Kevin Y. Pei

Comparison of Internal Medicine and General Surgery Residents' Assessments of Risk of Postsurgical Complications in Surgically Complex Patients

Importance Anticipating postsurgical complications is a vital physician skill, particularly when counseling surgically complex patients on their risks of intervention. Although internists and surgeons both counsel patients on surgical risks, it is uncertain who is better equipped to accurately anticipate surgical complications. Objective To examine how internal medicine and general surgery trainees compare in their assessment of risk of surgically complex patients. Design, Setting, and Participants General surgery and internal medicine residents (urban, tertiary, and academic medical center) answered an anonymous, online assessment of 7 real-life, complex clinical scenarios. Participants estimated the chance of any morbidity, mortality, surgical site infection, pneumonia, and cardiac complications. Scenarios represented a diverse general surgery practice, including colectomy, duodenal ulcer repair, inguinal hernia repair, perforated viscus exploration, small-bowel resection, cholecystectomy, and mastectomy in surgically complex patients likely to be comanaged by surgical and internal medicine services. Main Outcomes and Measures Responses were compared with risk-adjusted outcomes reported by the American College of Surgeons’ National Surgical Quality Improvement Project (NSQIP) online calculator. Results A total of 76 general surgery residents (50 [65.8%] male and 26 [34.2%] female) and 76 internal medicine residents (36 [47.4%] male and 40 [52.6%] female) participated (64% overall response rate). General surgery residents were significantly more confident with their responses (general surgery residents’ mean response, 3.6 [95% CI, 3.4-2.8]; internal medicine residents’ mean response, 2.8 [95% CI, 2.6-3.0]; P < .001) and with not offering operations (general surgery residents’ mean response, 4.3 [95% CI, 4.1-4.4]; internal medicine residents’ mean response, 3.7 [95% CI, 3.4-3.9]; P = .006) but less likely to discuss code status (general surgery residents’ mean response, 3.2 [95% CI, 2.9-3.4]; internal medicine residents’ mean response, 3.8 [95% CI, 3.5-4.1]; P < .001) or consult risk-adjusted models, such as NSQIP (general surgery residents’ mean response, 2.9 [95% CI, 2.7-3.1]; internal medicine residents’ mean response, 3.7 [95% CI, 3.4-4.0]; P < .001). For 91% of clinical estimates, both groups similarly overestimated every type of risk; in 9% of estimates, internal medicine residents had higher overestimates. Estimates varied significantly, with wide 95% CIs; however, only 11% of the NSQIP estimates fell within the 95% CIs. Overall, the mean percentages of the estimates ranged from 26% to 33% over NSQIP estimates for all complications. Conclusions and Relevance General surgery and internal medicine residents demonstrated similar estimates of postoperative complications and death. Both groups overestimated risks in surgically complex patient scenarios compared with NSQIP risk calculator estimates. This near-universal overestimation of risk underscores the importance of developing risk-estimation resources for internists and surgeons.

Association of Lowering Default Pill Counts in Electronic Medical Record Systems With Postoperative Opioid Prescribing

Importance Reliance on prescription opioids for postprocedural analgesia has contributed to the opioid epidemic. With the implementation of electronic medical record (EMR) systems, there has been increasing use of computerized order entry systems for medication prescriptions, which is now more common than handwritten prescriptions. The EMR can autopopulate a default number of pills prescribed, and 1 potential method to alter prescriber behavior is to change the default number presented via the EMR system. Objective To investigate the association of lowering the default number of pills presented when prescribing opioids in an EMR system with the amount of opioid prescribed after procedures. Design, Setting, and Participants A prepost intervention study was conducted to compare postprocedural prescribing patterns during the 3 months before the default change (February 18 to May 17, 2017) with the 3 months after the default change (May 18 to August 18, 2017). The setting was a multihospital health care system that uses Epic EMR (Hyperspace 2015 IU2; Epic Systems Corporation). Participants were all patients in the study period undergoing 1 of the 10 most common operations and discharged by postoperative day 1. Intervention The default number of opioid pills autopopulated in the EMR when prescribing discharge analgesia was lowered from 30 to 12. Main Outcomes and Measures Linear regression estimating the change in the median number of opioid pills and the total dose of opioid prescribed was performed. Opioid doses were converted into morphine milligram equivalents (MME) for comparison. The frequency of patients requiring analgesic prescription refills was also evaluated. Results There were 1447 procedures (mean [SD] age, 54.4 [17.3] years; 66.9% female) before the default change and 1463 procedures (mean [SD] age, 54.5 [16.4] years; 67.0% female) after the default change. After the default change, the median number of opioid pills prescribed decreased from 30 (interquartile range, 15-30) to 20 (interquartile range, 12-30) per prescription (P < .001). The percentage of prescriptions written for 30 pills decreased from 39.7% (554 of 1397) before the default change to 12.9% (183 of 1420) after the default change (P < .001), and the percentage of prescriptions written for 12 pills increased from 2.1% (29 of 1397) before the default change to 24.6% (349 of 1420) after the default change (P < .001). Regression analysis demonstrated a decrease of 5.22 (95% CI, −6.12 to −4.32) opioid pills per prescription after the default change, for a total decrease of 34.41 (95% CI, −41.36 to −27.47) MME per prescription. There was no statistical difference in opioid refill rates (3.0% [4 of 135] before the default change vs 1.5% [2 of 135] after the default change, P = .41). Conclusions and Relevance Lowering the default number of opioid pills prescribed in an EMR system is a simple, effective, cheap, and potentially scalable intervention to change prescriber behavior and decrease the amount of opioid medication prescribed after procedures.

Model for End-Stage Liver Disease Underestimates Morbidity and Mortality in Patients with Ascites Undergoing Colectomy

BackgroundThe Model for End-Stage Liver Disease (MELD) score and ascites correlate with surgical morbidity and mortality. However, the MELD score does not account for ascites. We sought to evaluate whether the MELD score accurately risk stratifies patients with ascites.MethodsWe analyzed the American College of Surgeons National Surgical Quality Improvement Program (2005–2014) to examine the risk-adjusted morbidity and mortality of cirrhotic patients with and without ascites undergoing colectomy for diverticulitis. Patients were stratified by MELD score, and the presence of ascites and outcomes were compared between patients with and without ascites to the reference group of low MELD and no ascites. Multivariable logistic regression was used to control for demographic factors and comorbidities.ResultsA total of 16,877 colectomies were analyzed. For each MELD stratum, patients with ascites have increased risk of complications compared to those without ascites (P < 0.05 unless indicated): low MELD ascites OR 1.13, P = 0.69, moderate MELD no ascites OR 1.37, moderate MELD ascites OR 2.06, high MELD no ascites OR 1.93, and high MELD ascites OR 3.54. These trends hold true for mortality: low MELD ascites OR 2.91, P = 0.063, moderate MELD no ascites OR 1.47, moderate MELD ascites OR 5.62, high MELD no ascites OR 3.04, and high MELD ascites OR 9.91.ConclusionAscites predicts an increased risk for postoperative morbidity and mortality for cirrhotic patients undergoing colectomy for all MELD classifications. These findings suggest that the MELD score significantly underestimates postoperative risk as it does not account for ascites.

Recurrent Falls Among Elderly Patients and the Impact of Anticoagulation Therapy

BackgroundFalls are the leading source of injury and trauma-related hospital admissions for elderly adults in the USA. Elderly patients with a history of a fall have the highest risk of falling again, and the decision on whether to continue anticoagulation after a fall is difficult. To inform this decision, we evaluated the rate of recurrent falls and the impact of anticoagulation on outcomes.MethodsAll patients of age  ≥ 65 years and hospitalized for a fall in the first 6 months of 2013 and 2014 were identified in the nationwide readmission database, a nationally representative all-payer database tracking patient readmissions. Readmissions for a recurrent fall within 6 months, and mortality and bleeding injuries (intracranial hemorrhage, solid organ bleed, and hemothorax) during readmission were identified. Logistic regression evaluated factors associated with mortality on repeat falls.ResultsOf the 331,982 patients admitted for a fall, 15,565 (4.7%) were admitted for a recurrent fall within 6 months. The median time to repeat fall was 57 days (IQR 19–111 days), and 9.0% (1406) of repeat fallers were on anticoagulation. The rate of bleeding injury was similar regardless of anticoagulation status (12.8 vs. 12.7% not on anticoagulation, p = 0.97); however, among patients with a bleeding injury, those on anticoagulation had significantly higher mortality (21.5 vs. 6.9% not on anticoagulation, p < 0.01).ConclusionAmong patients hospitalized for a fall, 4.7% will be hospitalized for a recurrent fall within 6 months. Patients on anticoagulation with repeat falls do not have increased rates of bleeding injury but do have significantly higher rates of death with a bleeding injury. This information is essential to discuss with patients when deciding to restart their anticoagulation.

Revised cardiac risk index poorly predicts cardiovascular complications after adhesiolysis for small bowel obstruction

Background: The number of patients undergoing preoperative risk stratification in the United States is expected to increase as the population ages. A large percentage of patients undergo some form of preoperative testing, and society guidelines suggest that up to 50% of the testing in lower risk surgical subgroups is unnecessary. The Revised Cardiac Risk Index and the risk calculator of the American College of Surgeons National Surgical Quality Improvement Program are widely used tools as the first step of preoperative cardiac evaluation. The Revised Cardiac Risk Index was developed to fill a need for objective perioperative cardiac risk evaluation. Despite the ease of use of Revised Cardiac Risk Index, it is uncertain if the stratification is accurate for surgical patients because its accuracy in large surgical samples has not been tested. With the National Surgical Quality Improvement Program risk calculator having excellent accuracy in estimating cardiac complications (area under the receiver operating characteristic 0.895), a unique opportunity to test the predictive accuracy of postsurgical cardiac events became available. The purpose of this study is to determine the accuracy of the Revised Cardiac Risk Index for predicting cardiovascular complications after adhesiolysis for small bowel obstruction. Methods: From 2005 to 2015, 34,032 cases of open or laparoscopic adhesiolysis (Current Procedural Terminology codes 44005 and 44180) for small bowel obstruction (International Classification of Diseases, 10th edition [ICD‐10]) were analyzed using the National Surgical Quality Improvement Program dataset. Revised Cardiac Risk Index estimates were calculated for each case and compared to reported cardiovascular complications (myocardial infarction or cardiac arrest) using univariable logistic regression. Overall predictive accuracy was assessed by measuring model discrimination (area under the receiver operating characteristic) and model calibration (Hosmer‐Lemeshow chi‐squared statistics). Results: Although the Revised Cardiac Risk Index predicted cardiovascular complications with an odds ratio of 2.3 and a 95% confidence interval of 1.9 to 2.8 (P < .001) and the Hosmer‐Lemeshow chi‐square was significant (0.22, P = 0.64), the area under the receiver operating characteristic was poor (0.63, 95% confidence interval 0.59–0.67). Conclusion: Despite its relative simplicity, the Revised Cardiac Risk Index performed poorly as a predictor of cardiovascular complications after adhesiolysis for small bowel obstruction. These findings question the utility of the Revised Cardiac Risk Index in this patient population. Future studies should aim to develop models that are computationally simple while retaining predictive accuracy.

Ascites: A marker for increased surgical risk unaccounted for by the model for end-stage liver disease (MELD) score for general surgical procedures

Background: Ascites and the Model for End‐Stage Liver Disease score have both been shown to independently correlate with surgical morbidity and mortality. We evaluated if incorporating the presence of ascites changed postoperative risk as assessed by the Model for End‐Stage Liver Disease score. Methods: Data originated from the National Surgical Quality Improvement Program database from 2005−2014. Patients undergoing hernia repair, adhesiolysis, and cholecystectomy were included. Univariate analysis and logistic regression stratified by Model for End‐Stage Liver Disease score and presence of ascites was performed. Results: A total of 30,391 patients were analyzed. When compared to low Model for End‐Stage Liver Disease stratum without ascites, the presence of ascites predicted increased risk for complications (low Model for End‐Stage Liver Disease with ascites odds ratio 3.22, 95% confidence interval [2.00−5.18], moderate Model for End‐Stage Liver Disease with ascites odds ratio 3.70, 95% confidence interval [2.64−5.19], high Model for End‐Stage Liver Disease with ascites odds ratio 6.38, 95% confidence interval [4.39−9.26]). These findings hold true for mortality as well (low Model for End‐Stage Liver Disease with ascites odds ratio 9.40 95% confidence interval [3.53−25.01], moderate Model for End‐Stage Liver Disease with ascites odds ratio 15.24 95% confidence interval [8.17−28.45], high Model for End‐Stage Liver Disease with ascites odds ratio 28.56 95% confidence interval [15.43−52.88]). Conclusions: Ascites increased the risk of morbidity and mortality across multiple general surgery operations. Model for End‐Stage Liver Disease may underestimate surgical risk in patients with ascites. Predictive models inclusive of ascites may more accurately predict the perioperative risk of these complex patients.

Evaluating the adoption of primary anastomosis with proximal diversion for emergent cases of surgically managed diverticulitis

Background: Although Hartmann procedure is common for operatively managed acute diverticulitis, there is accumulating evidence that primary anastomosis with proximal small bowel diversion is safe, even in emergent cases. This study seeks to clarify the current adoption of primary anastomosis with proximal small bowel diversion among emergent, operatively managed cases of acute diverticulitis and compare outcomes between primary anastomosis with proximal small bowel diversion and Hartmann procedure. Methods: Patients who underwent open, emergent Hartmann procedure or primary anastomosis with proximal small bowel diversion for a primary diagnosis of diverticulitis between 2005 and 2015 were identified in the database of the American College of Surgeons National Surgical Quality Improvement Program. Outcomes were compared with logistic regression adjusted for patient and operative characteristics. Results: From 2005–2015 the proportion of primary anastomosis with proximal small bowel diversion decreased from 33% to 17% among emergent cases. Although mortality and complications were similar, primary anastomosis with proximal small bowel diversion resulted in a greater risk of returning to the operating room in emergent cases (odds ratio = 1.35, 95% confidence interval: 1.06–1.74). Conclusion: Despite previous suggestions of clinical equipoise, the adoption of primary anastomosis with proximal small bowel diversion for emergent, operatively managed acute diverticulitis among National Surgical Quality Improvement Program hospitals appears to be decreasing. Primary anastomosis with proximal small bowel diversion resulted in increased return to the operating room for emergent cases, suggesting that caution should be exercised in selecting primary anastomosis with proximal small bowel diversion for emergent cases.

Surgeons overestimate postoperative complications and death when compared with the National Surgical Quality Improvement Project risk calculator

BACKGROUND The assessment of postoperative morbidity and mortality is difficult particularly for complex patients. We hypothesize that surgeons overestimate the risk for complications and death after surgery in complex surgical patients. MATERIALS AND METHODS General surgery residents and attending surgeons estimated the likelihood of any morbidity, mortality, surgical site infection, pneumonia, and cardiac complications for seven complex scenarios. Responses were compared with the American College of Surgeons National Surgical Quality Improvement Project Surgical Risk Calculator. RESULTS From 101 residents and 48 attending surgeons, overall response rate was 61.7%. For all seven clinical scenarios, there was no difference between resident and attending predictions of morbidity or mortality, with significant variation in estimates among participants. Mean percentages of the estimates were 25.8%-30% over the National Surgical Quality Improvement Project estimates for morbidity and mortality. CONCLUSIONS General surgery residents and attending surgeons overestimated risks in complex surgical patients. These results demonstrate broad variance in and near universal overestimation of predicted surgical risk when compared with national, risk-adjusted models.

A simple predictor of post-operative complications after open surgical adhesiolysis for small bowel obstruction

BACKGROUND Small bowel obstruction is common and often requires surgical management. Simple preoperative models are lacking to predict post-operative complications after surgical management of adhesive small bowel obstruction. METHODS We retrospectively analyzed data from 15,036 patients who underwent open lysis of adhesions for small bowel obstruction from 2005 to 2013 using the American College of Surgeons (ACS) National Surgical Quality Improvement Program (NSQIP) database. Predictors of post-operative complications were identified using logistic regression. Predictive models were compared using areas under the receiver operating characteristic curves (AUROC). RESULTS A three-parameter model was constructed, termed FAS: Functional status, American Society of Anesthesiologists (ASA) classification, and prior Sepsis. FAS predicted post-operative complications with odds ratio (OR) 1.11, 95% CI (1.10, 1.12), P < 0.001 and AUROC of 0.69, 95% CI (0.67, 0.70). CONCLUSIONS FAS predicts post-operative complications after open lysis of adhesions using three readily available clinical parameters.

Caring for the Geriatric Combat Veteran at the Veteran Affairs Hospital

Purpose of ReviewThe US population continues to grow older, and their needs pose a challenge to the healthcare system. The nation’s aging veterans are no exception to this trend.Recent FindingsThe geriatric patient is physiologically distinct from younger adults. Geriatric veterans are unique in terms of their social history and the illnesses they risk encountering. Veterans of our recent conflicts will in the decades to come also have their own unique needs as they grow older that are yet to be fully understood.SummaryIn this review, we discuss several conditions that clinicians who care for geriatric veterans may expect to encounter.