Surgeon Variation in Complications With Minimally Invasive and Open Colectomy: Results From the Michigan Surgical Quality Collaborative | Colorectal Surgery | JAMA Surgery | JAMA Network
[Skip to Navigation]
Sign In
Figure 1.  Variation in Adjusted Complication Rate by Surgeon for Minimally Invasive Colectomy
Variation in Adjusted Complication Rate by Surgeon for Minimally Invasive Colectomy

Dotted line represents cohort mean.

Figure 2.  Variation in Adjusted Complication Rate by Surgeon for Open Colectomy
Variation in Adjusted Complication Rate by Surgeon for Open Colectomy

Dotted line represents cohort mean.

Figure 3.  Difference in Surgeon Ranking for Mean Adjusted Complication Rate in Minimally Invasive and Open Colectomy
Difference in Surgeon Ranking for Mean Adjusted Complication Rate in Minimally Invasive and Open Colectomy

Orange lines represent the top 10 surgeons for open colectomy (lowest adjusted complication rate), and blue lines represent the top 10 surgeons for minimally invasive colectomy.

Table 1.  Patient and Hospital Characteristics for the Minimally Invasive and Open Colectomy Cohortsa
Patient and Hospital Characteristics for the Minimally Invasive and Open Colectomy Cohortsa
Table 2.  Unadjusted Complications by Surgical Approach
Unadjusted Complications by Surgical Approach
Original Investigation
September 2017

Surgeon Variation in Complications With Minimally Invasive and Open Colectomy: Results From the Michigan Surgical Quality Collaborative

Author Affiliations
  • 1Department of Surgery, University of Michigan, Ann Arbor
  • 2Center for Healthcare Outcomes & Policy, University of Michigan, Ann Arbor
  • 3Michigan Surgical Quality Collaborative, University of Michigan, Ann Arbor
JAMA Surg. 2017;152(9):860-867. doi:10.1001/jamasurg.2017.1527
Key Points

Question  How do rates of complications and surgeon ranking compare for minimally invasive colectomy vs open colectomy among surgeons?

Findings  In this cohort study of 97 surgeons and 5196 patients, rates of complications varied nearly twice as much among surgeons for minimally invasive colectomy compared with open colectomy.

Meaning  The study findings imply a need for improved training in adoption of minimally invasive colectomy techniques among some surgeons.

Abstract

Importance  Minimally invasive colectomy (MIC) is an increasingly common surgical procedure. Although case series and controlled prospective trials have found the procedure to be safe, it is unclear whether safe adaptation of this approach from open colectomy (OC) is occurring among surgeons.

Objective  To assess rates of complications for MIC compared with OC among surgeons.

Design, Setting, and Participants  We analyzed 5196 patients who underwent MIC or OC from January 1, 2012, through December 31, 2015, by 97 surgeons in the Michigan Surgical Quality Collaborative, with each surgeon performing at least 10 OCs and 10 MICs. Hierarchical regression was used to assess surgeon variation in adjusted rates of complications and the association of these outcomes across approaches.

Main Outcomes and Measures  Primary study outcome measurements included overall 30-day complication rates, variation in complication rates among surgeons, and surgeon rank by complication rate for MIC vs OC.

Results  Of the 5196 patients (mean [SD] age, 62.9 [14.4] years; 2842 [54.7%] female; 4429 [85.2%] white), 3118 (60.0%) underwent MIC and 2078 (40.0%) underwent OC. Overall, 1149 patients (22.1%) experienced complications (702 [33.8%] in the OC group vs 447 [14.3%] in the MIC group; P < .001). For MIC, the rates of complications varied from 8.8% to 25.9% among surgeons. For OC, rates of complications were higher but varied less (1.7-fold) among surgeons, ranging from 25.9% to 43.8%. Among the 97 surgeons ranked, the mean change in ranking between OC and MIC was 25 positions. The top 10 surgeons ranged in rank from 6 of 97 for OC to 89 of 97 for MIC.

Conclusions and Relevance  Surgeon-level variation in complications was nearly twice as great for MIC than for OC among surgeons enrolled in a statewide quality collaborative. Moreover, surgeon rankings for OC outcomes differed substantially from outcomes for those same surgeons performing MIC. This finding implies a need for improved training in adoption of MIC techniques among some surgeons.

Introduction

Minimally invasive colectomy (MIC) approaches have become common. In the United States, rates of laparoscopic colectomy increased from 2.2% in 1996 to 31.4% in 2009.1 By 2012, laparoscopy had become the most common approach to elective colectomy, accounting for 59.3% of cases.2 Patient safety concerns can arise any time that a new surgical technique is introduced and again as it is widely adopted into surgical practices. Some surgeons may more effectively and safely adopt new techniques, whereas others may be less well trained or experienced with a newer approach yet feel a myriad of pressures to apply the technique.

Although the safety of MIC approaches has been established, it is unclear whether the safe adoption of these techniques is occurring broadly among all surgeons who now perform these complex procedures.3 In addition, colectomy is often treated as a single procedure type in the measurement and reporting of surgical outcomes. It is unclear whether this is appropriate because favorable surgeon outcomes in open colectomy (OC) may or may not be associated with similar outcome rates for MIC approaches. The correlation of surgical approach with surgeon-level outcomes must be better elucidated. If there is not a strong correlation between the best outcomes in OC and the best outcomes in MIC, novel approaches may be needed to help train surgeons in new techniques. Understanding the association between surgeon performance in OC and MIC could better inform surgeons, hospitals, accreditation bodies, and surgical societies in ensuring the safety of patients.

We sought to characterize the association between outcomes in MIC and OC at the surgeon level. We studied patients undergoing MIC and OC to examine the variation in 30-day complication rates across surgeons and evaluate the associations with surgeon performance. We hypothesized that differences in variation in complication rates across surgical approaches exist, implying inconsistencies in the safe adoption of MIC among surgeons.

Methods
Data Source

The Michigan Surgical Quality Collaborative (MSQC) is a 73-hospital consortium that represents diverse practice settings (community and academic) throughout the state. The MSQC data abstraction and data quality assurance details have been described elsewhere.4-6 In brief, specially trained data abstractors prospectively collect information on patient characteristics, intraoperative processes of care, postoperative laboratory results, and 30-day postoperative outcomes for patients undergoing specified general, vascular, and gynecologic surgical operations, using a sampling algorithm that minimizes selection bias. This algorithm divides each calendar year into 46 eight-day cycles from which the first 25 consecutive surgical operations that meet the MSQC inclusion criteria are selected. The cycle rotates every 8 days to ensure that every cycle begins with a different day of the week. The MSQC does not collect data for patients younger than 18 years, those classified as American Society of Anesthesiologists (ASA) class 6, and those undergoing bariatric, trauma, or transplant operations or operations performed within 30 days of another operation captured by the MSQC. Regular data audits ensure registry data validity in accordance with established policies and procedures. Data collection and review for the MSQC are institutional review board exempt at participating hospitals; therefore, no separate informed consent was required for this study.

Patient Population

This study included all adult patients who underwent MIC or OC from January 1, 2012, through December 31, 2015, in the MSQC with surgeons who had a minimum of 10 recorded OCs and 10 MICs each (n = 5196 patients who underwent operations by 97 surgeons at 1 of 46 hospitals). Any laparoscopic, laparoscopy-assisted, robotic, robot-assisted, and single-incision laparoscopic colectomies were classified as MIC. Akin to an intention-to-treat analysis, all cases included in the database as laparoscopic converted to OC and robotic converted to OC were also included in the MIC cohort.

Outcomes

Data were collected on 25 different colectomy-related outcome measures that occurred within 30 days of the operation (corresponding to short-term complications). The primary outcome measure for this study was the occurrence of any complication (grades 1-3) within 30 days after operation. A secondary outcome was the change in surgeon rank for the primary outcome across OC and MIC approaches.

Recorded complications included acute renal insufficiency and/or failure, pneumonia, sepsis, superficial incisional surgical site infection, urinary tract infections, anastomotic leak, deep incisional surgical site infection, deep vein thrombosis that required therapy, organ or space surgical site infection, pulmonary embolism, severe sepsis, unplanned intubation, cardiac arrest that required cardiopulmonary resuscitation, myocardial infarction, stroke or cerebrovascular accident, and mortality.

Independent Variables

The primary exposure variable was the operating surgeon identified through a unique but deidentified variable in the MSQC registry. Patient characteristics, including demographic data and comorbidities, were used as covariates in the logistic regression model that determined complication rates for individual surgeons. Patient demographic data included age, sex, race, insurance type, body mass index, smoking history, functional status, and ASA classification. Comorbidities included preoperative cardiovascular, pulmonary, renal, hematologic, gastrointestinal, and endocrine diagnoses included as independent dichotomous variables. Comorbidity status was obtained from the MSQC registry and was defined by documentation of that condition or its treatment in the medical record. Other covariates included year of operation, patient admission source, and procedure urgency (elective, urgent, or emergency). To explore the degree to which surgeon variation in complication rates correlated with surgical approach, we compared overall complication rates and corresponding surgeon rank for OC with the same measures for MIC.

Statistical Analysis

For each surgeon, we generated risk- and reliability-adjusted rates of overall and serious short-term complications. Multilevel mixed-effects logistic regression was used to evaluate risk factors for postoperative complications. On the basis of backward stepwise selection methods, the final mixed-effects regression models included risk factors that were significant (P < .05) in univariate analysis as fixed patient-level effects in our multivariable model.

To account for variation in surgeon procedure volume, we used 2 approaches. First, we included only surgeons with at least 10 OCs and 10 MICs recorded in the MSQC data. Second, in our multivariable model, the surgeon identifier was incorporated as a random effect to account for clustering of patients by surgeon.7,8 We also performed reliability adjustment, a technique that is useful for surgeons with small numbers of cases whose crude outcome measures may be skewed because of atypical results in the setting of a small sample size. Reliability adjustment works by shrinking the point estimate for complication rate back toward the mean complication rate for the entire cohort, with the degree of shrinkage proportional to the reliability measure of each surgeon.8,9 These methods aim to produce a true measure of surgeon complication rates that allows for more valid comparisons of surgical quality for MIC across surgeons in the MSQC.

Before adjustment, characteristics of patients who underwent OC and MIC were compared using the Pearson χ2 test for categorical variables and the t test for continuous variables. Descriptive statistics were used to characterize the overall incidence of specific postoperative complications. After risk adjustment, variation in adjusted complication rates was assessed for OC and MIC approaches. Then, ranked surgeon complication rates were compared across approach types. Each single surgeon was assigned a rank based on complication rates by approach, and these were then compared. For example, a surgeon might have been ranked 1 of 97 (lowest complication rate overall) for OC but ranked 26 of 97 (26th lowest complication rate overall) for MIC.

Statistical analyses were performed using STATA/SE software, version 13 (StataCorp). All hypotheses were tested using a 2-sided approach with P < .05 considered to be statistically significant.

Results

Of the 5196 patients (mean [SD] age, 62.9 [14.4] years; 2842 [54.7%] female; 4429 [85.2%] white), 3118 (60.0%) underwent MIC and 2078 (40.0%) underwent OC. Overall, 1149 patients (22.1%) experienced complications (702 [33.8%] in the OC group vs 447 [14.3%] in the MIC group; P < .001). Patient and hospital characteristics are given in Table 1. Patients undergoing MIC were younger, with a slightly higher body mass index, more likely to be male, more likely to be white, and more likely to be classified as ASA class I or II. With the exception of statin use, all cardiovascular comorbidities were statistically significantly less common in the MIC cohort. Other endocrine, gastrointestinal, hematologic, oncologic, and alcohol-related comorbidities were less likely in the MIC group. Obstructive sleep apnea was slightly more common in the MIC group (425 [13.6%] vs 218 [10.5%], P = .001). All other pulmonary comorbidities were less common in the MIC group. Patients in the MIC group were less likely to be undergoing dialysis preoperatively and less likely to have a preoperative urinary tract infection.

Patients in the MIC group were also more likely to have private insurance and less likely to be uninsured or have government-sponsored insurance. There was also a much higher percentage of elective MIC cases than for OC (2719 [87.2%] vs 878 [42.3%], P < .001). Patients undergoing MIC were also more likely to have been admitted from home (2777 [89.1%] vs 1025 [49.3%], P < .001). Epidural use was more common in OC (354 [17.0%] vs 315 [10.1%], P < .001).

Differences in specific complications across surgical approaches are listed in Table 2. No statistically significant difference in complication rates was found by approach for catheter-associated bloodstream infection, cardiac arrest requiring CPR, and stroke or cerebral vascular accident, with low event rates in both groups. The rates of individual complications were statistically significantly higher in OC for all other complication types.

Variation in adjusted complication rates for MIC is shown in Figure 1. There was a nearly 3-fold variation across surgeons, with adjusted rates ranging from 8.8% to 25.9%. Similarly, variation in adjusted complication rates for OC is shown in Figure 2. There was 1.7-fold variation among surgeons, with rates ranging from 25.9% to 43.8%.

Next, we compared surgeon ranking for adjusted complication rates in MIC vs OC (Figure 3). Among the 97 surgeons ranked, the mean change in ranking for complications between OC and MIC approaches was 25 positions in the rankings. The top 10 surgeons for OC are highlighted in the figure with darker coloration. These same surgeons, when their performance is assessed for MIC, ranged in rank from 6th to 89th place of 97 surgeons.

Sensitivity Analyses

To ensure that these results were not skewed disproportionately by the morbidity of cases that were converted from MIC to OC, we performed a sensitivity analysis that excluded these cases and found generally similar results (eTable and eFigure 1 in the Supplement). Furthermore, because of the large heterogeneity of case factors in the cohort, we also performed a sensitivity analysis in a more uniform patient cohort (elective partial colectomy with anastomosis: Current Procedural Terminology codes 44140, 44160, and 44204-44205 for surgeons performing ≥5 OCs or MICs). Again, similar results were found with one exception: nearly all of the top 10 surgeons for MIC also fared in the top half of surgeons for OC (eFigure 2 in the Supplement).

Discussion

We found that (1) there is substantial variation in adjusted complication rates for MIC among surgeons in Michigan participating in the MSQC, (2) this variation is nearly twice as wide as the variation in complication rates for OC among these same surgeons, and (3) surgeons who rank highly in patient outcomes for OC do not necessarily rank similarly in outcomes for MIC. These results imply that there is variation in the safe adoption of MIC among surgeons.

Many previous studies10-12 have examined hospital-level outcomes related to surgical approach in colectomy. Fox et al10 found that hospitals with higher rates of laparoscopy had slightly lower postoperative morbidity and length of stay. A multi-institution study12 using Surveillance, Epidemiology, and End Results Medicare data identified a hospital effect for length of stay and in-hospital mortality after laparoscopic colectomy. In addition, a study of colectomy by Xu et al13 found that postoperative complications varied more by surgeon (30.0%) than by hospital (18.2%), suggesting that quality improvement efforts should perhaps be aimed more at the surgeon level. We were able to complement the results of these studies by examining the within-surgeon effects of surgical approach. This type of surgeon-level information is, in large part, less commonly available for study at a multi-institution level and is perhaps more important for quality improvement. The MSQC data set allows us to examine outcomes by surgeon, and in doing this, we identified substantial, clinically significant variation in surgeon-level outcomes.

There are several implications of these findings. We found that variation in complication rates for MIC is 2-fold higher among surgeons in MIC than for these same surgeons for OC despite a much higher risk profile and higher overall complication rate for OC. In addition, we found that surgeons who rank highly in patient outcomes for OC do not necessarily rank similarly in outcomes for MIC. This finding may mean that those surgeons who are the most skilled at OC may be the most experienced in this technique. Perhaps they trained before the advent of MIC approaches or in an environment in which these approaches were rarely used, which would potentially explain the substantial decrease in surgeon ranking for MIC approaches for many of these surgeons who do well with OC. These findings support a potential need for improved training in MIC techniques with the goal of attenuating this observed variation to make it proportionally the same as or less than the variation observed in OC. Many possible approaches to improve surgeon skill after training are being explored, including the use of video-based coaching and feedback.14 However, the best way to train surgeons who are currently in the workforce is still unclear. Moreover, the best entity to facilitate this training remains unclear, although there are many clear stakeholders in improving the safe adoption of new techniques.15,16 As such, a potential remedy to our findings may be improved access to resources for training in the adoption of new surgical techniques.

Limitations

There are several potential limitations in our study. First, these data represent the results of a retrospective cohort study. Inherent in this approach is selection bias and unmeasured confounders. Namely, each surgeon made an individualized decision with regard to their own skill and experience and patient-specific factors on which approach to use. The reasons behind this choice may not be captured in this data set. Although we were not able to eliminate these differences in case mix, we were able to mitigate them using the multiple granular covariates supplied by the MSQC data and the use of hierarchical modeling with reliability adjustment. A particularly poignant supplement to this work would be the performance of interviews with surgeons regarding preferences for surgical approach and reasons for any hesitation to use one approach over the other in certain circumstances. Another possible limitation is that the observed differences for surgeons, by approach, is attributable to differences in case mix. For example, some surgeons may have simply been “dealt a worse hand” in terms of patients with more or worse comorbidities and risk factors. Although the potential for unmeasured confounders always exists, we were able to use a particularly robust and granular data set to adjust for patient-related factors. In addition, because of the deidentified nature of our surgeon data, we did not have access to surgeon-specific training histories (general surgery, colon and rectal surgery, or surgical oncology), which could certainly have an effect on comfort with performing MICs and OCs.

We were similarly not able to adjust for hospital-level effects attributable to hospital type (community, teaching, or academic centers). Although some data sets, such as the National Cancer Data Base, track these center-specific variables, our MSQC data set did not include these data; thus, we were unable to account for these potential effects.

Conclusions

There is substantial variation in adjusted complication rates among surgeons for MIC in Michigan. This variation is proportionally wider than the variation in complication rates for OC among these same surgeons. Last, many surgeons whose patients fared the best for OC did not have similar standing, relative to their peers, for MIC. It is likely that new approaches should be implemented to better train those surgeons who are already skilled in OC to perform newer MIC techniques safely and consistently.

Back to top
Article Information

Corresponding Author: Mark A. Healy, MD, MS, Department of Surgery, University of Michigan, 1500 E Medical Center Dr, Ann Arbor, MI 48109 (healym@umich.edu).

Accepted for Publication: March 12, 2017.

Published Online: June 14, 2017. doi:10.1001/jamasurg.2017.1527

Author Contributions: Dr Healy had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Healy, Kanters, Suwanabol, Varban, Campbell, Dimick, Byrn.

Acquisition, analysis, or interpretation of data: Healy, Regenbogen, Dimick.

Drafting of the manuscript: Healy, Dimick, Byrn.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Healy, Dimick.

Obtained funding: Healy, Dimick.

Administrative, technical, or material support: Dimick, Byrn.

Study supervision: Regenbogen, Campbell, Dimick, Byrn.

Conflict of Interest Disclosures: Dr Dimick is a cofounder of ArborMetrix, a company that makes software for profiling hospital quality and efficiency. No other disclosures were reported.

Funding/Support: Dr Healy received support from grant T32CA009672-25 from the National Institutes of Health. Dr Dimick received grant funding from the National Institutes of Health, the Agency for Healthcare Research and Quality, and the BlueCross BlueShield of Michigan Foundation.

Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

References
1.
Bardakcioglu  O, Khan  A, Aldridge  C, Chen  J.  Growth of laparoscopic colectomy in the United States: analysis of regional and socioeconomic factors over time.  Ann Surg. 2013;258(2):270-274.PubMedGoogle ScholarCrossref
2.
Moghadamyeghaneh  Z, Carmichael  JC, Mills  S, Pigazzi  A, Nguyen  NT, Stamos  MJ.  Variations in laparoscopic colectomy utilization in the United States.  Dis Colon Rectum. 2015;58(10):950-956.PubMedGoogle ScholarCrossref
3.
Abraham  NS, Young  JM, Solomon  MJ.  Meta-analysis of short-term outcomes after laparoscopic resection for colorectal cancer.  Br J Surg. 2004;91(9):1111-1124.PubMedGoogle ScholarCrossref
4.
Campbell  DA  Jr, Englesbe  MJ, Kubus  JJ,  et al.  Accelerating the pace of surgical quality improvement: the power of hospital collaboration.  Arch Surg. 2010;145(10):985-991.PubMedGoogle ScholarCrossref
5.
Campbell  DA  Jr, Kubus  JJ, Henke  PK, Hutton  M, Englesbe  MJ; The Michigan Surgical Quality Collaborative.  The Michigan Surgical Quality Collaborative: a legacy of Shukri Khuri.  Am J Surg. 2009;198(5)(suppl):S49-S55.PubMedGoogle ScholarCrossref
6.
Hendren  S, Fritze  D, Banerjee  M,  et al.  Antibiotic choice is independently associated with risk of surgical site infection after colectomy: a population-based cohort study.  Ann Surg. 2013;257(3):469-475.PubMedGoogle ScholarCrossref
7.
Birkmeyer  NJ, Dimick  JB, Share  D,  et al; Michigan Bariatric Surgery Collaborative.  Hospital complication rates with bariatric surgery in Michigan.  JAMA. 2010;304(4):435-442.PubMedGoogle ScholarCrossref
8.
Dimick  JB, Ghaferi  AA, Osborne  NH, Ko  CY, Hall  BL.  Reliability adjustment for reporting hospital outcomes with surgery.  Ann Surg. 2012;255(4):703-707.PubMedGoogle ScholarCrossref
9.
Dimick  JB, Welch  HG, Birkmeyer  JD.  Surgical mortality as an indicator of hospital quality: the problem with small sample size.  JAMA. 2004;292(7):847-851.PubMedGoogle ScholarCrossref
10.
Fox  JP, Desai  MM, Krumholz  HM, Gross  CP.  Hospital-level outcomes associated with laparoscopic colectomy for cancer in the minimally invasive era.  J Gastrointest Surg. 2012;16(11):2112-2119.PubMedGoogle ScholarCrossref
11.
Lucas  DJ, Ejaz  A, Bischof  DA, Schneider  EB, Pawlik  TM.  Variation in readmission by hospital after colorectal cancer surgery.  JAMA Surg. 2014;149(12):1272-1277.PubMedGoogle ScholarCrossref
12.
Zheng  Z, Hanna  N, Onukwugha  E, Bikov  KA, Mullins  CD.  Hospital center effect for laparoscopic colectomy among elderly stage I-III colon cancer patients.  Ann Surg. 2014;259(5):924-929.PubMedGoogle ScholarCrossref
13.
Xu  T, Makary  MA, Al Kazzi  E, Zhou  M, Pawlik  TM, Hutfless  SM.  Surgeon-level variation in postoperative complications.  J Gastrointest Surg. 2016;20(7):1393-1399.PubMedGoogle ScholarCrossref
14.
Greenberg  CC, Dombrowski  J, Dimick  JB.  Video-based surgical coaching: an emerging approach to performance improvement.  JAMA Surg. 2016;151(3):282-283.PubMedGoogle ScholarCrossref
15.
Birkmeyer  JD, Finks  JF, O’Reilly  A,  et al; Michigan Bariatric Surgery Collaborative.  Surgical skill and complication rates after bariatric surgery.  N Engl J Med. 2013;369(15):1434-1442.PubMedGoogle ScholarCrossref
16.
Pradarelli  JC, Campbell  DA  Jr, Dimick  JB.  Hospital credentialing and privileging of surgeons: a potential safety blind spot.  JAMA. 2015;313(13):1313-1314.PubMedGoogle ScholarCrossref
×