve

The technical complexity of LRYGB involves a learning curve during which these patients, with their substantial comorbidities, could experience poor outcomes. Many studies have estimated the experience needed to achieve proficiency in LRYGB. Schauer et al. [12] suggested that the learning curve for LRYGB is 100 cases. Higa et al. [26] detailed declining operating times that stabilized after 100 cases. Oliak et al. [27] demonstrated mortality and conversion rates that declined during the early period, reaching a plateau only after 75 cases. Our data do not permit such a detailed analysis. However, we surmise that the apparent decrease in morbidity in the latter half of the training period is a result of the fellows having surmounted the learning curve.

Our data support the hypothesis that the learning curve is overcome by fellows during their training. Each measured morbidity decreased in frequency from the early to the late period. Fellow cases from January to June had a 25 % reduced likelihood of any morbidity compared to cases from July to December. Despite a uniform trend, there was no statistical difference, except for overall morbidity, because the low incidence of complications precluded adequate statistical power. Fellow cases from January to June had a 25 % reduced likelihood of any morbidity compared to cases from July to December. All outcomes during the second half of the training period were equivalent for fellow and attending cases. These data bolster our conclusion that by the end of a fellowship year, trainees have become safe bariatric surgeons, completely competent at LRYGB.

Limitations

Our study has several limitations. Discrepancies in designating trainee level may have miscategorized senior-level residents who took nonclinical time as fellows. While we cannot estimate the potential this error, our study conclusions are still valid for advanced trainees. Because surgeon and trainee identifiers are not included, we cannot directly compare the outcomes of the same trainees over time. In this study, the temporal analysis assumes that at the end of the academic year trainees are more experienced. This assumption is likely true even if they are not the same trainees as reported in the initial training period. While we provide a multivariable model to adjust for potential confounding, no model can account for all factors; our analysis is limited by unobserved influences on outcomes. Institutional factors could be controlled for by hierarchical modeling, but this was not possible because hospital identifiers are excluded from this data set. In spite of these qualifications, our study employed a large, validated, prospective surgical outcomes data set to demonstrate the morbidity associated with fellow participation in LRYGB and how this effect changes over the training period.

Conclusion

Our study showed that the participation of fellows in laparoscopic Roux-en-Y gastric bypass may be associated with increased postoperative morbidity during the early months of training. This is due mainly to an increased risk of reoperation, superficial surgical wound infections, and urinary tract infections. Examining only the first half of the training year, this effect was prominent, supporting a “July effect” in cases where fellows were present. A decrease in complications between the first and second halves of the year supports an attenuation of this risk over the training period. Furthermore, there were no significant differences in outcome between fellow and attending cases during the second half of the year. These data support the need for continued close supervision of even advanced trainees when learning LRYGB. Fellowship is both necessary and adequate for trainees to overcome the learning curve associated with laparoscopic Roux-en-Y gastric bypass and optimize patient outcomes.