Racial Disparities in Treatment of Ruptured Abdominal Aortic Aneurysms

Abstract

Objectives:

SVS recommends immediate transfer of patients with ruptured abdominal aortic aneurysm (rAAA) to a regional center when feasible. However, Black patients may be less likely to be transferred, and more likely to be turned down for repair. We therefore examined transfer rates, turndown rates, and outcomes in Black versus White patients presenting with rAAA in two large databases.

Methods:

We examined all rAAA repairs in the Vascular Quality Initiative (VQI) from 2003–2020 to evaluate transfer rates and outcomes in Black versus White patients. We used the National Inpatient Sample (NIS) from 2004–2015 to examine turndown rates. Mixed effects logistic regression and cox regression as well as marginal effects modeling were used to study the interaction between race, insurance status, type of operation (open repair vs EVAR), and hospital volume.

Results:

We identified 4,935 rAAA in the VQI (6.2% Black) and 48,489 in the NIS (6.0% Black). Rates of transfer were high, but Black patients were significantly less likely to undergo transfer prior to repair compared with White patients (Black 49% vs White 62%; P=.002). This was consistent in both crude and adjusted analyses when considering only stable patients, and was not modified by insurance status, type of operation, or hospital volume. There was no significant difference in perioperative mortality (22% vs 26%; P=.098) or complications (52% vs 52%; P=.64). However, Black patients were significantly more likely to be turned down for repair (37% vs 28%, OR 1.5 [1.2 – 1.9], P <.001). There was a significant interaction between race and insurance status with respect to turndown. Patients with private insurance underwent operations at similar rates regardless of race, but among patients with Medicare or Medicaid/self-pay, Black patients were less likely than Whites to undergo repair (Medicare: 64% vs 72%, P=.001; Medicaid/self-pay: 43% vs 61%, P=.031). Patients with Medicaid/self-pay were also less likely to undergo repair when compared with patients of the same race with either Medicare or private insurance (P <.05).

Conclusion:

Black patients with rAAA are poorly served by current systems of interhospital transfer in the United States, as they less often undergo transfer prior to repair. Although postoperative outcomes appear similar, this may be false optimism, as Black patients, especially the underinsured, are more often turned down for repair even after adjustment. Significant work is needed to better understand the reasons underlying these disparities and identify targets to improve the care of Black patients with rAAA.

Introduction:

In recent years, increased focus has been brought to bear on racial disparities in the treatment of patients with cardiovascular diseases. A wealth of data currently exists about inequities in the medical and surgical management of entities ranging from lower extremity amputations to elective abdominal aortic aneurysm repairs.^1–10 However, less is known about the management of ruptured abdominal aortic aneurysms (rAAA).

Aortic rupture carries a high rate of early morbidity and mortality. The management of rAAA is complex and involves both local and regional factors. Patients with rupture first need prompt diagnosis and triage, and then either rapid transfer to the operating room, or to another facility with the appropriate capabilities. Disparities can manifest in prolonged waiting in the emergency department before imaging and evaluation, failure to transfer to another facility if needed, delay in operating, and being turned down for repair. Therefore, rAAA management and outcomes provide a useful spotlight into the potential barriers to equitable care for patients with rAAA, but also vascular patients at large. We utilized two large, national databases to study these factors and identify key areas for improvement.

Methods:

Study Population:

This analysis was carried out in two parts, leveraging one clinical registry and one administrative database. For the first part, we examined all White or Black patients undergoing repair of rAAA without prior aortic surgery in the Society for Vascular Surgery’s Vascular Quality Initiative (VQI) from 2003–2020. The VQI records patient demographics, transfer status, anatomic variables, comorbid conditions, perioperative complications, one-year follow-up, and long-term mortality through linkage to the Social Security Death Index. The granular detail captured in this registry makes it ideal for studying perioperative outcomes, volume-outcome relationships, and long-term survival. However, the VQI only includes patients who underwent an operation, and only captures a subset of hospitals. Therefore, we supplemented our analysis by studying all White or Black patients with a diagnosis of rAAA in the National Inpatient Sample (NIS) from 2004-September 2015 using the ICD-9 diagnosis code 441.3-Abdominal Aneurysm with Rupture. Patients undergoing repair were identified with ICD-9 procedure codes (38.34-Aorta resection and Anastomosis, 39.25 Aorta-Iliac femoral bypass, 38.44-Replacement of Abdominal Aorta, 38.64-Excision of Aorta, 39.52-Other repair of Aneurysm, 39.71-Endovascular Abdominal Aorta Repair). Although the NIS lacks the granular detail of the VQI, it captures patients who presented with rAAA but did not undergo repair (the turndown group), and records the type of facility (Rural, Urban Teaching, and Urban Non-teaching). For consistency, we terminated the NIS period at the changeover from ICD-9 to ICD-10 coding. The Beth Israel Deaconess Medical Center Institutional Review Board approved this study and waived the need for patient consent due to the nature of the design and minimal risk to human subjects. This manuscript was written in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines.¹¹

Definitions and Variables:

VQI:

Race in the VQI is entered by surgeons or trained data entry nurses and reflects the patient’s self-identification reflected in the medical record. We calculated the estimated glomerular filtration rate (eGFR) by the CKD-EPI formula. This allowed us to stratify chronic kidney disease by the National Kidney Foundation Guidelines into eGFR >60 mL/min/1.73m², 45 – 60 mL/min/1.73m², 30 – 44 mL/min/1.73m², and < 30 mL/min/1.73m². We classified smoking status as never, former, or current (i.e. within 30 days of the procedure), and stratified chronic obstructive pulmonary disease (COPD) into none, mild-moderate, and requiring home oxygen. Congestive heart failure (CHF) we categorized as none, mild (New York Heart Association; NYHA Class I or II), or moderate to severe (NYHA Class III or IV). Insurance status was based on the primary insurer listed (Private, Medicare, Medicaid, self-pay). Based on previous work, we considered preoperative anemia as a hemoglobin < 10 mg/dL.¹² We classified postoperative renal dysfunction, or acute kidney injury (AKI), by the RIFLE criteria based on the peak postoperative creatinine value.¹³ The VQI defines stroke as any new motor or sensory loss, speech abnormality, or documentation of any other neurological symptoms related to the right or left hemisphere, lasting ≥24 hours. We defined perioperative death as occurring within 30 days or discharge dead from index hospitalization. Myocardial infarction (MI) includes elevated troponins or electrocardiographic changes with symptoms. We defined any complication as death, re-intervention, AKI, stroke, transient ischemic attack (TIA), access site complication, MI, new arrhythmia, new onset CHF, pneumonia, re-intubation, or mesenteric ischemia. Hospital and surgeon volumes were calculated based on the previous year’s volume of elective, symptomatic and ruptured aneurysms to allow for changes in volume over time, and classified into quartiles. We defined a patient as unstable if they met one of the following criteria: a cardiac arrest prior to the operation, a disoriented or unconscious mental status, or a preoperative blood pressure of <80mm Hg.

NIS:

For all patients, we used the predefined Elixhauser covariates to identify comorbidities.¹⁴ Insurance was based on their primary insurer (Private, Medicare, Medicaid, self-pay). Turndown was defined as presence of ICD-9 diagnosis code for rAAA without any associated ICD-9 procedure code for repair (as defined above). Outcomes of interest were admission, turn-down rates, treatment (open vs EVAR), and in-hospital mortality for patients who presented with rAAA stratified by race.

Statistical Analysis:

VQI:

We compared patient characteristics using chi-squared and rank-sum tests where appropriate. Linear regression was used to assess changes over time. For adjusted analyses, we constructed mixed effects logistic regression models with clustering at the center level, and used these to calculate the marginal effects of race, transfer status, and insurance. Interaction terms were used to study their interaction. Cox regression was used for long-term outcomes. Variables were chosen a priori, and included age, race, sex, open/endovascular repair, aortic diameter, year of surgery, transfer status, insurance, coronary artery disease (CAD), COPD, chronic kidney disease, concomitant iliac aneurysm, smoking status, CHF, prior MI, anemia, pre-operative medications (statin, aspirin, other antiplatelet, beta-blocker, ACE-inhibitor/Angiotensin receptor blocker, anticoagulants), and hospital and surgeon volume. Baseline variables are well-captured by the VQI, and rates of missing were <3%.

NIS:

Patients were classified based on the final hospital they arrived at. We used the chi-squared test to compare patient characteristics. We used mixed effects logistic regression to calculate the marginal effects on race, turn down and insurance. These were adjusted for age, heart failure, hypertension, diabetes, renal failure and obesity per predefined Elixhauser definitions.

All statistical analyses were performed using STATA14.2 (StataCorp, College Station, Texas). A p value < .05 was considered statistically significant.

Results:

There were 4,935 repairs of rAAA in the VQI during the study period; 6.2% were Black patients, and 59% were treated with EVAR. Demographics of the VQI cohort are presented in Table 1. Black patients were younger, more often female, had smaller aortic diameters, more iliac aneurysms, and higher rates of comorbidities such as hypertension, smoking and diabetes (all P<.001). In addition, Black patients were more often treated at low volume centers by low volume surgeons (P<.01), but had similar rates of hemodynamic instability on presentation (32% vs 37%, P=.07). White patients were more often on aspirin (P<.001). Of the 1,047 patients with anatomic data regarding the aortic neck, Black patients had longer aortic necks (median 25mm [Interquartile Range 15.5–32] vs 20mm [15–30], p=.02), and similar neck diameters and neck angulation.

Table 1.

Demographics of patients with ruptured abdominal aortic aneurysms in the Vascular Quality Initiative.

		White n=4631	Black 304	p-value
Age, mean (SD)		72.4 (9.6)	69.5 (11.4)	<0.001
Male Sex	Male	3579 (77.3%)	204 (67.1%)	<0.001
Insurance	Private	1501 (37.5%)	101 (35.3%)	<0.001
	Medicare	2226 (55.6%)	144 (50.3%)
	Medicaid/Self-Pay	274 (6.8%)	41 (14.3%)
Diameter	mean (SD)	73.2 (19.4)	68.0 (23.3)	<0.001
	<4.5 cm	221 (5.1%)	41 (14.5%)	<0.001
	4.5–5.0 cm	188 (4.3%)	13 (4.6%)
	5.0–5.5 cm	232 (5.3%)	28 (9.9%)
	5.5–6.0 cm	380 (8.7%)	31 (11.0%)
	6.0–6.5 cm	385 (8.8%)	23 (8.1%)
	6.5–7.5 cm	901 (20.6%)	51 (18.0%)
	7.5+ cm	2061 (47.2%)	96 (33.9%)
Aortic Neck Length (mm)	median [IQR]	20 [15–30]	25 [16–32]	0.02
Aortic Neck Diameter (mm)	median [IQR]	24 [22–28]	23 [20–28]	0.054
Aortic Neck Angle	< 45 degrees	768 (79%)	63 (89%)	0.27
	45–60 degrees	128 (13%)	6 (8.5%)
	>60 degrees	80 (8%)	2 (3%)
Iliac Aneurysm		791 (17.1%)	96 (31.6%)	<0.001
Ever Smoker		3670 (79.2%)	217 (71.4%)	0.001
Hypertension		3481 (75.8%)	260 (87.4%)	<0.001
Diabetes		681 (14.7%)	71 (23.4%)	<0.001
CAD		1531 (33.1%)	90 (29.6%)	0.21
Chronic Kidney Disease	None	2118 (45.7%)	135 (44.4%)	<0.001
	eGFR 30–60	1952 (42.2%)	102 (33.6%)
	eGFR <30	522 (11.3%)	54 (17.8%)
	Dialysis	39 (0.8%)	13 (4.3%)
COPD	None	3173 (69.1%)	237 (78.2%)	0.010
	Yes	530 (11.5%)	25 (8.3%)
	On Medications	676 (14.7%)	30 (9.9%)
	Home Oxygen	214 (4.7%)	11 (3.6%)
Congestive Heart Failure	None	4123 (89.8%)	257 (84.8%)	0.014
	Mild	397 (8.6%)	37 (12.2%)
	Moderate-Severe	70 (1.5%)	9 (3.0%)
Body Mass Index	mean (SD)	29.2976 (55.65973)	28.42 (8.67)	<0.001
	Underweight	165 (3.6%)	22 (7.2%)	0.001
	Normal	1163 (25.1%)	94 (30.9%)
	Overweight	1514 (32.7%)	91 (29.9%)
	Obese	1286 (27.8%)	71 (23.4%)
	Morbidly Obese	503 (10.9%)	26 (8.6%)
Aspirin		1799 (39.6%)	95 (31.6%)	0.033
P2Y12 Inhibitor		343 (7.6%)	17 (5.6%)	0.22
Dual Antiplatelet		236 (5.2%)	10 (3.3%)	0.15
Statin		1908 (42.8%)	116 (39.2%)	0.24
Beta Blocker		1686 (37.1%)	110 (36.5%)	0.84
ACE or ARB	0	1205 (30.6%)	89 (31.4%)	0.78
Quartile of Hospital Volume	1	1148 (24.8%)	90 (29.6%)	0.008
	2	1152 (24.9%)	86 (28.3%)
	3	1154 (24.9%)	76 (25.0%)
	4	1177 (25.4%)	52 (17.1%)
Quartile of Surgeon Volume	1	1203 (26.0%)	107 (35.2%)	<0.001
	2	1145 (24.7%)	79 (26.0%)
	3	1113 (24.0%)	68 (22.4%)
	4	1170 (25.3%)	50 (16.4%)
Unstable		1722 (37.2%)	97 (31.9%)	0.065

SD: standard deviation. IQR: interquartile range. COPD: chronic obstructive pulmonary disease. ACE: angiotensin converting enzyme inhibitor. ARB: angiotensin receptor blocker.

Black patients were significantly less likely to undergo transfer prior to repair (Black: 48% vs White: 62%, P<.001). This was consistent in both stable (47% vs 63%, P<.001) and unstable patients (48% vs 61%, P=.017), and after multivariable adjustment. Similarly, Black patients were less likely to undergo transfer independent of insurance status or operative type (Figure 1). Black patients were more likely to undergo EVAR than White patients, irrespective of transfer status (73% vs 58%, P<.001). Differences in transfer rates between Black and White patients did not change over time (P=.2)

Figure 1.

Rates of Transfer Prior to Repair in Patients Presenting with Ruptured AAA by A) overall, B) stable and unstable patients (cardiac arrest prior to the operation, disoriented or unconscious mental status, or systolic blood pressure < 80 mmHg C) primary insurer D) repair type.

Low volume hospitals less often treated patients that were transferred in compared with higher quartiles of volume (lowest quartile: 48% vs upper three quartiles: 63–68%, P<.001). At each quartile, Black patients were less likely to be transferred in than White patients, although the gap did narrow at the highest quartile (58% vs 69%, P=.07). There was no difference in transfer rates for Black patients based on region.

Perioperative Outcomes

Perioperative mortality in the VQI was 28% overall. Although Black patients experienced lower perioperative mortality in crude analyses (22% vs 29%, P=.01), there was no difference after multivariable adjustment (22% vs 26%, P=.10). This was consistent across stable and unstable patients. Black patients experienced similar adjusted perioperative mortality regardless of whether they were transferred in for care (transfer: 21% vs non-transfer: 22%, P=.83), but White patients who were transferred experienced lower perioperative mortality than White patients who were not (25% vs 29%, P<.01).

Similarly, there were no differences in adjusted rates of in-hospital reoperation or perioperative complications between Black and White patients. Black and White patients both had adjusted reoperation rates of 22% (P=.85), and had complication rates of 52% (P=.98).

Long-term mortality was similar between White and Black patients. Black patients had 1,5, and 10-year survival of 70%, 60% and 58%, compared to 71%, 61% and 53% for White patients (P=.98). This was consistent after multivariable adjustment (HR 1.00 [0.8–1.3], P=.94).

Turndown rates

In the NIS there were 48,489 patients who presented with rAAA, of whom 6.0% were Black, and 31% were turned down for repair. In crude and adjusted analyses, Black patients were significantly more likely to be turned down for repair than White patients (37% vs 28%, OR 1.5 [1.2–1.9], P<.001). Characteristics of Black and White patients turned down for repair are shown in Table II. Black patients who were turned down were 5 years younger, with higher rates of renal dysfunction but similar rates of cardiac and pulmonary comorbidities. In addition, there was a significant interaction between insurance status and race with respect to turndown rates. Overall, patients with private insurance were less likely to be turned down for repair (Private: 27%; Medicare: 29%; Medicaid/self-pay: 40%). Moreover, Black patients were more likely to be turned down for repair than White patients if they had Medicare (36% vs 28%, P=.001) or Medicaid/self-pay (57% vs 39%, P=.03) (Figure 2). However, there was no statistically significant difference in turndown rates between Black and White patients with private insurance (38% vs 27%, P=.06). Rural hospitals had the highest rate of turndown for repair (rural; 38% vs urban non-teaching: 30% vs urban teaching: 27%, P < .001). Black patients were more often turned down than White patients at urban non-teaching (38% vs 29%, P = .023) and urban teaching hospitals (37% vs 26%, P <.001), but not rural hospitals (31% vs 38%, P=.5).

Table 2.

Patient characteristics of Black and White patients turned down for repair.

		White n=12,622	Black n=868	p value
Female		42%	44%	0.62
Age, Mean (SD)		80.5 (0.2)	74.8 (0.98)	<0.001
Congestive Heart Failure		13%	13%	0.98
Hypertension		60%	77%	<0.001
Diabetes		13%	19%	0.008
Renal Failure		18%	30%	0.001
Pulmonary circulation disorder		2%	6%	0.001
Chronic Pulmonary Disease		3%	2%	0.009
Insurance	Medicare	88%	77%	<0.001
	Medicaid/Self-Pay	3%	10%	<0.001
	Private Insurance	9%	13%	<0.001
Depression		6%	3%	0.056
Pyschosis		2%	3%	0.3
Drug Abuse		0%	4%	<0.001
Alcohol abuse		2%	2%	0.74
Anemia		15%	27%	0.0001
Weight loss		5%	9%	0.038
Obesity		4%	7%	0.058
Coagulopathy		6%	12%	0.003
Arthritis		2%	2%	0.79
Tumor		3%	2%	0.3
Metastatic disease		2%	3%	0.25

SD: standard deviation.

Figure 2.

Turndown rate. A) Overall turndown rate. P<.001 between White and Black patients. B) Turndown rate stratified by Primary Insurer. * denotes P<.05 for comparison.

Discussion:

This study demonstrates significant racial inequities in the care of patients with ruptured AAA, but also highlights several key areas for improvement. Black patients were less likely to be transferred prior to repair, irrespective of comorbid burdens or clinical stability. In addition, these data reveal that one important source of disparities in care is the significantly higher rates at which Black patients with rAAA are turned down for repair. Even after adjusting for comorbidities, Black patients had 50% higher odds of being turned down for repair compared to White patients.

Differences in turndown rates are one of the ways that disparities can persist undetected, and even provides false reassurance. Prior works have shown that Black patients have equivalent or higher survival compared with White patients following repair of rAAA, which is corroborated by our data.^15–18 Working only off those studies, one might conclude that no racial disparities exist. However, our data demonstrate that the outcomes of patients who underwent repair tell only one half of the story, as Black patients were significantly more likely to be turned down for repair. The only real hope for survival after aneurysm rupture is prompt repair, but Black patients were less often given that chance.

A retrospective analysis such as this cannot prove causation, but the weight of our data strongly suggests that systemic racism and/or implicit bias may play an outsized role in the triage and management of patients with rAAA. Black patients were less often transferred and less often operated on when presenting with rupture. Although we cannot definitively say that racism and/or bias was the reason underlying this discrepancy, our analysis eliminates several of the most commonly cited alternative explanations centered on inherent differences in presentation and comorbid burdens. Black and White patients had similar rates of presenting with hemodynamic instability (and indeed Black patients had a trend towards being less likely to be unstable). In addition, Black patients who were turned down had higher rates of renal dysfunction, but were five years younger and had similar rates of cardiac and pulmonary diseases. As such, it is hard to attribute such a large difference in turndown rates to either differences in presentation or comorbid burdens, which begs the question: what alternative causes remain? Although there may be some residual confounding such as the site of presentation; patient preferences; and other social determinants of health such as level of education; economic stability, and neighborhood; the association between race and insurance status with both transfer and turndown remains highly troubling, and should prompt reassessment of transfer and operative protocols to ensure they are more robust against bias.

This report adds to the growing body of literature demonstrating a concerning relationship between insurance status and the type and manner of care a patient receives. We found that patients with Medicaid were one and a half times as likely to be turned down for repair after adjusting for comorbidities. In addition, private insurance seemed to mitigate some, but not all, of the disparity between Black and White patients with respect to turndown rates. With each improvement in insurance status, from Medicaid to Medicare to private insurance, the gap between Black and White patients narrowed. Previous works have found that underinsured patients have worse disease at time of presentation, more often present for emergent surgery rather than elective, have more comorbidities, and suffer worse outcomes following repair.^19–28 Moreover, expanded Medicaid access has been shown to increase access to vascular surgery care.²³ Improving access to insurance and vascular surgical care therefore has the potential to improve outcomes on several fronts. Better access to primary care would improve the diagnosis and management of comorbid conditions, expanded access to vascular care would allow for earlier diagnosis of AAAs to facilitate repair on an elective rather than emergent basis, and patients with rupture would be more likely to undergo life-saving treatment. In conjunction with efforts to expand access to insurance and to care, efforts need to be taken to ensure that a patient’s insurance does not impact the decision to offer a life-saving operation. Although the difference between turndown rates between racial groups was not statistically significant in those with private insurance, it is worth noting that we are not concluding that the turndown rates in that subset are equivalent. The p value was still only .06, with an absolute difference of 11% (38% of Black patients with private insurance were turned down compared to 27% of privately insured White patients). With such a potentially large effect size and a close p value, there is a substantial potential for a type 2 error due to a lack of power. We therefore cannot conclude there is a statistically significant difference, but similarly cannot claim equivalence. It is encouraging that the gap narrows with increasing insurance status, but it is likely that a gap still remains that we are just underpowered to detect. Expanding access to insurance likely represents one solution to improve care and narrow the existing gaps, but more work will be needed to fully close the gap.

This study must be interpreted in the context of its design. This is a retrospective study that leverages the unique advantages of two large databases (one registry-based and one administrative), but it is also subject to their limitations. We cannot ascertain the reason patients do not undergo repair, and whether that was a surgeon or patient decision, and we lack anatomic details in the NIS database. However, in the subset of VQI patients with anatomic details regarding the aortic neck, there was no difference in neck anatomy between Black and White patients except that Black patients actually had longer aortic necks. This would argue against the notion that anatomic differences played an outsize role in the differential turndown rates. Similarly, we do not know why patients were or were not transferred. In addition, we only know the characteristics of the hospital a patient arrived at, not where they came from, so we don’t know if there are differences in the types of hospitals White and Black patients are presenting to. Furthermore, both databases only capture the primary insurance of each patient, so we cannot tell if patients have supplemental insurance, Medicare Advantage, etc. Similarly, we cannot ascertain if patients are dual-eligible for Medicare and Medicaid, which has been identified previously as a highly vulnerable population. In addition, although we can use insurance as a surrogate measure, we do not have direct measures of socioeconomic status. These limitations notwithstanding, we believe that the weight of our data suggests highly concerning trends that are worthy of further study as well as quality improvement initiatives.

Conclusion:

Black patients with rAAA are poorly served by the current systems of interhospital transfer in the United States, as they less often undergo transfer prior to repair. Although postoperative outcomes appear similar, this may be false optimism, as Black patients, especially the underinsured, are more often turned down for repair even after adjustment. Significant work is needed to better understand the reasons underlying these disparities and identify targets to improve the care of Black patients with rAAA.

Article Highlights.

Type of Research:

Multicenter, multi-database retrospective review of data from the Vascular Quality Initiative and National Inpatient Sample databases.

Key Findings:

Black patients who presented with ruptured abdominal aortic aneurysms were less likely to be transferred and were more often turned down for repair when presenting with ruptured aneurysms compared to White patients. However, private insurance did mitigate some of the association between race and turndown for repair.

Take Home Message:

Significant racial disparities exist in the disposition and treatment of Black and White patients who present with ruptured abdominal aortic aneurysms.

Table of Contents Summary:

This study compared transfer rates, turndown rates, and outcomes in Black versus White patients with ruptured abdominal aortic aneurysms in the Vascular Quality Initiative and National Inpatient Sample databases. Black patients were less likely to be transferred and were more often turned down for repair when presenting with ruptured aneurysms.