Associations Between Video Evaluations of Surgical Technique and Outcomes of Laparoscopic Sleeve Gastrectomy

Key Points

Question

Can peer review of surgical video be used to evaluate associations between surgical technique and outcomes?

Findings

This cohort study of 30 surgeons participating in a statewide quality improvement program found that peer reviews of bariatric surgery technique were associated with statistically significant differences in surgical outcomes.

Meaning

Variations in surgical technique, as measured by peer video review, may be associated with consistent and significant differences in patient outcomes.

Abstract

Importance

In any surgical procedure, various aspects of technique may affect patient outcomes. As new procedures enter practice, it is difficult to evaluate the association of each aspect of technique with patient outcomes.

Objective

To examine the associations between technique and outcomes in laparoscopic sleeve gastrectomy.

Design, Setting, and Participants

In this cohort study of bariatric surgery programs participating in a statewide surgical quality improvement collaborative, 30 surgeons submitted intraoperative videos from representative sleeve gastrectomies performed on 6915 patients with morbid obesity. These videos were reviewed by blinded peer surgeons on key technical elements, and 605 reviews were linked to sleeve gastrectomy outcomes of all of the surgeons’ patients from January 1, 2015, to December 31, 2016.

Exposures

Surgeons’ technical approaches to 5 controversial aspects of laparoscopic sleeve gastrectomy: dissection of the proximal stomach, sleeve caliber, sleeve anatomy, staple line reinforcement, and leak testing.

Main Outcomes and Measures

The 30-day outcomes were rate of postoperative hemorrhage and staple line leak. The 1-year outcomes were percentage of total weight lost and reflux severity (Gastroesophageal Reflux Disease Health-Related Quality of Life instrument).

Results

A total of 30 surgeons submitted 46 videos of operations performed on 6915 patients (mean [SD] age, 45.4 [11.7] years; 5494 [79.5%] female; 4706 [68.1%] White). Complete dissection of the proximal stomach was associated with reduced hemorrhage rates (higher ratings for complete mobilization of fundus were associated with a decrease in hemorrhage rate from 2.1% [25th percentile] to 1.0% [75th percentile], P = .01; higher ratings for visualization of the left crus were associated with a decrease in hemorrhage rate from 1.5% to 0.94%, P = .006; and higher ratings for complete division of the short gastrics were associated with a decrease in hemorrhage rate from 2.8% to 1.2%, P = .03). The reduction in hemorrhage rates came at the expense of higher leak rates (higher ratings for complete mobilization of fundus were associated with an increase in leak rate from 0.05% [25th percentile] to 0.16% [75th percentile], P < .001; higher ratings for visualization of the left crus were associated with an increase in leak rate from 0.1% to 0.2%, P = .003; and higher ratings for complete division of the short gastrics were associated with an increase in leak rate from 0.02% to 0.1%, P = .01). Surgeons who stapled more tightly to the bougie had smaller decreases in reflux than those who stapled less tightly (−2.0 to −1.3 on a 50-point scale, P = .002). Staple line reinforcement (buttressing and oversewing) was associated with a small (2 of 1000 cases) decrease in hemorrhage rates. Staple line buttressing was also associated with a similarly small increase in leak rates (1 of 1000 cases). Leak testing was associated with a statistically insignificant change in the staple line leak rate (0.16%-0.22%, P = .47).

Conclusions and Relevance

Variations in surgical technique can be measured by video review and are associated with differences in patient outcomes.

This cohort study examines the associations between surgical technique as reviewed by peers on video and outcomes in laparoscopic sleeve gastrectomy performed on patients with morbid obesity.

Introduction

Surgical technique is often driven by a surgeon’s training, personal preference, and expert opinion—not by high-quality evidence.¹ Because even simple operations involve dozens of technical steps, it is impractical to expect a randomized trial for each step. As a result, novel surgical techniques often enter widespread practice before researchers have had a chance to study them carefully.²

Intraoperative videos offer rich data on the way surgery is performed and may represent a novel way to evaluate surgical technique. Peer reviews of surgical video, for instance, reliably measure skill and can even identify surgeons who have higher rates of surgical complications.³ However, they have never been used to assess the effects of technique as distinct from skill. Although skill refers to how well a surgeon performs a procedure, technique refers to the details of how an operation is conducted, which includes steps of the procedure, the technology used, and the anatomical results intended.⁴ Decisions regarding technique are highly idiosyncratic and surgeon specific.⁵ Perhaps as a result, even once surgeons reach the peak of their learning curves, they still vary widely in their technical skill and outcomes.

New methods of evaluating surgical technique may provide much needed evidence to decrease practice variation and improve patient outcomes. We linked video-based evaluations of surgeons’ technique to their clinical outcomes for sleeve gastrectomy, which recently overtook gastric bypass to become the most common weight loss procedure performed in the US.⁶ Outcomes of sleeve gastrectomy vary widely, and this variation is not fully explained by conventional factors, such as hospital volume and care processes, surgeon experience, or even surgical skill; thus, assessing and standardizing surgical technique may be the remaining opportunity to reduce variation and improve the safety of this common operation.^7,8,9,10,11 Specifically, we tested whether 5 recently recommended procedural steps have any association with the outcome they are thought to affect based on a review of the literature and society guidelines (Table 1).^{12,13,14,15,16,17,18} This novel approach could be applied to any operation in which practicing surgeons’ technique and outcomes vary.

Table 1. Components of Technique Evaluated Through Video Review.

Aspect of operation	Significance	Peer-rated components
Dissection of proximal stomach	Complete dissection of the proximal stomach is thought to improve sleeve anatomy, improve weight loss, and decrease risk of GERD.	Complete mobilization of fundus, visualization of left crus, complete division of short gastric vessels
Sleeve caliber	Sleeve caliber is affected by both how large a bougie is used and how tightly the surgeon follows the bougie. Too tight a sleeve is thought to increase leak and stricture risk; too loose is thought to impede weight loss.	Distance from bougie, bougie size
Sleeve anatomy	Stapling too close to the pylorus and incisura angularis is thought to increase the risk of stricture. Incomplete resection of the gastric fundus is thought to increase reflux and impede weight loss, but excessive dissection can promote ischemia and leaks.	Distance from incisura >3 cm12, distance from pylorus 2-6 cm13,14
Stapling technique	Staple line buttressing is thought to decrease bleeding rate but potentially increase leak rate.15,16,17	Any reinforcement, buttressing, oversewing
Leak test	Staple line leaks are 1 of the leading perioperative sleeve complications. Some simple techniques can test for leaks in the operating room.	Any leak test

Abbreviation: GERD, gastroesophageal reflux disease.

Methods

Data Source and Study Population

For this cohort study, we used 2 complementary sources of data from the Michigan Bariatric Surgery Collaborative (MBSC), a payer-supported statewide quality improvement consortium. First, we used the MBSC registry (an all-payer bariatric-specific registry) to extract detailed information on patients’ baseline characteristics, operations, and outcomes. Second, we used intraoperative videos of laparoscopic sleeve gastrectomies voluntarily submitted by MBSC surgeons and reviewed by their peers. Surgeons were asked to submit representative videos of themselves performing a sleeve gastrectomy for morbid obesity; each surgeon was able to submit 1 or more videos. A total of 30 eligible MBSC surgeons (44.8%) submitted videos, each typically submitting 1 or 2 videos (eTable 1 in the Supplement). Videos were collected from January 1, 2015, to December 31, 2016; edited to remove patient and clinical staff identifiers and focus on critical portions of the case; and uploaded to a Health Insurance and Portability and Accountability Act–compliant portal. Each video was reviewed multiple times by various peer surgeons in a blinded fashion. To ensure consistent ratings, clear and specific definitions of each domain were provided in real time during the video rating process. The study was approved by the institutional review board of the University of Michigan, and surgeons provided written informed consent before participation. All data were deidentified. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.

Exposures

The video review form asked raters to evaluate each video on key technical components thought to affect sleeve gastrectomy outcomes (Table 1). These components were based on a review of the literature, including expert consensus statements establishing optimal technique.^{12,13,14,15,16,17,18} In addition to these aspects of surgical technique, peer reviewers also rated videos on the surgeon’s skill using a version of the Objective Structured Assessment of Technical Skills tailored for bariatric surgery.^3,19

Outcomes

The MBSC registry contains detailed information on patient outcomes obtained through medical record review by trained nurse abstractors using hospital medical records up to 30 days from surgery and from patient surveys obtained at 1, 2, and 3 years after surgery. Centrally trained abstractors review medical records using a standardized and validated instrument; each hospital within the MBSC is audited annually by nurses from the coordinating center to verify that the data are complete and accurate. We reported 30-day and 1-year sleeve gastrectomy outcomes.

30-Day Outcomes

We measured the rates of staple line leak and postoperative hemorrhage as short-term operative complications. Leaks were identified by clinical documentation referring to a gastric leak or by treatment of a leak (percutaneous drain placement or additional operation). Hemorrhage was identified by documentation referring to bleeding or by treatment for hemorrhage (transfusion, splenectomy, or endoscopy). All 6915 patients had 30-day follow-up.

1-Year Outcomes

We measured two 1-year outcomes. First was the percentage of total weight loss at 1 year after surgery. Second was the patient-reported outcome of reflux severity, rated on the validated Gastroesophageal Reflux Disease Health-Related Quality of Life (GERD-HRQL) scale. This scale consists of 10 items graded from 0 to 5, with 0 indicating no symptoms and 5 indicating incapacitating symptoms, for an overall composite score ranging from 0 to 50. We subtracted the overall GERD-HRQL score before surgery from the score at 1 year from surgery to arrive at the change in GERD severity (with more negative values indicating greater improvement). One-year outcomes were calculated only for those patients with 1-year follow-up (4359 patients [63%]).

Statistical Analysis

Within surgeons, peer evaluations of technique varied, with intraclass correlations of 0.4 to 0.9. We thus averaged all reviewers’ evaluations of each surgeon’s technique to arrive at a summary measure of that surgeon’s technical practices. We then divided each technique rating into quartiles to show variations in technique within a realistic distribution (25th and 75th percentiles).

To estimate independent associations between technique components and outcomes, we calculated risk-adjusted outcome rates at the 25th and 75th percentiles of each technique rating. We applied multivariable logistic regression models for binary outcomes (hemorrhage and leak) and multivariable linear regression models for continuous outcomes (percentage of excess weight loss and reflux severity), adjusting for age, sex, race, baseline body mass index (calculated as weight in kilograms divided by height in meters squared), comorbidity burden, smoking status, functional status, and percentage of baseline weight lost preoperatively. We performed a sensitivity analysis that additionally adjusted for surgeon skill. Risk-adjusted outcomes were calculated as marginal means, using robust SEs to account for clustering at the surgeon level. Missing data were handled with case-wise deletion. All analyses were conducted in Stata software, version 15 (StataCorp LLC) using 2-sided tests at the P < .05 significance level.

Results

A total of 30 surgeons submitted 46 videos of operations performed on 6915 patients (mean [SD] age, 45.4 [11.7] years; 5494 [79.5%] female; 4706 [68.1%] White). The videos were reviewed a total of 605 times, and the video reviews were linked to surgeons’ outcomes across their entire panel of patients undergoing sleeve gastrectomy from 2015 to 2016 as represented in the MBSC registry (Table 2). Patients had a mean (SD) body mass index of 47.4 (8.1), and 3589 (51.9 %) had a diagnosis of gastroesophageal reflux disease (GERD). They lost a mean (SD) of 29.1% (9.4%) of their total weight, and GERD severity decreased by a mean (SD) of 1.6 (8.1) on the 50-point scale.

Table 2. Patient Characteristics.

Characteristic	Finding (N = 6915)a
Age, mean (SD), y	45.4 (11.7)
Female sex	5494 (79.5)
White	4706 (68.1)
BMI, mean (SD)	47.4 (8.1)
Commercial insurance	4619 (66.8)
Year of operation
2015	3510 (50.8)
2016	3405 (49.2)
Comorbidities
Diabetes	2133 (30.8)
Hypertension	3500 (50.6)
GERD	3589 (51.9)
Baseline GERD-HRQL score, median (IQR)	3 (0-10)
1-Year GERD-HRQL score, median (IQR)	2 (0-7)
OSA	3110 (45.0)
Outcomes
Leak	16 (0.2)
Hemorrhage	62 (0.9)
TWL, mean (SD), %	29.1 (9.4)
Change in GERD-HRQL score, mean (SD)	−1.6 (8.1)

Abbreviations: BMI, body mass index (calculated as weight in kilograms divided by height in meters squared); GERD, gastroesophageal reflux disease; GERD-HRQL, Gastroesophageal Reflux Disease Health-Related Quality of Life scale; IQR, interquartile range; OSA, obstructive sleep apnea; TWL, total weight loss.

^aData are presented as number (percentage) unless otherwise indicated.

A total of 34 surgeons (69.4%) who submitted reviews practiced in teaching environments, with a median volume of 91 (interquartile range, 27-142) sleeve gastrectomies per year and a mean (SD) skill rating of 3.5 (0.5) on a 5-point scale (eTable 2 in the Supplement). Surgeons varied in their use of the techniques evaluated (Table 3). For instance, the average (50th percentile) surgeon in the study was rated as completely dividing the short gastric vessels in 93% of reviews, but the 25th percentile of this metric was 85% and the 75th percentile was 100%. Surgeons also varied in how tightly they stapled to the bougie: on a 5-point scale from too loose (score of 1) to too tight (score of 5), the 25th percentile score was 2.38 and the 75th percentile was 3.08. In 42 of 46 videos (91.3%), surgeons used some form of staple line reinforcement.

Table 3. Surgeon-Level Technique Patterns.

Technique	Median (IQR)a
Proximal dissection
Complete mobilization of fundus	85 (79-96)
Visualization of left crus	87 (71-96)
Complete division of short gastric vessels	93 (85-100)
Sleeve caliber
Distance from bougie, cm	2.8 (2.4-3.1)
Bougie size, F	34.0 (34.0-36.0)
Sleeve anatomy
Distance from incisura >3 cm	16 (9-19)
Distance from pylorus 2-6 cm	88 (71-92)
Staple line reinforcement
Any reinforcement	100 (100-100)
Buttressing	100 (0-100)
Oversewing	0 (0-0)
Leak testing	88 (75-100)

Abbreviation: IQR, interquartile range.

^aData are presented as percentages unless otherwise indicated.

Figure 1 displays adjusted associations between technique and 30-day outcomes. Higher ratings in any of the 3 domains of complete proximal dissection were associated with significant reductions in hemorrhage rates (higher ratings for complete mobilization of fundus were associated with a decrease in hemorrhage rate from 2.1% [25th percentile] to 1.0% [75th percentile], P = .01; higher ratings for visualization of the left crus were associated with a decrease in hemorrhage rate from 1.5% to 0.94%, P = .006; and higher ratings for complete division of the short gastrics were associated with a decrease in hemorrhage rate from 2.8% to 1.2%, P = .03). However, higher ratings for complete proximal dissection were also associated with higher leak rates (higher ratings for complete mobilization of fundus were associated with an increase in leak rate from 0.05% [25th percentile] to 0.16% [75th percentile], P < .001; higher ratings for visualization of the left crus were associated with an increase in leak rate from 0.1% to 0.2%, P = .003; and higher ratings for complete division of the short gastrics were associated with an increase in leak rate from 0.02% to 0.1%, P = .01). Staple lines beginning 2 to 6 cm from the pylorus were also associated with higher leak rates (from 0.07% to 0.18% of cases, P < .001).

Figure 1. Associations of Surgical Technique With Hemorrhage and Leak Rates.

Error bars indicate 95% CIs. Hemorrhage and leak rates were measured up to 30 days from surgery. Table 1 provides descriptions and definitions of each technique measurement. Statistically significant differences in outcomes between surgeons at the 25th and 75th percentiles for each element of technique are shown.

^aP < .05.

^bP < .01.

^cP < .001.

Increasing use of staple line buttressing was associated with a small but statistically significant decrease in hemorrhage rates (absolute risk reduction of 0.2%, P = .049) and a statistically insignificant increase in leak rates (absolute risk increase of 0.1%, P = .08). Leak testing was associated with a statistically insignificant change in the staple line leak rate (0.16% to 0.22%, P = .47).

On evaluating 1-year outcomes, we found that complete dissection of the proximal stomach was consistently associated with more weight loss (higher ratings for complete mobilization of fundus were associated with a higher percentage of total weight lost from 27.1% [25th percentile] to −28.7% [75th percentile], P = .02; higher ratings for visualization of the left crus were associated with a higher percentage of total weight lost from 27.7% to 28.9%, P = .03; and higher ratings for complete division of the short gastrics were associated with a higher percentage of total weight lost from 24.9% to 28.0%, P = .02).

Higher ratings on the 3 domains of complete proximal dissection were also associated with larger improvements in patient-reported reflux severity (on a 50-point scale consisting of 10 items graded from 0 to 5, with 0 indicating no symptoms and 5 indicating incapacitating symptoms, for an overall composite score ranging from 0 to 50). Specifically, higher ratings for complete mobilization of fundus were associated with a larger decrease in reflux severity from −0.16 [25th percentile] to −1.3 [75th percentile], P < .001; higher ratings for visualization of the left crus were associated with a larger decrease in reflux severity from −0.8 to −1.5, P = .006; and higher ratings for complete division of the short gastrics were associated with a change in direction of reflux severity from 0.8 to −1.0, P = .001) (Figure 2). Bougie size had no association with weight loss or reflux severity. However, surgeons who stapled more tightly to the bougie had smaller decreases in reflux than those who stapled less tightly (from −2.0 to −1.3 on the 50-point scale, P = .002). Stapling greater than 3 cm from the incisura, as currently recommended, was associated with statistically insignificant decreases in reflux (from −1.8 to −2.6 on the 50-point scale, P = .26). Stapling 2 to 6 cm from the pylorus was associated with improved weight loss (27.7% to 28.8% total weight loss, P = .24) and reflux severity (−1.1 to −1.5, P = .48).

Figure 2. Associations of Surgical Technique With Weight Loss and Gastroesophageal Reflux Disease (GERD) Severity.

Error bars indicate 95% CIs. Weight loss and GERD severity were measured at 1 year from surgery. GERD severity was measured with the Gastroesophageal Reflux Disease Health-Related Quality of Life patient-reported outcome instrument. This scale consists of 10 items graded from 0 to 5, with 0 indicating no symptoms and 5 indicating incapacitating symptoms, for an overall composite score ranging from 0 to 50. Table 1 gives descriptions and definitions of each technique measurement. Statistically significant differences in outcomes between surgeons at the 25th and 75th percentiles for each element of technique are shown.

^aP < .05.

^bP < .001.

^cP < .01.

In a sensitivity analysis, the following associations persisted after further adjusting for surgical skill: complete proximal dissection and higher leak rates (higher ratings for complete mobilization of fundus were associated with an increase in leak rate from 0.04% [25th percentile] to 0.16% [75th percentile], P < .001; higher ratings for visualization of the left crus were associated with an increase in leak rate from 0.09% to 0.19%, P = .006; and higher ratings for complete division of the short gastrics were associated with an increase in leak rate from 0.01% to 0.09%, P = .08); staple line 2 to 6 cm from pylorus and an increase in leak rates (from 0.05% to 0.17%, P < .001); tighter staple line to bougie and a decrease in GERD improvement (from −2.0 to −1.3, P = .001); and complete proximal dissection and improvement in GERD (higher ratings for complete mobilization of fundus were associated with a larger decrease in reflux severity from −0.16 [25th percentile] to −1.3 [75th percentile], P < .001; higher ratings for visualization of the left crus were associated with a larger decrease in reflux severity from −0.8 to −1.5, P = .006; and higher ratings for complete division of the short gastrics were associated with a change in direction of reflux severity from 0.8 to −1.0, P = .001). The following associations were attenuated: complete proximal dissection and lower hemorrhage rates, buttressing or oversewing and lower hemorrhage rates, and complete proximal dissection and increased weight loss (eFigure in the Supplement).

Discussion

In this cohort study, there were consistent and significant associations between video-based peer ratings of sleeve gastrectomy technique and surgical outcomes. This study demonstrates that variations in surgical technique may be associated with important differences in patient outcomes.

A complete dissection of the proximal stomach was associated with lower hemorrhage rates and greater weight loss and reflux improvement at the expense of a higher leak rate. The decreased rate of hemorrhage may be the result of identifying and controlling the short gastric vessels while completely dissecting the proximal stomach. The improved weight loss and reflux severity are consistent with the notion that an incomplete dissection increases the risk of retained fundus, which has been associated with worse postoperative hunger control and reflux severity.^20,21 The increase in leak risk was 1 leak per 1000 cases vs increases in weight loss of up to 5% of excess weight and decreases in reflux severity of 1 to 2 points on the 50-point scale associated with complete proximal dissection.

The study also found that stapling more tightly to the bougie was associated with statistically insignificant improvements in weight loss (0.6% total weight loss) at the expense of decreased GERD improvement. Although it may be counterintuitive that a smaller gastric reservoir does not improve weight loss, this approach is mechanistically supported: weight loss from sleeve gastrectomy has more to do with complex physiologic (eg, gastric emptying time) and neurohormonal factors (eg, eliminating ghrelin secretion from the gastric fundus) than with simple volume restriction.^22,23,24 The decreased capacitance of narrower sleeves may thus increase reflux risk without improving weight loss. In addition, the operations studied were performed after a specialty-wide shift toward more consistent bougie sizing, between 34F and 36F, potentially leading the analysis to underestimate the potential association of bougie size with outcomes.^12,14

In addition, the use of staple line reinforcement (buttressing and oversewing) was associated with a small (2 of 1000 cases) decrease in hemorrhage rates. Staple line buttressing was associated with a similarly small increase in leak rates (1 of 1000 cases). Intraoperative leak testing was not associated with lower leak rates. These finding are consistent with prior research^{10,15,16,25,26} using surveys, operative notes, and registry data.

A sensitivity analysis adjusting for surgical skill in addition to the patient factors included in the primary models was performed. The strongest observed associations, such as those pertaining to complete proximal dissection, were robust to this change, but some associations were attenuated. For those techniques for which associations with outcomes were attenuated after adjusting for skill, one possible explanation is that skill was associated with technique; that is, highly skilled surgeons may adhere to more recommended techniques. There may have also been a halo effect that influenced reviewers; for instance, reviewers may have been more likely to assess technique as optimal if watching an apparently skillful surgeon. However, the associations that persisted suggest that technique is an important factor associated with outcomes independent of surgical skill. This sensitivity analysis suggests that the most skilled surgeons do not always adhere to recommended technique and may still have room for improvement by optimizing their technique. Future studies could determine the percentage of outcome variation attributable to technique vs to skill, although both would likely be addressed similarly through peer-to-peer feedback.

Video analysis has several advantages over other methods of reviewing technique. Randomized clinical trials, although the criterion standard, are difficult and time-consuming to conduct in surgery. Moreover, each operation consists of dozens of technical steps—if each step needed to be prospectively and independently evaluated, the pace of surgical innovation would be glacial.^1,2 On the other hand, observational research^27,28 has relied heavily on operative reports and surveys, which are notoriously unreliable. In fact, the perceptual errors that often lead to surgical complications (eg, bile duct injury in laparoscopic cholecystectomy) are rarely self-reported because by definition they arise from surgeon misperception. Video, on the other hand, provides an accurate and unbiased record of intraoperative findings and technique, allowing objective evaluation from fellow expert surgeons.⁴ In laparoscopic surgery, during which critical aspects of each case are routinely captured on camera, obtaining video recordings is straightforward and unobtrusive. This method is also particularly efficient for evaluating multiple aspects of technique simultaneously.

These findings have important implications for surgical quality improvement through peer coaching. On the basis of prior work^{4,29,30,31,32} showing the association between surgical skill and outcomes, multiple efforts are now underway to address skill directly via peer coaching informed by intraoperative video. The findings of this study suggest that video-enabled surgical coaching may also help address variation in surgical technique by identifying areas in which a particular surgeon is not adhering to recommended practice and opportunities for improvement. For any surgical coaching program to succeed, trust and collaborative relationships are crucial because the primary barriers to successful surgical coaching are the surgical culture’s emphasis on autonomy and competency.³³ To overcome these barriers, we recommend that these programs build on existing relationships among surgical colleagues, potentially starting within individual institutions or local surgical societies before scaling more broadly.⁴

Limitations

This study has limitations. First, although it used detailed clinical data for risk adjustment, it is important to avoid causal interpretations of these observational data. Some results may be subject to confounding by indication; for instance, surgeons may use leak testing for patients whom they perceive to be at higher risk for leak, which would lead us to underestimate the association between leak testing and leak rates.

Second, some of the effect sizes noted in this study, although statistically significant, may not be clinically significant. The study reported effect sizes as absolute risk reductions in the interest of clarity, but given that complication rates after sleeve gastrectomy are low at baseline, even a sizable relative risk reduction will appear small in absolute terms. In addition, bariatric surgeons have changed practice in pursuit of similarly small risk reductions—for instance, many surgeons have incorporated staple line reinforcement on the basis of a 1% to 2% absolute decrease in bleeding complications.^{12,14,15,18,34} Further decreasing these complications through video review remains a worthwhile goal.

Third, an assumption in this study was that technique was consistent between the videos submitted and all other operations performed by each surgeon during a 2-year period. Prior evidence confirms that skill evaluations from one operation to another are highly correlated (R = 0.85).³ Like skill, technical decisions appear to be relatively stable over time.⁵

Fourth, evaluations of the same video may vary from reviewer to reviewer, leading to some subjectivity in video ratings. However, this study used the mean of multiple evaluations to account for individual differences in interpretation, and these mean ratings still varied meaningfully at the surgeon level and had significant associations with patient outcomes.

Fifth, although our exposure was surgeon-level technique patterns, we performed a patient-level analysis to perform detailed risk adjustment. However, we accounted for clustering at the surgeon level, reducing effective sample size, to limit the potential for type I error related to the patient-level analysis.

Sixth, we chose to evaluate the independent effect of each aspect of technique and did not measure synergistic or interacting effects among each of these practices. Future studies may explore what combinations of technical practices are associated with optimal outcomes.

Seventh, surgeons self-selected into the video evaluation program and chose which videos to submit for review, so these videos may not be generalizable to the broader population of surgeons and operations. However, characteristics of the study participants mirror those of the broader MBSC community. Surgeons may have submitted videos with particularly exemplary technique and skill, but this would bias our results in favor of the null hypothesis by minimizing variation in the exposure variable.

Conclusions

These findings suggest that complete dissection of the proximal stomach during sleeve gastrectomy is associated with improved hemorrhage rates, weight loss, and reflux improvement at the expense of a higher leak rate and that stapling tightly to the bougie is associated with a small increase in weight loss at the expense of reduced reflux improvement. More broadly, surgical technique, as measured by peers reviewing intraoperative video, appears to have consistent and statistically significant associations with the key outcomes of sleeve gastrectomy. Video analysis thus appears to be a feasible way to evaluate surgical technique. In conjunction with assessments of surgical skill, video-based assessments of surgical technique may help inform surgical coaching and quality improvement.

Supplement.

eTable 1. Overview of Data Collected

eTable 2. Surgeon and Reviewer Characteristics

eFigure. Associations Adjusted for Surgical Skill