Skip Navigation LinksHome > Blogs > Annals of Surgery Journal Club
Annals of Surgery Journal Club
Interactive resource for surgery residents and surgeons to discuss and critically evaluate articles published in Annals of Surgery selected by a monthly guest expert who will review an article each month, offer questions and respond to reader's comments.
Wednesday, April 01, 2015

Moderator: Mary T. Hawn, MD

Kasotakis G, Lakha A, Sarkar B, et al. Trainee participation is associated with adverse outcomes in emergency general surgery: an analysis of the National Surgical Quality Improvement Program database.Ann Surg. 2014;260:483-493.

Kasotakis and colleagues attempt to define whether surgical trainee participation in emergency general surgery cases results in adverse outcomes. They use the American College of Surgeons National Surgical Quality Improvement Program (NSQIP) data to asses the effect o trainee participation. They matched procedures with and without trainee participation and found that when a trainee was involved, nearly all adverse outcomes measured occurred at a higher frequency. They suggest that much of this may be mediated through the increased operative time associated with resident participation in an operation.

In my opinion, this paper has two major flaws. First, given that no hospital identifiers are provided in the NSQIP participant use file, there is no ability to control for the hospital level effect on post-operative outcomes. Second, the authors made no effort to exclude patients that were transferred to a hospital from a dataset. It is my experience that complicated procedures such as obstructing cancers, multiply recurrent hernias with loss of domain and cholecystitis in the setting of cirrhosis are much more likely to be transferred to a tertiary referral center rather than being managed at a community hospital. While the authors did control for the differences in comorbid conditions that are observed in tertiary care hospitals, there is little ability to control for the challenging surgical presentation that often results in transfer to a higher level of care because of the known attendant higher rates of adverse outcomes.

I would caution any conclusions being drawn from the data presented in this paper. A previous publication from the College found that trainee participation was not associated with adverse outcomes (Ravel et al, J Am Coll Surg. 2011;212:889–898). The investigators at the ACS are able to control for hospital level variation and produce more reliable findings.


  1. Trainee participation in emergency surgical procedures is vitally important for comprehensive surgical training. Based on the findings in this study, should the potential harm from resident participation in an emergency case be addressed in the informed consent?
  2. Given that healthcare reform has placed increasing emphasis on transparence in outcomes and quality, what are potential consequences for academic training programs from this study?
  3. The authors suggest that the potential increase in harm comes from trainee participation, however, could this be a surrogate for the quality of care provided by attending surgeons or academic medical centers?

Tuesday, March 03, 2015

Moderator: Ali Salim, MD

Minei JP, Fabian TC, Guffey DM, et al. Increased trauma center volume is associated with improved survival after severe injury. Ann Surg. 2014;260:456-465.

There has been a long-standing debate on the hospital factors that influence better patient outcomes among trauma centers. While some studies indicate that larger volume centers have better outcomes, others have shown that higher trauma center designation, not volume, predicts better outcomes. Because many previous studies have collected either retrospective clinical data or administrative data with limited clinical information, the authors of this paper decided to use a prospectively collected clinical database to show the relationship between trauma center volume and patient outcomes.

The study was a secondary analysis of two completed randomized controlled clinical trials performed by the Resuscitation Outcomes Consortium: one enrolling patients that had hypovolemic shock and the other enrolling patients with severe traumatic brain injury (TBI). In both trials, the interventions were in three arms: pre-hospital hypertonic saline, hypertonic saline with 6% dextran 70, or normal saline. For this study, the patients were analyzed based on the hospitals they were admitted to. Trauma center volume was categorized as annual patient volume in increments of 1000 and trauma center designation was categorized as Level I or Level II. The primary outcomes were 24-hour and 28-day mortality, while secondary outcomes that measured hospital events and long-term functional outcomes were also analyzed. Variables that measured injury severity, pre-hospital care, and admission physiology were recorded. Various models were used to identify hospital factors that predict better outcomes while adjusting for potential confounders.

After adjusting for potential confounders, there was no difference in 24-hour mortality in either the shock cohort or the TBI cohort based on trauma center designation or volume. Analyzing all the patients together, however, showed that increased volume was associated with lower 24-hour mortality. A similar analysis of 28-day mortality showed the same association when all patients were combined. But this association only persisted in the TBI cohort on stratified analysis. Interestingly, Level I trauma center status was associated with reduced 28-day mortality in the shock cohort even after adjusting for hospital volume. When other secondary outcomes such as ventilator-free days and Multiple Organ Dysfunction were analyzed, higher volume seemed to predict better outcomes. However, higher volume centers had higher complication rates. Trauma center status did not seem to influence outcomes significantly.

The strengths of this study include the detail of clinical information that was collected prospectively, the long-term follow up of patients and the careful selection of patients. However, some limitations exist. Although the different arms of treatment in the original randomized trial were associated with different risks of mortality, the authors did not consider treatment arms in the analyses. In addition, only 11 centers were included in the study and individual hospital factors aside from volume and trauma center designation could easily bias the results of the study even after accounting for within-hospital correlation. The authors concluded that higher volume centers are more likely to have better systems to provide trauma care. Therefore, to optimize outcomes in patients with severe injuries, they should preferentially be taken to higher volume centers.


  1. Based on the results of this study, should hospitals focus on increasing their volume instead of improving their practices and procedures?
  2. 2.
  3. If the protocols demand directing severe injuries to higher volume centers, is there a potential for these centers to get overwhelmed by the volume of patients? In other words, is there a maximum volume threshold for improved outcomes?
  4. 3.
  5. Would adjusting for other hospital factors have changed the results?
  6. 4.
  7. How generalizable are the results? The practices and procedures involved, within the included trauma centers, to perform a high quality parallel randomized prospective trial may not be applicable to other trauma centers.

Thursday, February 05, 2015

Moderator: Steven M. Strasberg, MD

Featured Article: Uzunoglu FG, Ree M, Vettorazzi E, et al. Preoperative Pancreatic Resection (PREPARE) Score: A Prospective Multicenter-Based Morbidity Risk Score. Ann Surg 2014;260:857–864.


The aim of the study was to develop a predictive score for the development of complications after pancreatic surgery. 703 patients from 8 high volume pancreatic surgery centers were used to develop the model. Multiple potential risk factors such as serum albumin, weight, preoperative albumin were examined and the outcome was postoperative complications evaluated by the Clavien-Dindo complication severity grading system. This system has multiple levels but in this study complications were broken up into two groups, minor (grade 1-2) and major (grades 3-5). Thus major complications range from invasive procedures to postoperative death. The model was developed by multivariate analysis using major complications as the dependent variable and the potential risk factors as the independent variables. All the risk factors were measurable preoperatively. Eight risk factors were identified. Using these a scoring system was developed which categorized risk as low, intermediate, or high. A prospective validation cohort of 429 patients was used to validate the model and it performed with an accuracy of 78% in its ability to predict whether the risk of developing major complications was low, intermediate or high. The C statistic was 0.709.


1. This scoring system was developed from eight high volume centers. Would it be applicable to low volume centers? Would it be accurate if surgeons in low volume centers to use this system to predict the risk of major complications in their patients?

2. What is the C-statistic and how is it used? Based on the C-statistic would this model be considered to be “weak”, “reasonable” or “strong”?

3. In explaining risk of complications to patients in the preoperative period the surgeon has an obligation to provide the outcomes a)in his/her practice b) in the hospital c) in the literature d) predicted by a calculator. (choose none or one or more and explain).

4. What is the ACS-NSQIP risk calculator and how does it differ from this system?

Monday, January 05, 2015

Moderator: Justin B. Dimick, MD, MPH

Featured Article: Brown EG, Burgess D, Li CS, et al. Hospital readmissions: necessary evil or preventable target for quality improvement. Ann Surg. 2014;260:583-589.


This study uses the University HealthSystem Consortium (UHC) database to study patterns of readmission after cancer surgery. This analysis yields several interesting findings that potentially challenge whether readmissions are a valid quality metric for surgery. First, they find that high volume cancer centers and designated cancer centers have higher readmission rates. Since these centers are known to have better clinical outcomes (e.g., mortality and morbidity) in prior studies, this finding raises questions about whether readmission is a proxy for high quality rather than poor quality care. Second, the authors found that “a minority” of early readmissions (within 7 days) were due to “potentially preventable causes”, such as nausea, vomiting, dehydration, and pain (accounting together for 33% together). The biggest cause of readmissions was infectious complications (46%), which the authors deemed non-preventable. The authors therefore conclude that the majority of readmissions are not preventable (or modifiable), suggesting further that they are not a good measure of quality.


(1) Do you agree with the authors that readmissions should be correlated with other quality metrics such as morbidity and mortality? Or can readmissions represent a unique and important domain of quality, even if they do not correlate with traditional metrics?

(2) Do you agree with the authors that infectious complications are not preventable? Their argument that a “minority” of readmissions were due to potentially preventable causes hinges on this assumption.

(3) The authors raise the point that a readmission may sometimes represent high quality care, essentially “rescuing” the patient from downstream complications including death. How could a readmission metric be designed so it would not penalize hospitals for readmitting a patient who needs it?

Please feel free to comment on any or all of the questions. We look forward to hearing from you, the Annals readers. This article can be accessed for free.

Monday, December 01, 2014
Moderator: David Efron, MD
Featured Article: Neal M et al. Prehospital Use of Nonsteroidal Anti-inflammatory Drugs (NSAIDs) Is Associated With a Reduced Incidence of Trauma-Induced Coagulopathy. Ann Surg 260: 378-382, 2014.

Trauma induced coagulopathy (TIC) remains a challenge to the care of the severely injured patient.  Whether it is the result of a large burden of tissue damage or secondary to acute blood loss and the need for massive resuscitation, it is directly associated with clinical outcome morbidities and mortalities.  To date inflammation-modulating therapies have been non-existent, not only because of the inability to predict the onset of injury and septic insult, but also because of the unclear link between the anti-inflammatory medications and the subsequent response to injury.  This study attempts to explore one such potential association.  Identifying a link between the inflammation blunting effects of non-steroidal anti-inflammatory medications (NSAIDs) and the emergence of trauma induced coagulopathy is an attractive prospect for a potentially significant, relatively simple and inexpensive therapeutic intervention.

The Inflammation and the Host Response to Injury Large Scale Collaborative Program remains one of the most comprehensive compilations of data yet available to study questions such as these.  In a multicenter effort, clinical, proteomic and genomic data were prospectively collected regarding adults (aged 18-90) who suffered severe blunt injury.  Patients at the various centers were managed according to strict clinical protocols to minimize treatment variation and clinical history includes pre-injury, injury, in hospital and outcome variables. 

The goal of the study was to analyze patients for an association between pre-injury medication use and the incidence of TIC.  Coagulopathy was defined in two ways: 1) TIC was defined as an admission INR of >1.5 (which was the study’s primary outcome measure) and 2) a clinically relevant coagulopathy defined as transfusion of more than 2 units of fresh frozen plasma or 1 unit of platelets within the first 6 hours of admission.   Medications assessed and recorded as taken or not taken included NSAIDs, aspirin, other antiplatelet agents, beta-blockers, angiotensin receptor blockers,  ACE inhibitors, vasodilators, diuretics, calcium channel blockers, other anti-hypertensive medications (exclusive of the above list), oral contraceptives, amphetamines, statins, corticosteroids and cocaine. 

Spanning the years 2003 through 2010, of the 2007 patients who were captured in the study, 72 were excluded for pre-existing liver disease and 38 were excluded for preadmission warfarin use (the only exclusion criteria in this complete data set).  Of the 1897 patients analyzed, the majority (1455; 76.7%) were not on any pre-injury medications.  The remaining 442 patients were on at least one of the listed medications.  The three most common medication classes in decreasing order were non-specified anti-hypertensives, statins and beta-blockers.  NSAIDs use was identified in 47 patients (10.6% of those taking medications and 2.5% of all patients studied).  Under rigorous multivariate analysis, pre-injury NSAID use was associated with a 72% lower risk of TIC identified on admission and a 66% lower risk of clinically relevant coagulopathy.  NSAIDs were the only medications to demonstrate any beneficial effect and this remained significant when controlling for comorbidities associated with NSAID use.  Interestingly, neither NSAIDs nor other antiplatelet medications were associated with TIC or clinically defined coagulopathy.

The limitations of the study include the retrospective analysis, the relatively low numbers of patients on the listed medications and the inability to assure that patients were actually taking those medications. These are balanced against the high data fidelity of this program and the rigorously applied statistical analysis.  The authors conclude that pre-injury NSAID use was associated with a decreased incidence of early elevated INR or clinically relevant coagulopathy. 


1)      The authors do not report an association between medication use and mortality.  Are the end points of defined trauma induced coagulopathy or clinically relevant coagulopathy as defined in this study valid? 

2)      Is it important to know which specific NSAIDs were being taken?

3)      Should such data be specifically censored for early transfusion secondary to surgically relevant blood loss (non-coagulopathic bleeding as may occur from a splenic injury requiring splenectomy)?

4)      Should we be considering early administration of NSAID medication in the field prior to transport to the hospital?

Please feel free to comment on any or all of the questions. We look forward to hearing from you, the Annals readers. This article can be accessed for free.




About the Author

Dr. Mary T. Hawn
Dr. Mary T. Hawn is Professor of Surgery, Chief of the Division of Gastrointestinal Surgery and Vice Chair for Quality and Clinical Effectiveness at the University of Alabama at Birmingham. Her research interests are in the area of safety and effectiveness of surgical procedures and their policy implications.