Use of National Burden to Define Operative Emergency General Surgery

Importance  Emergency general surgery (EGS) represents 11% of surgical admissions and 50% of surgical mortality in the United States. However, there is currently no established definition of the EGS procedures.

Objective  To define a set of procedures accounting for at least 80% of the national burden of operative EGS.

Design, Setting, and Participants  A retrospective review was conducted using data from the 2008-2011 National Inpatient Sample. Adults (age, ≥18 years) with primary EGS diagnoses consistent with the American Association for the Surgery of Trauma definition, admitted urgently or emergently, who underwent an operative procedure within 2 days of admission were included in the analyses. Procedures were ranked to account for national mortality and complication burden. Among ranked procedures, contributions to total EGS frequency, mortality, and hospital costs were assessed. The data query and analysis were performed between November 15, 2015, and February 16, 2016.

Main Outcomes and Measures  Overall procedure frequency, in-hospital mortality, major complications, and inpatient costs calculated per 3-digit International Classification of Diseases, Ninth Revision, Clinical Modification procedure codes.

Results  The study identified 421 476 patient encounters associated with operative EGS, weighted to represent 2.1 million nationally over the 4-year study period. The overall mortality rate was 1.23% (95% CI, 1.18%-1.28%), the complication rate was 15.0% (95% CI, 14.6%-15.3%), and mean cost per admission was $13 241 (95% CI, $12 957-$13 525). After ranking the 35 procedure groups by contribution to EGS mortality and morbidity burden, a final set of 7 operative EGS procedures were identified, which collectively accounted for 80.0% of procedures, 80.3% of deaths, 78.9% of complications, and 80.2% of inpatient costs nationwide. These 7 procedures included partial colectomy, small-bowel resection, cholecystectomy, operative management of peptic ulcer disease, lysis of peritoneal adhesions, appendectomy, and laparotomy.

Conclusions and Relevance  Only 7 procedures account for most admissions, deaths, complications, and inpatient costs attributable to the 512 079 EGS procedures performed in the United States each year. National quality benchmarks and cost reduction efforts should focus on these common, complicated, and costly EGS procedures.

Emergency general surgery (EGS) encompasses the care of the most acutely ill, highest risk, and most costly general surgery patients.^1-5 There are more than 3 million patients admitted to US hospitals each year for EGS diagnoses, more than the sum of all new cancer diagnoses.^1-5 This is a particularly high-risk population of surgery patients—those who undergo an EGS operation are up to 8 times more likely to die postoperatively than are patients undergoing the same procedures electively.² In addition, approximately half of all patients undergoing EGS will develop a postoperative complication,^2,4,6 and up to 15% will be readmitted to the hospital within 30 days of their surgery.³ Despite the inordinate burden of EGS among operative surgical patients, to our knowledge, surgical benchmarks do not exist for EGS procedures.

Earlier efforts to define the scope of EGS have used various data sources: billing data and relative value units⁷ from EGS surgeons, Current Procedural Terminology codes of acute care surgery fellows,⁸ and single-institution assessments.⁹ In 2013, the American Association for the Surgery of Trauma (AAST) Committee on Severity Assessment and Patient Outcomes addressed this issue by publishing a landmark list¹⁰ of 621 International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) diagnosis codes.¹¹ This list was created to include “any patient (inpatient or emergency department) requiring an emergency surgical evaluation (operative or nonoperative) for diseases within the realm of general surgery as defined by the American Board of Surgery.”^10(p1092) A follow-up study proposed 149 ICD-9-CM procedure codes that were identified as procedures that would or could treat any of the aforementioned EGS diagnoses.^1,12-15 These lists represent a significant advance toward a unified definition of EGS; however, the wide clinical heterogeneity limits the ability to translate these studies to existing surgical benchmarking efforts.^1,10,12-15

To build on this broadly inclusive list of diagnoses that encompass operative and nonoperative EGS patients,^1,10,16,17 a more focused, clinically relevant, and nationally representative list of EGS procedures is needed. The goal of the present study was to expand the current diagnosis-based definition of EGS¹⁰to define a standardized, representative set of procedures that make up the bulk of the national clinical burden of EGS. This set will simplify efforts to establish EGS benchmarks, guide standardized and focused research priorities, and inform quality improvement programs. As such, the aim of the present study was to identify a set of urgent and EGS procedures that account for more than 80% of the national burden of EGS in terms of frequency, morbidity, mortality, and cost.

Box Section Ref ID

From November 15, 2015, to February 16, 2016, we queried data from the 2008 to 2011 Hospital Cost and Utilization Project’s (HCUP)¹⁶ National Inpatient Sample (NIS). The data query was performed The NIS is the largest all-payers claims database that is publicly available in the United States. The NIS sampling framework consists of a 20% sample of hospitals stratified based on geographic region, ownership control, urban or rural location, teaching status, and number of hospital beds. This framework provides a 90% sample of all hospital discharges in the United States and includes patient-level and hospital-level variables, comprising up to 25 diagnosis codes and 15 procedure codes per patient encounter.^11,16 The HCUP-provided discharge weights are provided annually to allow for nationally representative population weighting of discharges. The Partners Human Research Committee, the institutional review board of Partners Healthcare, approved this study.

The analytic data set consisted of all patients 18 years or older who had a principal diagnosis consistent with the AAST EGS definition based on ICD-9-CM diagnosis codes^11,18,19 who were admitted urgently or emergently or who were admitted through the emergency department if the admission type was unknown. The sample was further limited to include only patients who underwent an operative procedure on the day of or the day following admission. To focus the analysis on the scope of practice common to emergency general surgeons across a variety of facility types, patients who underwent primarily obstetric, cardiac, vascular, endovascular, or endoscopic procedures were excluded. Finally, patients undergoing very rare procedures were excluded by removing the bottom 5% of the sample when ranked by procedure frequency.

Patient- and facility-level demographic data were derived from NIS-provided variables and included age, sex, race/ethnicity (non-Hispanic white, non-Hispanic black, Hispanic, other, and missing), patient zip code income quartile, insurance payer (private, public, uninsured, and other), US census region (Northeast, Midwest, South, and West), hospital teaching status and location (urban teaching, urban nonteaching, and rural), and hospital bed size (small, medium, and large; actual cutoffs vary according to urban or rural location and US census region).

The NIS-provided ICD-9-CM procedure codes were analyzed in their original 4-digit format (NN.nx) as well as collapsed into 3-digit procedure groups (NN.n). This was done to ensure that similar procedures were clustered together according to the preexisting ICD-9-CM hierarchy of procedure codes and to avoid underestimation or overestimation of procedural burden owing to an increased or decreased number of second-order divisions. For example, a right-sided hemicolectomy (45.73), a left-sided hemicolectomy (45.75), and a sigmoid hemicolectomy (45.76) would all be considered partial colectomies (45.7x). However, these would remain a distinct procedure group from all variants of total colectomies (45.8x).

Procedure and procedure group counts were based on the NIS-provided principal procedure code. Mortality rates were also based on NIS-provided outcome data. Complications were based on secondary ICD-9-CM codes and included 13 complication types: pneumonia, deep vein thrombosis and/or pulmonary embolism, acute renal injury, stroke, myocardial infarction, cardiac arrest, acute respiratory distress syndrome, sepsis, septic shock, mechanical wound failure, wound infection, postoperative gastrointestinal tract complications, and other or unspecified postoperative complications including retained foreign body and postoperative hemorrhage.^17,20 Each patient could theoretically have 1 to 13 complications during an inpatient stay. Complication rate was calculated as a binary variable, indicating the presence of any complication, whereas the total complication count was calculated as the sum of all complications by procedure group. The NIS-provided inpatient hospital charges were converted to estimated costs using HCUP-provided cost to charge ratios, and these estimates from 2008 to 2011 were normalized to 2015 US dollars using annual hospital consumer-price indices.^{1,10,18,19,21}

Procedure count, total deaths, total complications, and total index hospitalization costs were calculated for each 3-digit ICD-9-CM procedure group included in the final analytic data set. A mortality score and a complications score were defined as the product of procedure frequency and mortality rate or complications rate, respectively. Scores were calculated per procedure groups and then ranked, such that each 3-digit ICD-9-CM procedure group received 1 mortality rank and 1 complication rank. These 2 ranks were then summed to generate a final burden rank by procedure group.

Once sorted by burden rank, the cumulative attributable burden was determined regarding the frequency, total deaths, total complications, and total hospital costs. For procedure groups ranked 1 to 35, the cumulative attributable burden of procedure frequency, for example, was calculated as the sum of the number of procedures ranked 1 to n divided by the total number of procedures in the entire analytic sample. This measure is akin to a percentile value for each ranked procedure. The cumulative attributable burden for frequency, deaths, complications, and costs was then calculated. For the procedure group with rank x, the cumulative attributable burden for an outcome of interest represents the summation of the proportion of total burden for that outcome attributable to procedures ranked 1 through x. A threshold of 80% of the total population burden was established as an a priori cutoff to represent significant representation of population-level burden for each of the 4 outcomes of interest. Costs were not used to calculate burden scores, but they were reported as an outcome measure to determine whether highly ranked procedures also account for significant surgical health care spending. All counts and means were weighted using NIS-provided population design-weights generalized with Stata’s svy command to account for patient clustering within hospitals and to attain nationally weighted estimates. All analyses were completed using Stata, version 13.0 (StataCorp, Inc). Data analysis was conducted from November 15, 2015, to February 16, 2016.

This study identified a sample of more than 3.1 million patients with an EGS diagnosis (Figure 1) weighted to represent more than 35.1 million patient encounters nationally in 4 years. From this sample, 512 079 patients (16.5%) were excluded because they were not admitted urgently or emergently. An additional 1 813 888 patients (58.4%) were excluded because they did not undergo a major operation during their hospital stay. Another 319 675 individuals (10.3%) were excluded because they did not undergo an operative procedure on the first or second day of admission. After excluding a small proportion of cardiac, vascular, obstetric, endoscopic, nonoperative procedure codes (13 532 [0.4%]), as well as rare procedure codes (23 532 [0.8%]), the final analytic sample included 421 476 patient encounters (13.6% of the original sample). This representative cohort exhibits appropriate patient-level, facility-level, and geographic diversity (eTable 1 in the Supplement). This sample included adults of all ages (range, 18-105 years), was racially/ethnically diverse (61.3% non-Hispanic white, 9.0% non-Hispanic black, and 14.5% Hispanic individuals), comprised a diverse payer mix (46.1% private, 37.8% public, and 11.6% uninsured), and had even distribution across all income quartiles (range, 24.6% in quartile 1 to 25.4% in quartile 3). This sample also represented a diverse array of facilities in which 61.2% were large hospitals and 11.8% were urban teaching hospitals. Applying NIS-provided population weights, the final analytic sample represented a total of 2.1 million operative patient encounters nationwide within 4 years, ranging from 502 484 (95% CI, 453 226-551 742) in 2008 to 537 031 (95% CI, 484 819-589 243) in 2009. The mean mortality in the overall sample was 1.23% (95% CI, 1.18%-1.28%), and 15.0% (95% CI, 14.6%-15.3%) of the patients experienced 1 or more complication. The mean estimated inpatient hospital cost per admission was $13 241 (95% CI, $12 957-$13 525). Nationally, the annual estimated cost ranged from $6.1 billion (95% CI, $5.5 billion to $6.7 billion) in 2008 to $6.6 billion (95% CI, $6.0 billion to $7.3 billion) in 2010.

The final analytic sample included 137 unique 4-digit ICD-9-CM procedure codes, which mapped into 35 distinct 3-digit procedure group codes. When ordered by burden rank 1 to 35, the cumulative attributable burden for total procedure count, total deaths, total complications, and total costs increased sharply through procedures ranked 1 to 7, but all 4 curves notably flattened thereafter (Figure 2). Each additional procedure increased the cumulative burden (for any of the 4 outcomes) by a mean of 9% to 12% for the first 7 procedures, whereas the next 7 procedures increased the cumulative attributable burden by a mean of 2% with each additional procedure. These first 7 procedure groups accounted for 80.0% of all procedures, 80.3% of all deaths, 78.9% of all complications, and 80.2% of all inpatient costs in this nationally representative data set (Figure 2). The top 7 procedures included partial colectomy (ICD-9-CM procedure group 45.7x), small-bowel resection (group 45.6x), cholecystectomy (group 51.2x), operative management of peptic ulcer disease (group 44.4x), lysis of peritoneal adhesions (group 54.5x), appendectomy (group 47.0x), and laparotomy (group 54.1x) (Table 1). The top 10 primary diagnosis codes associated with each of these top-ranked procedures as well as the most common diagnoses associated with procedure group 54.1x (laparotomy) are presented in eTable 2 and eTable 3, respectively, in the Supplement.

A complete list of procedure frequency, mortality rate, complication rate, burden ranks, and mean inpatient costs of all 35 procedure groups is reported in Table 2. Among the top 7 ranked procedures, the frequency varied from 682 043 primary appendectomies to 9418 primary laparotomies, the mortality rate ranged from 0.08% for appendectomy to 23.76% for laparotomy, and the complication rate ranged from 7.27% for appendectomy to 46.94% for small-bowel resection. Finally, mean inpatient costs ranged from $9664.30 for appendectomy to $28 450.72 for small-bowel resection. Figure 3 shows the associations between frequency and mortality rate as well as between frequency and complication rate for these 7 procedure groups.

To our knowledge, this study is the first to use a nationally representative sample to identify the EGS procedures that account for the greatest number of cases, deaths, complications, and inpatient costs. Starting with a previously defined list of 621 diagnoses, this study has identified 7 procedure groups that account for approximately 80% of the burden of operative EGS throughout the United States. This focused and clinically relevant list of procedures is an appropriate starting point for efforts of surgeons and researchers to improve quality, reduce costs, and develop national benchmarks for EGS.

To our knowledge, this list is the first to be derived from a nationally representative data set that includes patients of all ages, races/ethnicities, income levels, facility types, and geographic regions. In addition, ranking by the cross-products of both frequency × mortality rate and frequency × complication rate enabled identification of procedures that are common, deadly, and represent significant morbidity (Figure 2). By excluding procedures that are not common to the scope of practice of all general surgeons in the United States (ie, cardiac, thoracic, vascular, obstetric, and endoscopic), the resultant list represents sufficient homogeneity that can be interpreted in a clinically meaningful way throughout the country. Although cost was not used to rank the procedures, it is notable that the same 7 procedures ranked by clinical burden also accounted for 80% of all EGS-related inpatient costs (Figure 2). This finding further emphasizes the usefulness of these 7 procedures to serve as the basis for understanding ways to improve quality and reduce cost among patients undergoing EGS.

These findings build on important prior work seeking to define EGS.^1,10,21 The diagnosis codes put forth by the AAST’s 2013 committee¹⁰ are broadly inclusive of all operative and nonoperative diagnoses that may require urgent care provided by a surgeon. However, these analyses demonstrate that less than 14% of all inpatients with one of these diagnoses required a major operation within 2 days of an emergent or urgent admission. Ogola and Shafi²¹ subsequently identified 9 diagnoses that constituted 80% of all admissions; however, these diagnoses were not limited to operative admissions. Although a significant proportion of surgical care is nonoperative, many patients throughout the United States who are admitted for diagnoses such as diverticulitis, bowel obstruction, or soft-tissue infections are never evaluated or treated by a surgeon. As such, basing quality benchmarks on such an inclusive list may not provide specific information on the quality of surgical care.

In 2014, Gale and colleagues¹ proposed a list of 149 procedures that could be used to categorize patients with any of the AAST-defined EGS diagnoses. However, this list does not take into account which of these procedures represent the most significant proportion of national burden. In addition, the list’s broad clinical heterogeneity may make derivation of surgical national quality targets and national benchmarks difficult. Thus, although existing work has done much to establish a novel, sensitive, and inclusive definition of EGS, a focused sample that represents the true national burden is needed to focus benchmarking, quality improvement, and cost reduction efforts.

Existing quality efforts in health care focus on the conditions that represent common reasons for admissions nationwide and account for significant morbidity, mortality, and costs.¹² For example, the 2012 Centers for Medicare & Medicaid Services’ Readmission Reduction Program was applied only to patients admitted for acute myocardial infarction, heart failure, and pneumonia.¹² When this program was expanded to surgical patients, it likewise was limited to 3 procedures: total hip arthroplasty, total knee arthroplasty, and coronary artery bypass graft.^12,13 Focusing on a subset of important procedures is in keeping with the Centers for Medicare & Medicaid–sponsored Surgical Care Improvement Project,¹³ the American College of Surgeons’ National Surgical Quality Improvement Program,^5,14 and the Society of Thoracic Surgeons’¹⁵ quality benchmarks, which are all focused on patients who have undergone one of a specific, clearly defined set of procedures. Such measures are thought to be valuable because they are indicative of care provided by surgeons and because they can be applied to a variety of hospitals throughout the United States. Given their high prevalence nationally and the high proportion of burden that they represent, deriving EGS benchmarks from the 7 procedures identified in this study could lead to better clinical decision making, patient outcomes, and cost savings.

The findings presented should be interpreted in light of the study’s limitations. First, these results were derived from claims data rather than from prospective, clinically derived databases. However, the national representativeness of the NIS facilitates the identification of the highest burden procedures throughout the United States. Future studies are warranted to subsequently monitor these procedures in prospective clinical-derived databases. Second, the exclusion of nonoperative patients from the final analytic data set omits an important cohort that may not require an operative intervention but may otherwise require the care of a surgeon. As discussed above, the list provided in this study is meant to guide benchmarks for operative EGS akin to those that exist for nonemergent surgery. Analysis of nonoperative care is warranted and important, although many databases preclude the ability to determine which patients receive care that is managed primarily by surgeons and which receive care that is managed primarily by nonsurgeons, which may result in differences in both quality and costs of care.²² Third, this analysis was intentionally limited to operative procedures performed within 2 days of admission for an EGS-related diagnosis. As such, it does not account for the burden of trauma-related procedures or semi-elective procedures that take place many days after admission. Finally, these findings do not represent other forms of burden, such as postoperative quality of life, return to work, or other patient-centered outcomes. This limitation is common in studies using claims databases and warrants further prospective inquiry to identify the outcomes that matter most to patients undergoing EGS procedures.

Analysis of the largest available nationally representative database demonstrates that more than half of a million patients undergo urgent or emergent general surgery operations annually in the United States that account for more than $6 billion in annual costs. Only 7 representative procedures account for approximately 80% of all admissions, deaths, complications, and inpatient costs attributable to operative EGS nationwide. National quality benchmarks and cost reduction efforts should focus on these common, complicated, and costly EGS procedures.

Correction: This article was corrected on June 15, 2016, to fix an incorrect burden rank description in Table 2.

Corresponding Author: Joaquim M. Havens, MD, Center for Surgery and Public Health, Department of Surgery, Brigham & Women’s Hospital, 75 Francis St, Boston, MA 02115 (jhavens@partners.org).

Accepted for Publication: February 18, 2016.

Published Online: April 27, 2016. doi:10.1001/jamasurg.2016.0480.

Author Contributions: Drs Scott and Havens had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Scott, Olufajo, Rose, Zogg, Haider, Salim, Havens.

Acquisition, analysis, or interpretation of data: Scott, Olufajo, Brat, Rose, Zogg, Haider, Havens.

Drafting of the manuscript: Scott, Olufajo, Brat, Havens.

Critical revision of the manuscript for important intellectual content: Scott, Olufajo, Brat, Rose, Zogg, Haider, Salim.

Statistical analysis: Scott, Olufajo, Zogg.

Administrative, technical, or material support: Scott, Rose, Zogg, Haider.

Study supervision: Brat, Rose, Haider, Salim, Havens.

Conflict of Interest Disclosures: Dr Haider reported being the principal investigator of contract AD-1306-03980 with the Patient-Centered Outcomes Research Institute and a Harvard Surgery Affinity Research Collaborative Program grant 112227. Dr Haider also reported being the co-founder and an equity holder in Patient Doctor Technologies Inc, which owns and operates the website https://www.doctella.com. No other disclosures were reported.