eAppendix. Most Common Procedures Performed in Emergency General Surgery Patients (Overall and by Diagnosis Group).
eFigure. Schematic Diagram of the Selection Process for Emergency General Surgery Patients.
eTable. Major Reasons for Readmissions Among Emergency General Surgery Patients, California State Inpatient Database, 2007-2011.
Customize your JAMA Network experience by selecting one or more topics from the list below.
Havens JM, Olufajo OA, Cooper ZR, Haider AH, Shah AA, Salim A. Defining Rates and Risk Factors for Readmissions Following Emergency General Surgery. JAMA Surg. 2016;151(4):330–336. doi:10.1001/jamasurg.2015.4056
Copyright 2015 American Medical Association. All Rights Reserved.
Hospital readmission rates following surgery are increasingly being used as a marker of quality of care and are used in pay-for-performance metrics. To our knowledge, comprehensive data on readmissions to the initial hospital or a different hospital after emergency general surgery (EGS) procedures do not exist.
To define readmission rates and identify risk factors for readmission after common EGS procedures.
Design, Setting, and Participants
Patients undergoing EGS, as defined by the American Association for the Surgery of Trauma, were identified in the California State Inpatient Database (2007-2011) on January 15, 2015. Patients were 18 years and older. We identified the 5 most commonly performed EGS procedures in each of 11 EGS diagnosis groups. Patient demographics (sex, age, race/ethnicity, and insurance type) as well as Charlson Comorbidity Index score, length of stay, complications, and discharge disposition were collected. Factors associated with readmission were determined using multivariate logistic regression models analysis.
Main Outcomes and Measures
Thirty-day hospital readmission.
Among 177 511 patients meeting inclusion criteria, 57.1% were white, 48.8% were privately insured, and most were 45 years and older (51.3%). Laparoscopic appendectomy (35.2%) and laparoscopic cholecystectomy (19.3%) were the most common procedures. The overall 30-day readmission rate was 5.91%. Readmission rates ranged from 4.1% (upper gastrointestinal) to 16.8% (cardiothoracic). Of readmitted patients, 16.8% were readmitted at a different hospital. Predictors of readmission included Charlson Comorbidity Index score of 2 or greater (adjusted odds ratio: 2.26 [95% CI, 2.14-2.39]), leaving against medical advice (adjusted odds ratio: 2.24 [95% CI, 1.89-2.66]), and public insurance (adusted odds ratio: 1.55 [95% CI, 1.47-1.64]). The most common reasons for readmission were surgical site infections (16.9%), gastrointestinal complications (11.3%), and pulmonary complications (3.6%).
Conclusions and Relevance
Readmission after EGS procedures is common and varies widely depending on patient factors and diagnosis categories. One in 5 readmitted patients will go to a different hospital, causing fragmentation of care and potentially obscuring the utility of readmission as a quality metric. Assisting socially vulnerable patients and reducing postoperative complications, including infections, are targets to reduce readmissions.
Hospital readmission rates following surgery are increasingly used as a marker of quality of care and are used in pay-for-performance metrics. As such, reducing hospital readmission rates has become a focus of both physicians and hospital administrators as well as policy makers.1-4
Emergency general surgery (EGS) patients represent a unique population at high risk for medical errors and complications following surgery.5-8 Approximately half of all patients undergoing EGS will have a postoperative complication,7-9 and postoperative complications have been closely linked to hospital readmission.10-12 This disproportionate burden of complications is of critical importance because it would seem to place the EGS patient at increased risk for readmission. Furthermore, patients’ race/ethnicity and proxies for socioeconomic status, such as insurance payor status, differ in EGS patients when compared with patients undergoing elective procedures.13 Patients undergoing EGS are more likely to be minorities, uninsured, or on government insurance, and these demographic groups are known to have disproportionally higher readmission rates.14,15 Put together, these factors highlight the need to closely examine readmission rates in EGS patients and identify the patterns of readmissions in this population.
Studies on readmission rates in surgical patients are limited by the restricted number of procedures they examine, the exclusive age categories that are included, and the lack of distinction of EGS patients from other surgical patients.16-20 It has been shown that outcomes in surgical patients, including readmission rates, vary significantly by patient age, procedure type, and surgical specialty.10,16,20-22 Therefore, to define the rates, identify risk factors, and describe the patterns of hospital readmission among a representative group of EGS patients, while accounting for variations in patients’ age groups and diagnoses, we analyzed hospital discharge records from a statewide database. We hypothesized that surgical complications would be responsible for the majority of readmissions and that demographic factors and admission characteristics would contribute to rates of readmission in the EGS population.
Hospital discharge records in the California State Inpatient Database, Healthcare Cost and Utilization Project, Agency for Healthcare Research and Quality were extracted from 2007 to 2011. This database contains records from all participating hospitals in the state. The California State Inpatient Database was composed of 13 037 628 discharges during the study and encompasses more than 98% of the hospital discharges in the state.23 Patients had unique identifiers assigned to them that made it possible to track their admissions during the study as long as they were readmitted within the state.
We included patients with an EGS diagnosis as defined by the American Association for the Surgery of Trauma using a previously defined list of 467 International Classification of Diseases, Ninth Revision codes.24 To facilitate analysis and subsequent benchmarking, we categorized the patients into 11 diagnosis groups as described by the American Association for the Surgery of Trauma, namely: cardiothoracic, colorectal, general abdominal conditions, hepatic-pancreatic-biliary, hernias, intestinal obstruction, resuscitation, soft tissue, upper gastrointestinal tract, vascular, and others. The diagnoses codes associated with each diagnosis group have been previously published.24
The database contains information on whether admissions were scheduled or not scheduled. It also has information on the source of admission: eg, emergency department, other hospital, and other health care facility. We limited our selection to patients who had unplanned admissions through the emergency department and who underwent a surgical procedure on the day of admission. We excluded patients who had procedures that were not likely performed by general surgeons, such as interventional radiology procedures, cardiac procedures, or complex vascular procedures. From this cohort, we analyzed those patients who had 1 of the 5 most common procedures performed in each diagnosis group.
To ensure that we were capturing true emergencies and not cases that were scheduled during patients’ hospital stays, we excluded patients who underwent surgical procedures on any day other than the day of admission. We also excluded patients who were admitted or discharged in the last month of data collection (to ensure at least 30 days of postdischarge follow-up for each patient) and patients who died during admission or were transferred to another hospital. This study examined deidentified data and was deemed exempt from full review by the institutional review board of Partners Healthcare.
We included demographic information including age, sex, race/ethnicity (white, black, Hispanic, Asian/Pacific Islander, Native American, and others), and insurance type (private, Medicare/Medi-Cal, self-pay, and others). Medi-Cal is the State of California’s equivalent of Medicaid, and insurance type classified as others included worker’s compensation, county indigent programs, other government aids, and other indigent funds.25
Clinical variables measured included the Charlson Comorbidity Index, which was calculated by the CHARLSON program for Stata version 13 (StataCorp).26 The Charlson Comorbidity Index score includes 19 diseases weighted on the basis of their association with mortality, ranging from 0 to 37, with a higher score indicating greater comorbidity.27 Other variables included were length of hospital stay, complications during hospital stay, and discharge disposition (home, home health care, skilled nursing facility/intermediate care facility/other facilities, and against medical advice).
The 30-day readmission rate was calculated as the proportion of patients who had unplanned readmissions within 30 days after discharge. We stratified the patients by diagnosis groups and calculated 30-day readmission rates for each group. Only 9 of the 11 groups were included in further analyses because some groups either contained too few patients (resuscitation) or were too ambiguous (others).
We performed univariate logistic regression analyses to identify factors associated with 30-day readmission. Factors that were not significantly associated with readmission in the univariate analysis (P < .10), except sex (in which association with readmissions has been previously shown22), were excluded from the multivariate analyses. Multivariate logistic regression models were fitted to identify independent associations with readmissions. Reference categories used in making comparisons included age (18 to 44 years), sex (male), race/ethnicity (white), insurance type (private), Charlson Comorbidity Index score (<2), length of stay (≤3 days), complications during admission (absent), and discharge disposition (home).
To measure the degree of variation in readmission rates among diagnosis groups, we predicted the readmission rates for each diagnosis group using the multivariate models that adjusted for factors we observed were associated with readmission. Because of the potential differences in outcomes seen in elderly patients when compared with younger ones, we repeated these analyses on an age-stratified sample of patients.
Using the admitting International Classification of Diseases, Ninth Revision codes on readmission, we identified the major reasons for readmission and categorized them into previously described convenient clinically meaningful categories, which included gastrointestinal, surgical infections, malnutrition, wound complications, genitourinary, vascular, pulmonary, cardiac, pain, neurologic, and others.10 We also calculated the proportion of patients who were readmitted at hospitals different from the hospitals of index admission and identified factors associated with readmission at a different hospital using univariate and multivariate logistic regression models. All analyses were done using Stata Statistical Software release 13 (StataCorp LP), and the significance level was set as P < .05.
Among the procedures performed, we selected the 5 most common procedures in each diagnosis group, resulting in 44 unique procedures that were used for further analyses. As the diagnosis groups were categorized by International Classification of Diseases, Ninth Revision diagnosis codes and not procedure codes, several procedures were found in more than 1 diagnosis group. Overall, the 5 most common procedures were laparoscopic cholecystectomy, laparoscopic appendectomy, other incision with drainage of skin and subcutaneous tissue, other appendectomy, and other partial resection of the small intestine. Combined, these 5 procedures accounted for more than 50% of all the operative procedures that were performed on patients in this cohort (data not shown). The full list of the top 5 procedures per diagnosis group and the 44 most common procedures in EGS patients are included in the eAppendix in the Supplement.
Our final cohort included 177 511 patients (eFigure in the Supplement). Of these, 57.07% were white, 48.76% were privately insured, and most were 45 years and older (51.26%) (Table 1). The Charlson Comorbidity Index score was less than 2 in 90.37% of patients.
The mean length of stay was 3.2 days, and 72.63% of the patients were hospitalized for less than 4 days. Complications during admission were seen in 16.46% of patients, and 89.93% were discharged directly to their homes.
The upper gastrointestinal tract diagnosis group accounted for 51.23% of the primary diagnoses seen in this cohort. Other groups with large numbers of patients were the hepato-pancreatic-biliary group (20.95%) and the soft tissue group (12.19%). There were less than or equal to 10 patients in our cohort who were categorized under the resuscitation group, and they were not included in further analyses.
The overall 30-day readmission rate for the cohort was 5.91%. When we examined readmission rates in the different diagnosis groups, the rates ranged from 4.11% (upper gastrointestinal tract ) to 16.75% (cardiothoracic). Readmission rates for other diagnosis groups were 5.42% hepato-pancreatic-biliary, 7.19% soft tissue, 10.59% intestinal obstruction, 10.90% colorectal, 11.05% hernia, 14.45% general abdominal, and 15.24% vascular (Figure).
There were marked differences in readmission rates based on patient demographics and admission characteristics. Higher readmission rates were seen in patients 65 years and older (10.59%), black patients (11.01%), and patients with high Charlson Comorbidity Index scores (17.11%) compared with other patients (Table 2). Patients with lengths of stay less than 4 days (4.12%) and those who were discharged home (5.01%) had relatively lower readmission rates compared with other groups of patients.
On multivariate analyses, independent risk factors for readmission were having a Charlson Comorbidity Index score of 2 or greater (adjusted odds ratio [aOR]: 2.26 [95% CI, 2.14- 2.39]), being discharged against medical advice (aOR: 2.24 [95% CI, 1.89-2.66]), hospital stay greater than 7 days (aOR: 2.10 [95% CI, 1.96- 2.26]), and public insurance (aOR: 1.55 [95% CI, 1.47- 1.64]). Being black (aOR: 1.38 [95% CI, 1.15-1.65]) and being discharged to any type of posthospitalization care facility (aOR: 1.80 [95% CI, 1.50-2.16]) were also associated with higher readmission rates, but sex did not play a role in readmissions (aOR: 1.10 [95% CI, 0.98-1.23]).
Readmission rates in patients in each of the diagnosis groups, after adjusting for factors that were associated with readmission, are shown in Table 3. The rates were generally lower in patients younger than 65 years than in patients 65 years and older. The rates ranged from 5.04% (upper gastrointestinal tract ) to 11.87% (vascular) among the patients younger than 65 years and 8.70% (hepato-pancreatic-biliary) to 15.50% (vascular) in the patients 65 years and older.
Among patients younger than 65 years, the major reason for readmission was surgical infections (20.47%) followed by gastrointestinal illnesses (11.67%). For patients 65 years and older, 10.40% were readmitted owing to gastrointestinal illnesses, and 9.27% of readmissions were owing to surgical infections. Older patients were more often readmitted due to malnutrition, genitourinary, vascular, pulmonary, and cardiac reasons than patients younger than 65 years (P < .001). Details of the reasons for readmission are included in the eTable in the Supplement.
The mean length of stay of readmissions was 5.4 days, and 16.87% of patients had another surgery on readmission. Of note, 16.80% of the patients were readmitted at a hospital different from the hospital of the index admission. The main factors associated with readmission at a different hospital were being discharged against medical advice (aOR: 2.26 [95% CI, 1.61- 3.16]), being discharged to a skilled nursing facility (aOR: 1.80 [95% CI, 1.50 -2.16]), and paying out of pocket (aOR: 1.64 [95% CI, 1.31- 2.05]) (Table 4).
Because of the difficulties faced in characterizing EGS patients, the readmission rates of this group of patients have not previously been well described. The American Association for the Surgery of Trauma’s initiative toward recognizing emergency general surgery as a potential surgical subspecialty was an important step toward benchmarking surgical care for these patients.24 Our study found that 5.91% of EGS patients were readmitted within 30 days of discharge and that there was a wide variation in the readmission rates observed, depending on the admitting diagnoses of the patients. Variations in readmission rates based on the admitting diagnosis have previously been shown in surgical patients treated nonemergently.22,28 To our knowledge, this is the first study that provides estimates of readmission rates in a well-characterized and representative group of EGS patients. It is also the first study to describe readmission rates and characteristics of distinct diagnosis groups of EGS patients and indicates that hospital comparisons based on readmission rates should consider the type and volume of EGS to appropriately risk adjust.
Previous studies suggest that the characteristics of the patients seen at hospitals influence the readmission rates of such hospitals.16,29,30 We found that EGS patients who were on public insurance, had higher comorbidity status, had longer lengths of hospital stay, and had discharge dispositions other than home were more likely to be readmitted within 30 days. This pattern has also been seen with readmissions in elective surgical patients.20 Despite the fact that some of these factors are not modifiable, they serve as markers for high-risk patients in whom more aggressive approaches to management are needed, with an aim of reducing unwanted outcomes such as readmissions. Improving follow-up care, increasing the acceptability of outpatient treatment for appropriate conditions, and ensuring patient stability before discharge are potent strategies to reduce readmissions in high-risk patients.28
In addition to identifying high-risk patients, understanding the reasons for readmission among EGS patients provides an avenue for proactive intervention. Much work has gone into delineating the causes of readmission among patients.31-35 Surgical readmissions are multifactorial but often occur as a result of postoperative complications and infections.10 We found that the major reasons for readmission were different for those patients younger than 65 years compared with those 65 years and older. Among patients younger than 65 years, the most common reason for readmissions was surgical infections, accounting for 20.47% of their readmissions, while this only accounted for 9.27% of readmissions among patients 65 years and older. Ten percent of readmissions among both age groups were due to gastrointestinal complications. Previous studies have shown that the major driver for readmission differs between medical and surgical patients because surgical patients are more often readmitted for surgical complications and medical patients are more often readmitted for an underlying medical condition.10,36 It has been asserted that a major differentiating factor for surgical patients is that patients undergoing planned surgical procedures have an opportunity to undergo medical optimization, reducing the risk of readmission for medical causes.10 Unlike patients who undergo elective procedures, EGS patients do not have the opportunity to be properly optimized before surgery, which places them at an additive risk for readmission both from the surgical procedure and from their underlying conditions.
Some of the gains demonstrated from the National Surgical Quality Improvement Program include reduced complication rates in participating institutions.37,38 Reducing complication rates should result in a consequent reduction in readmission rates. However, there should be corresponding efforts to incorporate care for patients’ underlying conditions and prevention of potential medical complications to reduce overall readmission rates among EGS patients.
The debate about whether measuring readmission rates is an appropriate metric of surgical quality has been a heated one with divergent views.11,39 One of the major arguments against this measure is that many patients develop complications after discharge, which affects the ability of readmission rates to accurately measure the quality of care received during the admission. Furthermore, same-hospital readmissions, which is the method commonly used to measure readmissions, does not adequately capture hospital readmissions, as many patients get readmitted to different hospitals.40 We found that 16.80% of readmitted patients were readmitted at hospitals other than where their procedures were performed. These patients were more likely to have left the index hospital against medical advice, be on public insurance, or be uninsured. The evidence from a nationwide cohort of patients points to the fact that the outcomes of surgical patients readmitted at different hospitals are significantly worse than those readmitted to the same hospital.41 This indicates that hospitals that have a high retention rate for their readmissions (low level of fragmentation of care) may ultimately have better patient outcomes but may have worse ratings when using readmissions as a quality measure.
There were a few limitations to this study that should be considered. First, we used an administrative database that was able to capture greater than 98% of all hospital readmission in the state but did not have clinical data. This absence of clinical data limited our ability to identify other possible risk factors that may be associated with readmission based on previous studies.42 We also were not able to capture the hospital characteristics of the index or readmitting hospitals, which have been shown to be associated with readmissions.22 While these factors may play significant roles in patterns of readmission, our objective was to outline patient risk factors for readmission. Further investigation of readmission that incorporates hospital data would be useful for the creation of adequate strategies to reduce hospital readmissions.
With this study, we have been able to characterize the demographics of EGS patients and identify a representative group of surgical procedures that EGS patients commonly undergo. We have also identified patient-level risk factors for readmission to the index hospital and to other hospitals. Reducing readmissions is a noble cost-saving goal with benefits not only to the hospitals, but also to the patients. However, it is critical to understand the underlying factors associated with readmission to appropriately identify quality-improvement measures that address the true problem. Focused and concerted efforts should be made to incorporate readmission-reducing strategies into the care of EGS patients, particularly among those at higher risk for readmission.
Corresponding Author: Joaquim M. Havens MD, Division of Trauma, Burn and Surgical Critical Care, Brigham and Women's Hospital, Boston, MA 02115 (firstname.lastname@example.org).
Accepted for Publication: July 20th, 2015.
Published Online: November 11, 2015. doi:10.1001/jamasurg.2015.4056
Correction: This article was corrected on March 22, 2017, to correct the number of patients listed in Table 1 and revise the corresponding sentence in the Results section.
Author Contributions: Dr Havens 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: All authors.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Havens, Olufajo, Cooper, Shah.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Olufajo, Shah, Salim.
Administrative, technical, or material support: Havens, Cooper, Salim.
Study supervision: Havens, Cooper, Salim.
Conflict of Interest Disclosures: Dr Haider reports that he is cofounder and equity shareholder of Doctella. His involvement in the company is not related to the contents of this study. No other disclosures are reported.
Create a personal account or sign in to: