Longitudinal Spending on Endovascular and Open Abdominal Aortic Aneurysm Repair.

Supplemental Digital Content is available in the text. Background: Endovascular repair (EVR) has replaced open surgery as the procedure of choice for patients requiring elective abdominal aortic aneurysm (AAA) repair. Long-term outcomes of the 2 approaches are similar, making the relative cost of caring for these patients over time an important consideration. Methods and Results: We linked Medicare claims to Vascular Quality Initiative registry data for patients undergoing elective EVR or open AAA repair from 2004 to 2015. The primary outcome was Medicare’s cumulative disease-related spending, adjusted to 2015 dollars. Disease-related spending included the index operation and associated hospitalization, surveillance imaging, reinterventions (AAA-related and abdominal wall procedures), and all-cause admissions within 90 days. We compared the incidence of disease-related events and cumulative spending at 90 days and annually through 7 years of follow-up. The analytic cohort comprised 6804 EVR patients (median follow-up: 1.85 years; interquartile range: 0.82–3.22 years) and 1889 open repair patients (median follow-up: 2.62 years; interquartile range: 1.13–4.80 years). Spending on index surgery was significantly lower for EVR (median [interquartile range]: $25 924 [$22 280–$32 556] EVR versus $31 442 [$24 669–$40 419] open; P<0.001), driven by a lower rate of in-hospital complications (6.6% EVR versus 38.0% open; P<0.001). EVR patients underwent more surveillance imaging (1.8 studies per person-year EVR versus 0.7 studies per person-year open; P<0.001) and AAA-related reinterventions (4.0 per 100 person-years EVR versus 2.1 per 100 person-years open; P=0.041). Open repair patients had higher rates of 90-day readmission (12.9% EVR versus 17.8% open; P<0.001) and abdominal wall procedures (0.6 per 100 person-years EVR versus 1.5 per 100 person-years open; P<0.001). Overall, EVR patients incurred more disease-related spending in follow-up ($7355 EVR versus $2706 open through 5 years). There was no cumulative difference in disease-related spending between surgical groups by 5 years of follow-up (−$33 EVR [95% CI: −$1543 to $1476]). Conclusions: We observed no cumulative difference in disease-related spending on EVR and open repair patients 5 years after surgery. Generalized recommendations about which approach to offer elective AAA patients should not be based on relative cost.

T he advent and widespread adoption of endovascular repair (EVR) for abdominal aortic aneurysms (AAA) represent a paradigm shift in the armamentarium of vascular surgeons. Between 2000 and 2010 in the United States, the proportion of all repairs performed by an endovascular approach increased from 5% to 74%. 1 Endovascular techniques and their related devices are expensive. In the short term, this cost is offset by shorter postoperative hospitalizations and, most importantly, decreased morbidity and mortality for patients exposed to a less invasive procedure. 2 In the long-term, this trade-off becomes more complex. EVR can be complicated by endoleak requiring reintervention, with reported long-term rates of reintervention as high as 30%. 3,4 Society for Vascular Surgery guidelines also recommend annual surveillance imaging for these patients as compared to 5-year interval surveillance for open AAA repair patients, which may drive higher utilization of healthcare resources. 5,6 A robust body of literature has evolved to assess the short-and long-term outcomes of these alternative AAA interventions. Missing from most of these analyses, however, is an assessment of the cumulative spending on care for patients with different clinical courses following endovascular or open surgery. An estimated 30 000 to 40 000 AAA repair procedures incur ≈$1 billion in healthcare expenditure each year in the United States. 1 AAA repair patients often require significant healthcare resources following surgery as well. Cost-effectiveness analyses have compared EVR and open AAA repair, but these rely on assumptions about the probability and cost of postoperative events and have often used clinical trial data which may not accurately reflect real-world practice. [7][8][9][10] A recent characterization of real-world costs of EVR and open repair only considered the index operation. 11 Other efforts to measure the downstream costs of open AAA repair and EVR have come from single institutions. 12 In a national economic and healthcare reform context that prioritizes managing the rapidly growing costs of care, a better understanding of longitudinal spending on AAA repair patients is needed.
The aim of this study was to compare spending by Medicare on care for patients receiving elective EVR or open AAA repair. We used claims data to evaluate per-capita payments made by Medicare for disease-related care from index surgery through last follow-up. We leveraged data from the Vascular Quality Initiative (VQI) encompassing anatomic, procedural, and other clinical considerations to calculate adjusted differences in disease-related and total spending by Medicare on AAA repair recipients through 7 years of follow-up. This comparison allows for clinically relevant estimates of healthcare utilization and spending on AAA surgical patients that are consequential for health systems, providers, and payors looking to ensure highvalue care for their populations.

Data Sources
The VQI was launched by the Society for Vascular Surgery to improve the quality, safety, effectiveness, and cost of vascular health care. 13 Over 500 centers in the United States and Canada contribute validated data to this registry (https://www.vqi.org/). The VQI data set was used to identify patients receiving EVR or open AAA surgery between January 1, 2003, and September 30, 2015, and associated information on preoperative risk factors, procedural characteristics, and postoperative events.
These patients were linked to Medicare claims using an established social security number to beneficiary identifier crosswalk. 14 Use of these claims was pursuant to a data use agreement with the Centers for Medicare and Medicaid Services, and Institutional Review Board approval was obtained from the Dartmouth College Committee for Protection of Human Subjects with waiver of informed consent. Any data points generated from <11 beneficiaries were suppressed from data presentation per Centers for Medicare and Medicaid Services requirements. Because of the sensitive nature of the data collected for this study, requests to access a similar data set from qualified researchers trained in human subject confidentiality protocols must be made through the Centers for Medicare and Medicaid Services.

Study Population
The cohort selection process is outlined in Figure 1 procedures identified in the VQI to Medicare claims using a 100% sample of Medicare beneficiaries, International Classification of Diseases, Ninth Edition (ICD-9) procedure codes, and Medicare Provider Analysis and Review claims for inpatient hospital services. 14 Patients were followed from the date of their index operation to the end of the study period or death and were required to be enrolled in Medicare part A and B and exclusively fee-for-service throughout the study. To facilitate the comparison of similar patient populations, we considered only patients having elective surgery as categorized in the VQI. We also excluded EVR patients deemed unfit for open AAA repair by their surgeon. Finally, given the weight of the index operation on cumulative spending, we excluded patients whose index payment was missing or $0.

Patient and Procedure Characteristics at Index Surgery
Demographic covariates obtained from Medicare claims were age, sex, race/ethnicity, and dual-eligible status (defined as those fully eligible for Medicare and Medicaid in the year of surgery). Year of surgery was taken from the VQI. Preoperative clinical characteristics obtained from VQI data were smoking status, body mass index, coronary artery disease, diabetes mellitus or chronic obstructive pulmonary disease on therapy, congestive heart failure, and active dialysis. Additional variables obtained from the VQI were the center at which index surgery occurred, urgency of the operation (elective, symptomatic, ruptured), presence of concomitant iliac aneurysm, maximum AAA diameter, postoperative length of stay, and the occurrence of any postoperative complication: access site occlusion (EVR only), surgical site infection, hematoma, stroke, myocardial ischemia or infarction, cardiac dysrhythmia, congestive heart failure, respiratory failure, lower extremity embolization, new dialysis requirement, intestinal ischemia, and return to the operation room. Medicare uses diagnosis-related groups (DRGs) as the basis for reimbursing most inpatient care, and we extracted the DRG assigned at index surgery from Medicare claims.
These designations changed in 2008 for both EVR and open AAA repair, from 110 and 111 (major vessel operation except heart, with and without major complication or comorbidity, respectively) to 237 and 238 (major cardiovascular procedures, with and without major complication or comorbidity, respectively).

Outcomes
Spending on index surgery was calculated by summing Medicare's total hospital payment, professional service payment, and pass-through payment (temporary add-on fees for innovative technologies). Reintervention events were flagged using ICD-9-CM procedure codes; we categorized reinterventions as disease-related or other interventions. Diseaserelated reinterventions were composed of AAA-related procedures and abdominal wall procedures, which included procedures for incisional hernia repair and adhesiolysis. Other interventions were categorized as vascular or nonvascular. Reintervention payments were determined from the related inpatient hospital and professional services payment amounts for the reintervention claim record. Post-procedure imaging was defined by current procedural terminology codes using outpatient and part B/carrier line item files, and we considered only the line payment associated with the image claim. Other inpatient spending was measured using all claims that were not otherwise flagged as disease-related or other vascular/nonvascular procedures. All-cause readmissions occurring within 90 days of discharge were considered disease-related events.
The disease-related and total spending outcomes are summarized in Figure 2, and code lists used for event detection are provided in Tables A through G in the Data Supplement. Final code lists were the result of blinded categorization of appropriate codes by 2 clinicians based on lists used in prior work, and discrepancies were adjudicated by the senior author. 15 All payment amounts were adjusted to 2015 US dollars using the Bureau of Labor Statistics Consumer Price Index for Medical Care Services. 16

Statistical Analyses
We compared preoperative clinical and procedural characteristics of patients undergoing EVR versus open AAA repair using appropriate univariate tests. We included total inpatient spending in the 12 months before index surgery as a baseline spending variable. The rate of disease-related events was compared using unadjusted Poisson regression. We performed crude comparison of median differences in spending for index surgery, individual follow-up events, and cumulative spending at different follow-up intervals (90 days post discharge then annually through 7 years) using quantile regression. Finally, as described in other work examining longitudinal healthcare costs, we calculated the adjusted difference in cumulative AAA-related and total spending at these intervals using multivariable quantile regression, incorporating variables that were different (P<0.10) at index surgery. 17 The final model included clustering at the level of VQI center where the index operation occurred. A sensitivity analysis incorporating only patients who had surgery from 2011 to 2015 was also performed. All statistical analysis was performed using Stata/MP 15.1 (StataCorp LLC, College Station, TX).

Cohort and Index Surgery Characteristics
The analytic cohort comprised 6804 EVR patients (median follow-up: 1.85 years; interquartile range: 0.82-3. The preoperative and procedural characteristics of patients at index AAA repair are summarized in Table 1  for open repair with major complication or comorbidity (DRG 237/110). Furthermore, 46.1% of open repair patients were assigned the more complex DRG, compared with only 8.4% of EVR (P<0.001). About 5% of both EVR and open surgery patients were assigned a DRG not typically used for AAA repair. Spending on index surgery by approach and DRG assignment is summarized in Table 2.

Disease-Related Reinterventions
The rate of all-cause readmissions within 90 days of discharge was 12 CT scans ($254) and magnetic resonance imaging ($294) were more expensive.

Non-Disease-Related Spending
AAA surgery recipients underwent other vascular and nonvascular procedures during follow-up categorized as unrelated to their index operation ( Table 3). The rate of non-AAA related vascular operations was slightly lower for EVR patients (1. Inpatient admissions >90 days after discharge from the index operation were common. The rate of admissions, not including those for disease-related reinterventions or other vascular and nonvascular procedures described above, was 40.3 per 100 person-years for EVR patients and 38.9 per 100 person-years for open repair patients (P<0.001). Median spending on these admissions was roughly $9000 in both groups.

Cumulative Disease-Related Spending
Longitudinal analysis of spending after index AAA repair is shown in Figure 3A. Spending in the 90 days following surgery was similar between groups ($1387 EVR versus $1174 open). Median disease-related spending was greatest in the third year following open repair, at $621. By comparison, disease-related spending for EVR recipients was between $1400 and $1500 in postoperative years 3 to 5. In cumulative terms, disease-related followup spending at 5 years was $7355 for EVR patients compared with $2706 for open repair patients ( Figure 3B).  Figure 4B). A sensitivity analysis among only those patients having surgery between 2011 and 2015 demonstrated a similar trajectory, with significantly less spending on EVR patients through 3 years but no significant difference between the procedural cohorts at 4 to 5 years of follow-up.

DISCUSSION
We performed a longitudinal analysis of spending on endovascular and open AAA repair patients using Medicare claims linked to the VQI clinical registry. Index operations comprised the vast majority of spending for most AAA repair patients. Driven by a higher rate of postoperative complications, average spending on index open AAA repair was ≈$5000 more per patient than for EVR. EVR patients subsequently received more frequent and expensive surveillance imaging and reinterventions and accrued roughly $5000 in incremental disease-related spending by 5 years following surgery. After adjusting for patient factors and center-level effects, there was no difference in cumulative spending between the procedural cohorts at 5 to 7 years of follow-up. Given dramatic reductions in short-term morbidity and mortality as compared to open surgery, EVR quickly became the predominant method for AAA repair in the United States. 1 However, the high cost of endografts, potential need for reintervention, and studies demonstrating equivalent or superior long-term mortality with open repair, has led to cost-effectiveness concerns with an endovascularfirst approach. 4,18 Highlighting this issue, 2018 draft guide-lines from the UK National Institute for Health and Care Excellence outlined a recommendation that patients with unruptured infrarenal AAA should not be offered EVR if open surgical repair is suitable. 19 This recommendation was based, in part, on a statement that EVR entails higher net costs than open surgical repair. Final National Institute for Health and Care Excellence guidelines have yet to be submitted; clearly, there is a need to better understand both the long-term outcomes and costs of care for AAA surgery patients. Our analysis furthers this effort and has important implications for clinicians, health services researchers, device manufacturers, policymakers, and patients.
First, this analysis of Medicare data allows for longterm characterization of health care utilization and expenditure following AAA repair. We differentiated between AAA-related reinterventions, which were more common among EVR recipients, and abdominal wall procedures, which were more common among open repair patients. Our reported rate of abdominal wall operations is consistent with a prior study of the Medicare population, which reported rates of incisional hernia repair of 1.2 per 100 person-years for open repair and 0.3 per 100 person-years for EVR. 20 The rate of repeat AAA repair was lower in that study (1.0 per 100 person-years EVR versus 1.2 per 100 person-years open), reflecting a narrower code list used for event detection. The cost attributed to disease-related reinterventions by studies to date has ranged from $7000 to $17 000 and often were not stated explicitly. 7 These estimates may grossly underestimate the financial impact of AAA-related reinterventions in particular, which was comparable to index surgery in our analysis.
Incremental spending on abdominal wall procedures for open repair patients was nearly equivalent to incremental spending on AAA-related reinterventions for EVR patients. This supports surveillance imaging as a key driver of higher disease-related spending for EVR patients in follow-up. Current guidelines from the Society for Vascular Surgery recommend a CT at 1 month and, if normal, support a transition to ultrasound for annual surveillance after EVR; CT at 5-year intervals is recommended following open repair. 5 We found EVR patients received surveillance imaging at more than twice the rate of open repair patients and underwent a slightly greater proportion of CT scans. At the same time, compliance with surveillance imaging is an established concern in the EVR population, with 20% to 40% of EVR patients in our analysis having no surveillance study in a given 12-month interval. 21 Suboptimal surveillance in EVR patients, and the perpetual nature of this imaging in current guidelines, ensures that surveillance practices will remain a focus in the debate over the longterm cost of care for AAA surgery patients.
Our study was not designed to address the challenge of quantifying the utility of changing health states that accompany the varied natural histories following AAA repair. Review of prior cost-effectiveness analyses compar-  18 The true cost-effectiveness of EVR likely rests with individual willingness to pay thresholds and the relative utility of the short-term risks of open repair weighed against the follow-up requirements and longterm risks of EVR. These real tradeoffs should prompt providers to explore ways to align surgical decisionmaking with individual patient preferences. A randomized clinical trial evaluating the utility of a decision aid for patients facing AAA surgery is underway. 24 Medicare claims are a critical tool for exploring utilization and spending in the AAA surgical population given that the vast majority of these patients are Medicare beneficiaries. 25 This is especially timely in the United States amidst mounting calls for single-payer or Medicare for All healthcare reform. Medicare spending is not the same as cost, but DRG-based reimbursement-used by Medicare for most reimbursement of inpatient care-is designed to reflect cost. 26 In this study, Medicare spending can be interpreted as a surrogate for cost from a societal payer perspective. A comparison to direct cost estimates in other studies is also useful from a provider perspective. Gupta    into cases with or without major complication or comorbidity. As a result, differences in per patient spending by Medicare are functionally driven by DRG code assignment for the index hospital stay. In our analysis, only one in 10 index EVR procedures were billed with MCC, compared to half of open procedures. For provider systems, potential coding inaccuracy represents an important focus in ensuring needed services remain financially viable. 27 Many also felt the decision to lump EVR in with open repair DRG groups was untenable due, in large part, to high graft device costs, and Centers for Medicare and Medicaid Services introduced new DRGs specific to EVR in 2016. 28 These changes are estimated to increase reimbursement for EVR with and without MCC by 24% and 14%, respectively. 29 Evolving device technologies, perioperative care strategies and reimbursement schemes will perpetuate uncertainty about the most cost-effective approach to elective AAA repair. Using Medicare claims to evaluate and compare healthcare utilization and spending was subject to limitations. We relied on administrative coding for event detection. Coding error and heterogeneity make it difficult to attribute disease-related healthcare spending with both sensitivity and specificity. For example, our use of ICD-9 codes for detecting index and reintervention events, based on a previously developed VQI-Medicare matching algorithm, may be more sensitive but less specific than using current procedural terminology codes assigned by the surgeon. Most EVR procedures occurred in the final 3 years of our study period, leading to significantly higher censorship in the EVR cohort. As a result, longer-term data in the EVR cohort may not reflect contemporary outcomes with this approach. Sample size constraints limited our analysis to 7 years of follow-up, so our findings may not be generalizable to young, non-Medicare patients with longer life expectancies. With the exception of surveillance imaging, we did not evaluate spending beyond the inpatient setting. In the Veterans Affairs cost-effectiveness study, outpatient visits accounted for only 2% of AAA-related costs but nearly 40% of non-AAA related costs. 9 Our primary spending outcomes were confounded by center-level adjustments to DRG base payments based on regional price, teaching hospital status, and share of low-income patients. The difference in spending we observed between EVR and open index surgery within a given DRG suggests that EVR is performed disproportionately in more urban settings and teaching hospitals, where adjusted reimbursement tends to be higher. To mitigate this effect, we performed clustering at the level of index surgical center in our final model of incremental spending.

CONCLUSIONS
Medicare's increased spending on index open AAA repair is offset by higher disease-related spending on EVR patients during follow-up. If long-term cost and outcomes of EVR and open repair remain in relative equipoise, maximizing value of AAA surgical care may best be accomplished at the patient level by tailoring selection of surgical approach to individual preferences.