Reconceptualizing High-Quality Emergency General Surgery Care: Non-Mortality-Based Quality-Metrics Enable Meaningful and Consistent Assessment

Abstract

Background:

Ongoing efforts to promote quality-improvement in emergency general surgery (EGS) have made substantial strides but lack clear definitions of what constitutes “high-quality” EGS care. To address this concern, we developed a novel set of five non-mortality-based quality-metrics broadly applicable to the care of all EGS patients and sought to discern whether: (1) they can be used to identify groups of best-performing EGS hospitals, (2) results are similar for simple versus complex EGS severity in both adult (18–64 years) and older adult (≥65 years) populations, and (3) best performance is associated with differences in hospital-level factors.

Methods:

Patients hospitalized with 1-of-16 AAST-defined EGS conditions were identified in the 2019 Nationwide Readmissions Database. They were stratified by age/severity into four cohorts: simple adults, complex adults, simple older adults, complex older adults. Within each cohort, risk-adjusted hierarchical models were used to calculate condition-specific risk-standardized quality-metrics. K-means cluster analysis identified hospitals with similar performance, and multinomial regression identified predictors of resultant “best/average/worst” EGS care.

Results:

1,130,496 admissions from 984 hospitals were included (40.6% simple adults, 13.5% complex adults, 39.5% simple older adults, 6.4% complex older adults). Within each cohort, K-means cluster analysis identified three groups (“best/average/worst”). Cluster assignment was highly conserved with 95.3% of hospitals assigned to the same cluster in each cohort. It was associated with consistently best/average/worst performance across differences in outcomes (5x) and EGS conditions (16x). When examined for associations with hospital-level factors, best-performing hospitals were those with the largest EGS volume, greatest extent of patient frailty, and most complicated underlying patient case-mix.

Conclusion:

Use of non-mortality-based quality-metrics appears to offer a needed, promising means of evaluating high-quality EGS care. The results underscore the importance of accounting for outcomes applicable to all EGS patients when designing quality-improvement initiatives and suggest that, given the consistency of best-performing hospitals, natural EGS centers-of-excellence could exist.

Level of evidence:

Prognostic and Epidemiological, Level III.

Introduction

Accounting for upwards of 2.5 million hospitalizations¹ and 500,000 procedures² per year, emergency general surgery (EGS) has evolved from a subset of cases historically managed by general surgeons to a thriving specialty caring for some of the most high-risk and critically-ill patients.³ It was formally recognized as a part of “Acute Care Surgery” along with trauma and surgical critical care by the American College of Surgeons (ACS), American Association for the Surgery of Trauma (AAST), Western Trauma Association, and Eastern Association for the Surgery of Trauma at a joint meeting in 2003.^4,5 Since that time, efforts to understand EGS have increased awareness of its burden^1,2,6–10 (“higher than [that of] diabetes, coronary artery disease, cancer, and congestive heart failure”)¹ and drawn attention to the need to develop quality-improvement (QI) initiatives capable of optimizing outcomes for EGS patients.³

Following models established in the formalization of trauma centers and creation of organized trauma-systems across the United States (US), EGS-QI efforts have started to see: pilot-testing of a national EGS registry (inclusive of non-operative patients),¹¹ development of a AAST/ACS-sponsored EGS Verification Program,^3,12 and discussion about the potential role for establishing regional centers specializing in the care of EGS patients.^13,14 Yet, for all of these advances, EGS-QI remains “in [its] infancy” when compared to trauma and other surgical fields.³

One of the biggest challenges limiting more rapid dissemination of EGS-QI initiatives is a lack of clear definitions of what constitutes “high-quality” EGS care. While efforts to promote standardization in research and QI have helped (2013: introduction of a list of 621 operative/non-operative diagnosis codes,¹⁵ 2016: development of a standardized anatomic severity grading scale for 16 common EGS conditions,¹⁶ 2021: adaptation of the anatomic severity grading scale for use with national billing claims¹⁷), the unavoidable heterogeneity of EGS conditions, extensive variability in disease severity, and frequent low in-hospital mortality risk for common conditions like acute cholecystitis/appendicitis that make up a sizable portion of EGS’s burden make the best path forward unclear. It is a reality which has led to open calls for the needed development of quality-metrics capable of reflecting the outcomes of all patients presenting for EGS care.^3,18

To help address this concern, we developed a novel set of five non-mortality-based quality-metrics broadly applicable to the care of all EGS patients and sought to discern whether: (1) they can be used to identify groups of best-performing EGS hospitals, (2) results are similar among “simple” versus “complex” EGS severity¹⁷ in both adult (18–64 years) and older adult (≥65 years) populations, and (3) resultant best performance is associated with differences in hospital-level factors. We hypothesized that there are a set of quality-metrics that could be developed to address the above criteria. Considered quality-metrics included: major morbidity during index admission, index hospital length of stay (LOS), patients’ ability to be discharged home, need for readmission within 30 days, and patients’ total number of hospitalized days within 6 months (proxy measure for patients’ average number of healthy days at home [HDAH]).

Methods

Data source and study population

Records of index EGS admissions for adult (18–64 years) and older adult (≥65 years) patients were abstracted using January-June 2019 hospital inpatient claims contained with the US Agency for Healthcare Research and Quality’s (AHRQ) Nationwide Readmissions Database (NRD). Data through six months after (January-December 2019) were used to determine subsequent readmission (within 30 days) and hospitalized time (within 6 months).

NRD is the largest publicly-available, all-payer inpatient database in the US inclusive of longitudinal information.¹⁹ In 2019, it included data from 30 states, representing 61.8% of the US population and 60.4% of all hospital admissions. NRD includes information on patient encounters for US patients of all ages with all forms of health insurance/primary payer. It contains data on up to 25 ICD-10-PCS procedure and 40 ICD-10-CM diagnosis codes. All data included in NRD are collected from US hospitals by AHRQ in accordance with predetermined data standards.¹⁹

To be included, patients needed to present with a primary diagnosis code consistent with an EGS condition. Following the definition of EGS developed by AAST,¹⁵ we included patients with primary International Classification of Diseases, 10^th edition, Clinical Modification (ICD-10-CM) diagnosis codes that encompass 1 of the 16 common EGS conditions with a corresponding standardized anatomic severity grading scale.¹⁶ Using the definition of simple (“less complex”) versus complex (“more complex”) EGS severity developed for use in national billing claims based on the work of the AAST anatomic severity grading scale,¹⁷ included patients were categorized by EGS condition and divided into one of four cohorts stratified by age/severity: simple adults, complex adults, simple older adults, and complex older adults. Index admissions were defined as those without a prior EGS admission within the preceding 30 days. Included patients were limited to their first recorded admission during the six-month study period.

Risk-adjustment for potential confounders

Potential confounders utilized in the risk-adjustment of risk-standardized quality-metrics included: age in years on index admission, gender (categorized as male versus female), extent of multimorbidity (number of pre-existing conditions based on the Elixhauser Comorbidity Index), Hospital Frailty Risk Score²⁰ (categorized as low risk ≤4, intermediate risk 5–15, and high risk ≥16), and individual binary comorbidity indicators contained within the Elixhauser Comorbidity Index reported on index admission or up to 6 months prior. Hospital Frailty Risk Score and multimorbidity/comorbidities were calculated using ICD-10-CM primary/secondary diagnosis codes. Elixhauser Comorbidity Index scores were determined using the “elixhauser” program in Stata. Hospital Frailty Risk Scores were determined based on the list of ICD-10-CM codes provided by the risk score authors.²⁰

Proposed risk-standardized quality-metrics

1. “Major morbidity during index admission” was defined as the occurrence of ≥1 of the following complications: ^21–23 pneumonia, pulmonary embolism, acute renal failure (or acute on chronic renal failure), cardiovascular accident, acute myocardial infarction, cardiac arrest, acute respiratory distress syndrome, and sepsis. Complications were calculated using ICD-10-CM primary/secondary diagnosis codes.

2. “Index hospital length of stay” was defined as the length of time that patients spent hospitalized during their initial EGS admission.

3. “Ability to be discharged home” was defined as the percentage of patients routinely discharged among all patients discharged alive following index EGS admission. Patients discharged to higher levels of care, e.g., skilled nursing facilities or inpatient rehabilitation facilities, were not counted as home discharges.

4. “Readmission within 30 days” was measured from the data of discharge among patients discharged alive. It included patients readmitted for any reason. Patients discharged as transfers or who were discharged and readmitted within the same day (presumed transfers) were not counted as readmissions.

5. “Patients’ total number of hospitalized days within 6 months” was used as a proxy measure to mimic outcomes for patients’ average number of HDAH, which cannot be fully ascertained without access to clinical records or both inpatient and outpatient claims.²⁴ It was measured from the date of discharge among patients discharged alive and calculated as the total amount of patients’ time spent readmitted during the subsequent six months. Six months were used to ensure that all patients had complete follow-up after admission in January-June.

Following an approach defined by the Centers for Medicare & Medicaid Services to calculate risk-standardized quality-metrics^25–29 and more recent recognition of the need to incorporate volume-based reliability-adjustment of risk-standardized quality-metrics,^30,31 hospital performance on each of the proposed non-mortality-based quality-metrics was calculated using risk-adjusted hierarchical regression with hospital-level random intercepts and patient-level fixed-effects followed by reliability-adjustment of the resultant risk-standardized rates. Analyzing data in this way accounted for clustering of patients within hospitals, unstable estimates due to small sample-size, and risk-adjustment for known confounders. Similar methodology has been employed in the assessment of potential quality-metrics among older adult trauma patients.³²

More specifically, within each cohort (simple adults, complex adults, simple older adults, and complex older adults), we calculated condition-specific risk-standardized outcomes using mixed-effects hierarchical regression models. Models’ pooled random-effects intercepts were used to determine hospitals’ expected rates.^25–29 Predicted rates were taken from random intercepts. Risk-standardized rates were then calculated as the ratio of hospitals’ predicted/expected rates multiplied by their unadjusted rates.^25–29 To incorporate reliability-adjustment, we modeled a risk-standardization algorithm using an empirical-Bayes estimator.³³ To further allow for shrinkage of estimates closer to hospitals with similar volumes, median six-month EGS volumes were included as an additional hospital-level fixed-effect.³⁰

Categorizing hospital performance

Potential differences in hospital performance within each cohort were assessed using K-means cluster analysis. Prior to inclusion in cluster models, risk-standardized quality-metrics were mathematically “standardized” (set to have mean=0, variance=1) to ensure that they would be given equal weight. The elbow, silhouette, and gap statistics methods of evaluating clusters were used to determine the optimal number. Results were plotted along the first and second principal components to examine the amount of between hospital variability explained.

Assessing for associations with hospital-level factors

Using the results of the cluster analysis, multinomial regression models were used to determine what hospital-level factors influenced cluster assignment. They compared potential associations between overall performance and a hospital’s median number of EGS procedures within six months, mean number of comorbidities per patient, mean All Patients Refined Diagnosis Related Group (APR-DRG) Risk of Mortality (NRD-provided variable), mean APR-DRG Severity (NRD-provided variable), mean patient age, mean Hospital Frailty Risk Score, mean percentage of complex EGS patients, hospital bedsize (determined by NRD and categorized as small, medium, and large), hospital rurality (categorized as large metropolitan area with >1 million population, small metropolitan area with <1 million population, suburban, and rural), hospital teaching status, and hospital ownership (government, private non-profit, and private for profit).

Data analysis and cleaning were conducted using Stata Statistical Software: Version 17.0. Statistical analyses were conducted using Stata and R. Graphs were plotted in R. Missing data were minimal (<1.0% across all confounders) and, when present, were addressed using multiple imputation. Due to large sample size, two-sided p-values <0.01 were considered significant. The study was approved by the Human Investigation Committee of [investigators’ university]. Data were reported in accordance with STrengthening the Reporting of OBservational Studies in Epidemiology (STROBE) guidelines.

Results

Population characteristics

A total of 1,130,496 index admissions from 984 hospitals were included. Forty percent (40.6%; n=458,981) were simple adults. An additional forty percent (39.5%; n=446,546) were simple older adults. Fourteen percent (13.5%; n=152,617) were complex adults, and six percent (6.4%; n=72,352) were complex older adults. Distributions of patient-level demographic characteristics included in risk-adjustment are presented in Table 1.

Table 1.

Differences in patient-level demographic factors and unadjusted outcomes

	Simple Adult	Complex Adult	Simple Older Adult	Complex Older Adult
Patients January-June 2019	458,981 (40.6%)	152,617 (13.5%)	446,546 (39.5%)	72,352 (6.4%)
Hospitals	984	984	984	984
Mean age (SD)	48.1 (12.0)	45.0 (12.6)	77.3 (7.9)	75.5 (7.6)
Female	48.1%	47.5%	55.5%	51.3%
Mean number of comorbidities (SD)	3.1 (2.3)	2.6 (2.2)	4.4 (2.3)	4.2 (2.3)
Elixhauser Comorbidity Index
Congestive Heart Failure	10.4%	7.0%	28.5%	23.4%
Cardiac Arrhythmia	11.2%	9.0%	36.3%	33.0%
Valvular Disease	2.7%	2.0%	10.8%	8.2%
Pulmonary Circulation Disorder	3.1%	2.1%	7.4%	5.5%
Peripheral Vascular Disease	5.9%	5.0%	15.8%	18.5%
Hypertension, uncomplicated	34.6%	29.7%	42.7%	44.5%
Hypertension, complicated	14.7%	11.3%	37.5%	33.3%
Paralysis	1.7%	1.4%	1.4%	1.2%
Neurological Disorder	9.1%	6.0%	13.7%	11.6%
Chronic Pulmonary Disease	19.2%	15.4%	27.5%	23.3%
Diabetes, uncomplicated	10.9%	9.2%	13.3%	12.0%
Diabetes, complicated	18.2%	21.7%	22.9%	28.9%
Hypothyroidism	8.7%	6.2%	20.1%	17.1%
Renal Failure	12.5%	10.1%	29.1%	26.5%
Liver Disease	13.6%	7.2%	6.7%	5.6%
Peptic Ulcer Disease (not bleeding)	2.1%	0.5%	3.0%	1.0%
AIDS/HIV	0.8%	1.0%	0.1%	0.1%
Lymphoma	0.7%	0.5%	1.5%	1.4%
Metastatic Cancer	3.1%	1.9%	4.4%	4.0%
Solid Tumor, wilhout metastases	4.7%	3.2%	7.9%	7.6%
RA/Collagen Vascular Disease	3.2%	2.5%	4.7%	4.3%
Coagulopathy	8.5%	5.3%	9.3%	8.5%
Obesity	22.6%	21.8%	15.5%	16.5%
Weight Loss	8.5%	8.0%	13.3%	16.7%
Fluid and Electrolyte Disorders	35.7%	30.1%	45.1%	44.3%
Blood Loss Anemia	1.3%	0.9%	2.2%	1.2%
Deficiency Anemia	5.5%	4.6%	6.7%	0.5%
Alcohol Abuse	13.8%	6.1%	3.4%	2.9%
Drug Abuse	11.1%	15.6%	1.6%	1.7%
Psychoses	2.6%	2.2%	1.2%	1.1%
Depression	16.2%	12.9%	14.3%	12.6%
Hospital Frailty Risk Score
Low Risk < 5	59.5%	62.3%	35.2%	37.8%
Intermediate Risk 5–15	37.8%	35.7%	56.2%	55.0%
High Risk >15	2.7%	2.0%	8.7%	7.2%
Index admission mortality	1.6%	1.5%	4.4%	6.8%
Death on readmission (within 30 days)	0.9%	0.6%	2.2%	1.9%
Major morbidity	37.5%	46.9%	50.0%	54.6%
Readmission (within 30 days)	16.9%	13.8%	17.4%	17.6%
Able to be discharged home	74.4%	66.8%	44.2%	35.8%
Median index length of stay (IQR)	4 (2–7)	4 (3–8)	5 (3–8)	6 (4–11)
Median hosp. days within 6 mo. (IQR)	6 (3–15)	6 (3–15)	8 (4–17)	10 (5–21)

Two-sided p-values omitted due to large sample size (<0.001 in each case)

Abbreviations: SD standard deviation, RA rheumatoid arthritis, IQR interquartile range

Among adults, simple and complex EGS patients presented with average ages of 48.1±12.0 and 45.0±12.6 years, respectively. 48.1% and 47.5% were female. They had mean multimorbidities of 3.1±2.3 and 2.6±2.2. Forty percent of EGS patients in each group (simple adult: 40.5%, complex adult: 37.7%) presented with intermediate or high risk frailty. Among older adults, simple and complex EGS patients had average ages of 77.3±7.9 and 75.5±7.6 years, respectively. 55.5% and 51.3% were female. They had mean multimorbidities of 4.4±2.3 and 4.2±2.3. Sixty percent of EGS patients in each group (simple older adult: 64.8%, complex older adult: 62.2%) presented with intermediate or high risk frailty.

Unadjusted outcomes

During index admission, 1.6% of simple adults and 1.5% of complex adults died (Table 1). An additional 0.9% and 0.6% died on readmission within 30 days, yielding combined 30-day mortality rates of 2.5% and 2.1%, respectively. Major morbidity was more common. 37.5% of simple adults and 46.9% of complex adults experienced major morbidity during index admission. 74.4% (simple) and 68.8% (complex) of surviving adults were able to be discharged home. Both simple and complex adult EGS patients spent a median of 4 (IQR: 2–7, 3–8) days in the hospital during their index inpatient stay. 16.9% (simple) and 13.8% (complex) required readmission within 30 days, and, over the course of the subsequent 6 months, both simple and complex adults spent a median of 6 (IQR: 3–15, 3–15) days hospitalized.

Among older adults, 4.4% of simple and 6.8% of complex EGS patients died during index admission (Table 1). An additional 2.2% and 1.9% died on readmission within 30 days, yielding combined 30-day mortality rates of 6.6% and 8.7%, respectively. Major morbidity was again more common with 50.0% of simple older adults and 54.6% of complex older adults experiencing major morbidity during index admission. 44.2% (simple) and 35.8% (complex) of surviving older adults were able to be discharged home. Simple older adults spent a median of 5 (IQR: 3–8) days in the hospital during their index inpatient stay, while complex older adults spent a median of 6 (IQR: 4–11) days. 17.4% (simple) and 17.6% (complex) required readmission within 30 days. During the subsequent 6 months, simple older adults spent a median of 8 (IQR: 4–17) days hospitalized. Complex older adults spent a median of 10 (IQR: 5–21) days.

Risk-standardized (risk-adjusted) quality-metrics

Distributions of condition-specific risk-standardized quality-metrics followed expected patterns when stratified by age/severity. For example, among adults with a primary diagnosis of acute appendicitis, complex adults tended to have higher risk-standardized major morbidity (Figure 1A), greater risk-standardized need for readmission within 30 days (Figure 1B), longer risk-standardized median index hospital LOS (Figure 1C), a greater risk-standardized number of median hospitalized days within 6 months (Figure 1D), and a lower risk-standardized probability of being able to be discharged home (Figure 1E) when compared to simple adult patients.

Figure 1.

Distributions of condition-specific risk-standardized quality-metrics for simple and complex adults with acute appendicitis

Performance clusters: best, average, and worst overall EGS care

On K-means cluster analysis, three groups of hospitals within each cohort were identified: “best,” “average,” and “worst.” Results for simple adults are shown in Figure 2. Clusters collectively explained >60.0% percent of variability between hospitals based on the results of the first principal component alone (60.5% for simple adults). Across cohorts, cluster assignment was highly conserved with 95.3% of hospitals assigned to the same cluster in each cohort. Within cohorts (Figure 3), differences in standardized (mean=0, SD=1) risk-standardized quality-metrics revealed a consistent trend for best/average/worst performance that persisted across differences in outcomes (5x) and EGS conditions (16x). Taken together, such findings suggest that (1) hospitals that performed well on one condition-specific quality-metric tended to perform well on them all and (2) significant differences in cluster assignment (p<0.001) explained >60.0% of variability between hospitals based on the considered set of non-mortality-based quality-metrics.

Figure 2.

Results of K-means cluster analysis for simple adults showing an optimal solution of three clusters based on the elbow method of cluster determination (slowing of meaningful change in the percent of variation explained [reduction in the total within-cluster sum of squared errors] by adding an additional cluster, i.e., the elbow in the graph) (A), minimal overlap along the first principal component when using three clusters (B), and significant separation between centroids (significant difference in resultant cluster centers, two-sided p-value<0.001)

Figure 3.

Consistency of K-means cluster performance (best: green triangles, average: red circles, worst: blue squares) for simple adults across differences in “standardized” (mean=0, SD=1) risk-standardized outcomes (5x) and EGS conditions (16x)

Associations with hospital-level factors

Associations with hospital-level factors are presented in Table 2. Compared to worst- and average-performing hospitals, best-performing hospitals tended to be those with the highest EGS volume, most complicated patient case-mix (including differences in the average number of patient comorbidities, APR-DRG Risk of Mortality, and APR-DRG Severity), and greatest extent of frailty (two-sided p-value <0.001 for each). Differences in average EGS severity and patient ages were not significant (two-sided p-value >0.01). However, best performance was associated with significant differences (two-sided p-value <0.001 for each) in hospital bedsize (54.8% of best-performing hospitals were “large”), rurality (65.4% of best-performing hospitals were located in metropolitan areas with >1 million population), teaching status (76.5% of best-performing hospitals were teaching institutions), and ownership (81.1% of best-performing hospitals were private non-profit organizations).

Table 2.

Hospital-level associations with best, average, worst performance

	Worst Performance	Average Performance	Best Performance
Hospitals	313	454	217	p-value*
Volume: median in 6 months (IQR)	634 (440–912)	907 (625–1275)	939 (608–1387)	<0.001
Case-mix: mean number of comorbidities (SD)	3.1 (0.4)	3.7 (0.3)	4.1 (0.3)	<0.001
Case-mix: mean APR-DRG Risk of Mortality (SD)	1.9 (0.1)	2.1 (0.1)	2.2 (01)	<0.001
Case-mix: mean APR-DRG Severity (SD)	2.3 (0.1)	2.5 (0.11)	2.6 (0.1)	<0.001
Case-mix: mean patient age (SD)	60.6 (4.0)	61.2 (3.6)	60.5 (3.6)	0.033
Frailty: mean Hospital Frailty Risk Score (SD)	5.0 (0.7)	6.0 (0.5)	6.8 (0.7)	<0.001
AAST Anatomic Severity: mean % complex (SD)	19.5% (4.1%)	19.6% (3.4%)	20.4% (4.0%)	0.024
Hospital Bedsize, NRD determined				<0.001
Small	24.9%	16.3%	18.4%
Mediium	38.3%	36.1%	26.7%
Large	36.8%	47.6%	54.8%
Hospital Rurality				<0.001
Large metropolitan > 1 million	57.8%	55.5%	65.4%
Small metropolitan < 1 million	31.6%	41.2%	31.3%
Suburban	9.6%	3.3%	2.8%
Rural	1.0%	<0.1%	0.5%
Teaching Hospital	50.2%	65.6%	76.5%	<0.001
Hospital Ownership				<0.001
Government	15.0%	8.8%	8.8%
Private, non-profit	64.2%	75.3%	81.1%
Private, for profit	20.8%	15.9%	10.1%

^*Two-sided p-values <0.01 considered significant (bold) to account for the large sample size

Abbreviations: IQR interquartile range, SD standard deviation, NRD Nationwide Readmissions Database

Discussion

This national quality assessment of EGS patients conducted using a novel set of five non-mortality-based quality-metrics designed to be broadly applicable to the care of all EGS patients demonstrates that meaningful and consistent benchmarking of EGS is possible. Using the analytical approach outlined in the methods, best-performing hospitals were successfully identified based on the proposed set of condition-specific risk-standardized quality-metrics. Results were highly conserved among adults and older adults with both simple and complex EGS severity as well as across differences in outcomes and EGS conditions. Taken together, the results suggest that hospitals that tend to perform well on one condition-specific quality-metric in a given patient population (e.g. ability to be discharged home among simple adults with acute appendicitis) tend to perform well on them all. Those that perform well are likely to be large, private non-profit teaching hospitals located in urban areas managing the highest volume of EGS patients, greatest extent of patient frailty, and most complicated underlying patient case-mix. Analyzing data in this way showed that use of non-mortality-based quality-metrics appears to offer a needed, promising means of evaluating high-quality EGS care—one which yielded significant differences in best/average/worst performance and explained upwards of 60.0% of variability between hospitals in addition to providing meaningful associations with hospital-level factors.

Drawing from a historical focus in trauma on in-hospital mortality,³² early efforts to benchmark EGS quality primarily looked at differences in in-hospital mortality among operative patients. Their results suggest that for both adults^34,35 and older adults,^36,37 low mortality for one procedure is associated with low mortality for multiple procedures. The results of our study agree, confirming the persistence of best/average/worst performance across differences in EGS conditions, outcomes, and ages/severities. Where our results differ is in their inclusion of non-operative patients, the utilized quality-metrics’ ability to account for patients with low mortality-risk, and the resultant clusters’ ability to define meaningful associations with hospital-level factors. In contrast to prior mortality-based research,^34–37 our study demonstrated clear associations between EGS quality and factors likely to delineate the existence of natural centers-of-excellence specializing in the provision of high-quality EGS care. It showed better risk-standardized outcomes among hospitals with the highest volume of EGS patients managing the most sick and frail. Where available, similar results have been found for differences in EGS mortality among operative patients (e.g. higher hospital quality is associated with a greater proportion of EGS admissions to trauma centers, larger hospital bed-size, and more ready access to “high technology capability”).³⁵

What research is available looking at the quality of outcomes for lower mortality-risk conditions suggests that room for improvement remains,³⁸ which is not accounted for based on differences in mortality³⁹ or commonly-monitored patient-level factors.⁴⁰ For example, when benchmarking based on differences in total hospital costs in order to reflect differences in EGS value (value = quality/cost), variations in volume explained up to 12.4% of differences between hospitals among patients with acute cholecystitis and 9.9% of differences among patients with acute appendicitis.³⁸ Similar associations have been found to influence mortality.^13,14 The results of this study add to that work, highlighting the importance of hospital-level factors such as EGS volume and a hospital’s extent of experience dealing with frail and complex patients in explaining differences in outcomes beyond those captured based on risk-adjustment for variations in patient-level factors alone. When looking at differences in non-mortality-based quality-metrics, meaningful distributions of risk-standardized quality-metrics for conditions like acute appendicitis (Figure 2) could readily be attained.

By moving beyond in-hospital mortality, the proposed set of non-mortality-based quality-metrics take into account the heterogeneity of EGS conditions. They address the reality that common metrics used by other specialties (e.g. in-hospital mortality for trauma) might not be applicable to all EGS conditions and include consideration of important patient-level factors like frailty, the extent multimorbidity, and EGS severity which have historically been overlooked in prior efforts to monitor the quality of EGS care. By taking such an approach, we were able to look at more common adverse outcomes broadly applicable to the care of all EGS patients: >38% of adults and >50% of older adults experienced major morbidity during index admission versus 2% and 7% who died within 30 days (<0.1% among patients with acute cholecystitis and acute appendicitis). We accounted for differences in health-system performance and care coordination³² through the use of quality-metrics like LOS and the need for readmission within 30 days, and, critically, we accounted for the importance of patient preferences for increased time at home⁴¹ through the use of quality-metrics addressing patients’ ability to be discharged home and the extent of time patients spent at home during the available follow-up period.

Emerging literature in trauma^32,42 and among older adult EGS^43,44 patients has highlighted the need to account for patient-centered outcomes and patients’ extent of functional recovery through the use of quality-metrics that address patients’ average number of HDAH. Days in hospital during the six months following a patient’s index EGS admission was used as a proxy measure of this outcome to account for the nature of NRD data,¹⁹ which precludes the ability for patients to be tracked across calendar years or monitored for time spent not at home outside of a hospital inpatient setting. Future studies in datasets with access to both inpatient and outpatient longitudinal claims (e.g. MarketScan, Medicaid, and/or Medicare) are warranted to explore how formal use of HDAH compares.

Never intended to provide an exhaustive list of quality-metrics for hospitals to track, the 80 condition-specific risk-standardized quality-metrics (5x outcomes across 16x EGS conditions) utilized in this study to identify groups of best-, average-, and worst-performing hospitals demonstrate that meaningful quality benchmarking in EGS is possible. Moving forward from here, what the quality-metrics are intended to do is to provide a starting point to spur needed conversations about plausible means for how to evaluate the quality of EGS care that is reflective of the outcomes of all EGS patients and to support ongoing EGS-QI initiatives that urgently require novel definitions for high-quality EGS care beyond the results available based on in-hospital mortality alone.^3,18 On the whole, the results of this study suggest that the use of non-mortality-based quality-metrics appears to offer a needed, promising means for evaluating high-quality EGS care, one that can be easily measured using existing clinical data (e.g. “EGS-NSQIP”¹¹) and/or administrative billing claims (e.g. NRD, Medicare). For smaller, more rural hospitals operating outside of larger EGS-verified centers,^3,12 the results are hoped to help aid in emerging efforts to appropriately identify: (1) where triage and stabilization of EGS patients is warranted; (2) how to strengthen resources, experience, and training available to providers at all levels of care; (3) how to develop interhospital/interhospital-system partnerships that facilitate optimal outcomes for patients; and (4) how to help any poorly-performing hospital (regardless of EGS volume or size) identify areas in which they can improve.

The study has limitations. The most important reflect its reliance on administrative claims and their related lack of nuanced clinical detail, potential for absent or misreporting of events, and inability to detect adverse outcomes that occur outside of an inpatient setting. Use of NRD allowed for a large national assessment of EGS patients, including both adults and older adults. However, in relying on NRD claims, we were not able to capture the outcomes of EGS patients in non-participating states, measure outcomes experienced by patients outside of the inpatient setting (e.g. time spent in rehabilitation or skilled nursing facilitates), or follow-up patients across calendar years and beyond 180 days. Quality-metrics used in this study followed established methodology for the calculation of risk-standardized outcomes.^25–32

As efforts to understand EGS quality continue to develop, meaningful approaches are needed to gauge hospital quality and establish targets capable of best directing limited resources and improving future care for the increasing number of EGS patients. The results of our study suggest that when evaluating EGS quality based on a proposed set of five non-mortality-based quality-metrics, best-performing hospitals could be identified. They included differences in outcomes for all EGS patients; yielded consistent results across variations in patient age, severity, and each AAST-defined EGS condition; and identified associations with important hospital-level factors, including volume, frailty, and a hospital’s complexity of patient case-mix. As a first step in answering calls for the needed development of quality-metrics capable of reflecting the outcomes of all patients presenting for EGS care,^3,18 the results point toward important opportunities to increase ways of looking at quality and better capture the experience of all EGS patients. Ultimately, the proposed set of additional non-mortality-based quality-metrics provided a more complete picture of EGS care: one that overcomes limitations of relying on in-hospital mortality alone, accounted for care-transitions and multiple provider perspectives, and included patient-centered outcomes all in a way that could be easily measured using existing billing claims. Future research is needed to determine how they should best be used.