Cost Analysis of Endovascular versus Open Repair in the Treatment of Thoracic Aortic Aneurysms

Abstract

Objective

For descending thoracic aortic aneurysms (TAAs), it is generally considered that endovascular stents (TEVARs) reduce operative morbidity and mortality compared to open surgical repair. However, long-term differences in patient survival have not been demonstrated, and an increased need for aortic reintervention has been observed. Many assume that TEVAR becomes less cost effective through time due to higher rates of reintervention and surveillance imaging. This study investigated mid-term outcomes and hospital costs of TEVAR compared with open TAA repair.

Methods

This was a retrospective, single institution review of elective thoracic aortic aneurysm repairs between 2005 and 2012. Patient demographics, operative outcomes, reintervention rates, and hospital costs were assessed. The literature was also reviewed to determine commonly observed complication and reintervention rates for TEVAR and open repair. Monte Carlo simulation was utilized to model and forecast hospital costs for TEVAR and open TAA repair up to 3 years post-intervention.

Results

Our cohort consisted of 131 TEVARs and 27 open repairs. TEVAR patients were significantly older (67.2 vs. 58.7, p=0.02) and trended towards a more severe comorbidity profile. Operative mortality for TEVAR and open repair was 5.3% and 3.7%, respectively (p=1.0). There was a trend towards more complications in the TEVAR group, although not statistically significant (all p>0.05). In-hospital costs were significantly greater in the TEVAR group ($52,008 vs. $37,172, p=0.001). However, cost modeling utilizing reported complication and reintervention rates from the literature overlaid with our cost data produced a higher cost for the open group in-hospital ($55,109 vs. $48,006) and at 3 years ($58,426 vs. $52,825). Interestingly, TEVAR hospital costs, not reintervention rates, were the most significant driver of cost in the TEVAR group.

Conclusions

Our institutional data showed a trend toward lower mortality and complication rates with open TAA repair, with significantly lower costs within this cohort compared to TEVAR. These findings were likely at least in part due to the milder comorbidity profile within these patients. In contrast, cost modeling using Monte Carlo simulation demonstrated lower costs with TEVAR compared to open repair at all time points up to 3 years post-intervention. Our institutional data shows that with appropriate patient selection, open repair can be performed safely with low complication rates comparable to TEVAR. The cost model argues that despite the costs associated with more frequent surveillance imaging and reinterventions, TEVAR remains the more cost effective option even years after TAA repair.

Introduction

With the advent of less invasive approaches, the management and treatment of thoracic aortic aneurysms (TAAs) has undergone significant changes in recent years. There has been a movement away from the traditional open repair in favor of less invasive endovascular techniques, collectively known as TEVARs (thoracic endovascular aortic repairs) (1,2). Early data has demonstrated lower in-hospital complication and mortality rates in TEVAR compared to open repair (1,3-8). Therefore, TEVAR has become the first line choice for TAA treatment by many vascular surgeons.

Despite these advantages in the short-term, the long-term superiority of TEVAR has not been demonstrated. In most studies, the 1-year survival for TEVAR is not significantly higher than open repair (1,3,4,8-11). Additionally, TEVARs have higher reintervention rates, predominantly due to the development of endoleaks, a complication rarely seen in open repairs (6,9,10,12,13). The current medical and financial climate in health care necessitates attention to the cost effectiveness of our treatment practices in addition to standard postoperative outcomes. Several studies have demonstrated that TEVARs typically have in-hospital costs that are similar to or lower than open repair (3,6,7,10,11,14). However, many individuals presume that TEVARs generate higher costs through time due to more frequent reinterventions and more frequent surveillance chest CT scans (6,14). To date, there have only been small single-center studies to evaluate the relative costs of the respective TAA repair strategies over time.

Therefore, the purpose of this study is to investigate the cost effectiveness of TEVAR versus open repair for elective TAA repair, utilizing Monte Carlo simulation to model and forecasts hospital costs up to 3 years post-intervention. We hypothesized that TEVAR will become more expensive through time due to higher rates of reintervention and increased surveillance imaging compared to open repair.

Methods

Patient Selection and Data Acquisition

The Institutional Review Board at the University of Virginia (#16496) approved this study. Because of its retrospective design with de-identification of patient data, consent was waived. A retrospective review was performed of all elective repairs of isolated thoracic aortic aneurysms at our institution between March 2005 and July 2012. TAA repair patients were stratified into 2 primary cohorts: open repair and endovascular repair (TEVAR). Exclusion criteria included: emergent procedures (<24 hours from an unplanned admission), concomitant repair of the abdominal aorta (thoracoabdominal aneurysms), acute traumatic injury, repair for dissection without aneurysmal component (defined as aortic diameter > 4.5cm), repair for penetrating ulcers, planned elephant trunk procedures, and laparotomy for direct aortic cannulation for stent delivery and deployment. Aneurysms with concomitant aortic dissections were included, as were repairs for aortic pseudoaneurysms. Of note, TEVARs that required subsequent open repair during the same hospitalization were categorized within the endovascular repair group (intent-to-treat analysis).

Patients between the cohorts were compared on demographics, preoperative comorbidities, operative outcomes, complication rates, and occurrences of readmission and reintervention.

Cost Data Acquisition

Hospital cost data were obtained for each patient’s index hospitalization as well as subsequent readmissions that were related to complications of the aortic intervention from hospital inpatient discharge financial data. Of note, this was the cost of care as estimated by the institution, not the charges relayed to patients and insurers and not hospital reimbursement figures. The average cost of an index hospitalization without complications both for TEVAR and for open repair was determined, as well as the average marginal increase in hospital costs with various complications (e.g., paralysis, stroke, myocardial infarction, major bleeding event, respiratory complication). Major bleeding event was defined as hemorrhage requiring operative reintervention or bleeding within the mediastinum. Respiratory complications were defined as a mechanical ventilation requirement longer than 48 hours, reintubation, or discharge with a new home oxygen requirement. The average hospital costs of readmission with an open aortic intervention, readmission with an endovascular intervention, and readmission with no intervention were also determined.

Outcomes and Cost Modeling

The primary outcome of interest was the forecasted differences in hospital costs up to 3 years post-intervention as determined by Monte Carlo Simulation (described below). The forecasted model was utilized to create a more precise estimate of hospital costs that could be applied to other institutions, depending on their local experience and complication rates with TEVAR and open TAA repair, as well as to highlight the primary cost drivers in each cohort.

Monte Carlo Simulation was utilized to model and forecast hospital costs for TEVAR and open TAA repair while in-hospital, as well as at 1 year and 3 years post-intervention. Complication and reintervention rates were determined by the weighted averages and ranges of rates observed in the literature, as this was more likely to be representative of the complication and reintervention rates at other institutions, rather than using the observed rates within our single-institution cohort (1,3-22) [Supplemental Table]. Triangular probability distributions (utilizing a minimum, most likely, and maximum value) and lognormal distributions, rather than a normal probability distribution, were used for the majority of complication and reintervention rates given the relatively small sample of rates reported in the literature. Cost values for index hospitalizations as well as marginal cost increases associated with various complications and reinterventions were determined using our institutional cost data as described above. Patients were categorized based on their initial procedure (TEVAR vs open) but were able to undergo subsequent interventions from either procedure type within the model. The model was designed as a sequential cascade with multiple decision points in which potential rates of complications, rates of reinterventions, as well as unvaried costs such as surveillance imaging were utilized. The primary decision points were 1) during the in-hospital window, with variability in rates of complications between the TEVARs and open repairs, 2) at 1 year post-intervention, with variability in rates of readmission and reintervention between the two groups, and 3) at 3 years post-intervention, with similar variability in the cumulative rates of reinterventions. The values and sequential time points ultimately utilized to construct our model are presented in Table I of the results section.

Table I.

Cost Model Variables

	TEVAR		Open
	%	cost ($)	%	cost ($)
In-hospital   No Complications   Paralysis   Stroke   Myocardial Infarction   Major Bleeding Event   Respiratory Complication	82.5 2 3 2 0.5 10	42,110 + 110,498 +11,780 + 43,956 + 178,143 + 17,200	60 4 5 5 3 23	38,612 + 110,498 + 11,780 + 43,956 + 178,143 + 17,200
1 year post-intervention   Surveillance CT scans ($180)   No Readmission   Angiographic Reintervention   Open Reintervention   Readmission + No Intervention	× 3 82.2 6.8 4 7	540 from In-hospital + 24,643 + 41,114 + 24,460*	× 1 90 1.5 1.5 7	180 from In-hospital + 24,643 + 41,114 + 8,075*
3 years post-intervention   Surveillance CT scans ($180)   No Readmission   Angiographic Reintervention   Open Reintervention   Readmission + No Intervention	× 5 79.2 8.8 5 7	900 from In-hospital + 24,643 + 41,114 + 24,460*	× 1 88.5 2.25 2.25 7	180 from In-hospital + 24,643 + 41,114 + 8,075*

^*Open patients that were readmitted without an intervention demonstrated more significant and more costly reasons for admission (eg. wound infection, epidural abscess) compared to TEVAR patients, and therefore a higher cost for readmission was used in the model.

Marginal costs with each complication and reintervention were determined based on our institutional cost data and discounted to Year 0 dollars using a discount rate of 10%. Given the low complication rates within our institution’s open cohort, cost data for each complication from both TEVARs and open repairs were pooled to determine the most likely marginal cost for each complication. Similarly, reintervention cost data from both TEVARs and open repairs were pooled for the simulation.

Provided there were no complications, cost of surveillance imaging in the TEVAR cohort was added to our model at 1 month, 6 months, 12 months, and then every 12 months post-intervention. In open repairs, surveillance imaging was performed at 12 months post-intervention provided there were no complications. The forecasted costs from our simulations are reported with a mean, median, and 25^th and 75^th percentiles based on 100,000 iterations.

To determine the primary drivers of cost in both the TEVAR and open repair cohorts, a sensitivity analysis was performed. Each component utilized by our cost model was a potential driver of cost within the sensitivity analysis, including the rate of no intervention in TEVAR, cost of no intervention in TEVAR, rate of paralysis in TEVAR, rate of angiographic intervention in TEVAR, cost of angiographic intervention in TEVAR at 1 year, etc. Each potential driver of cost was then individually varied within a range of −10% to +10% of the mean value assumed by the model. Forecasts were produced at each end of this range, and drivers were ranked according to estimated change produced in the forecast. Therefore, the drivers that generated the largest effect on cost output when varied within the model were deemed the strongest drivers of hospital cost.

Statistical Analysis

Dichotomous variables were evaluated using the Chi-squared or Fisher’s exact test as appropriate and continuous variables were compared using the Student’s t-test. All data analysis was performed in Microsoft Excel (Redmond, WA), and the Monte Carlo Simulation was performed using Oracle Crystal Ball (Redwood Shores, CA).

Results

Comparison of Patient Risk Factors and Outcomes

Between March 2005 and July 2012, 131 TEVARs and 27 open TAA repairs were performed at our institution. Of note, three of the TEVAR patients required open repair during the same hospitalization. There was a trend towards higher incidence of comorbidities in the TEVAR group, particularly with coronary artery disease [Table II]. However, patient age was the only measure that was significantly different between the two cohorts, with TEVAR patients being significantly older (67.2 vs. 58.7 years old, p=0.02).

Table II.

Preoperative Patient Characteristics

Variable	TEVAR   (n = 131 )		Open   (n = 27)		p value
Age [years]	67.22		58.74		0.02
	%	n	%	n
Male gender Hypertension Coronary Artery Disease Stroke Diabetes Chronic Renal Insufficiency Dialysis-dependent Chronic Obstructive Pulmonary Disease	56.5% 84.7% 47.3% 17.6% 7.6% 20.6% 2.3% 26.7%	74 111 62 23 10 27 3 35	51.9% 88.9% 25.9% 11.1% 0.0% 18.5% 3.7% 18.5%	14 24 7 3 0 5 1 5	0.68 0.77 0.05 0.26 0.21 1.00 0.53 0.47

Postoperative outcomes were similar between the two groups. There was no significant difference in hospital length of stay (LOS), intensive care unit (ICU) LOS, discharge to home, 30-day mortality, or 1-year mortality [Table III]. However, when evaluating the subgroup of patients with no complications, the hospital LOS in TEVARs was significantly shorter than open repairs (5.5 vs. 7.3 days, p=0.03). There was a trend toward more complications and more reinterventions in the TEVAR group, although these differences were not statistically significant. Average length of follow-up documented within the medical record was similar among TEVARs and open repairs (24.6 vs. 28.1 months, p=0.55).

Table III.

Postoperative Outcomes, Complications, and Costs

Variable	TEVAR (n = 131 )		Open (n = 27)		p value
	mean	SD	mean	SD
LOS [days] ICU LOS [hours] LOS (no complications) [days] ICU LOS (no complications) [hours]	7.6 120.5 5.5 76.8	9.3 198.0 3.8 61.2	7.3 98.7 7.3 97.5	3.3 58.7 3.4 60.6	0.76 0.30 0.03 0.14
	%	n	%	n
Stroke Paralysis Major Bleeding Event Myocardial Infarction Respiratory Complication Any Complication Discharged to Hom 30-day mortality 1-year mortality Reinterventions at 1 year Reintervention at 3 years	5.30% 2.30% 2.30% 3.10% 14.50% 23.70% 79.40% 5.30% 13.00% 13.70% 17.60%	7 3 3 4 19 31 104 7 17 18 23	0.00% 0.00% 0.00% 3.70% 11.10% 14.80% 77.80% 3.70% 14.80% 3.70% 11.10%	0 0 0 1 3 4 21 1 4 1 3	0.60 1.00 1.00 1.00 1.00 0.45 0.80 1.00 0.76 0.20 0.57
	mean	SD	mean	SD
In-Hospital Cost Hospital Cost at 3 years	$52,008 $59,709	$40,256 $45,984	$37,172 $43,787	$14,483 $20,062	0.001 0.005

LOS = length of stay, SD = standard deviation

Comparison of Institutional Hospitalization Costs

In analyzing hospital costs during the admission of the index procedure, TEVAR patients had significantly higher costs compared to patients undergoing open TAA repairs ($52,008 vs. $37,172, p=0.001, Table III). However, among the subgroup of patients with no complications, there was no significant difference between TEVAR and open repair ($42,110 vs. $38,612, p=0.22). Of note, the average reimbursement for uncomplicated TAA repair was $43,468 for TEVAR (DRG 39.73) and $40,871 for open repair (DRG 38.44 and 38.45). Paralysis and major bleeding event were the complications that most dramatically increased the cost of the index hospitalization, generating increases of $110,498 and $178,143, respectively. At 3 years post-intervention, TEVAR costs remained significantly higher than open repair costs ($59,709 vs. $43,787, p=0.005, Table III).

Comparison of Forecasted Hospitalization Costs

The results of our literature review on complication and reintervention rates for both TEVAR and open TAA repair are shown in Table IV. Of note, there are some distinct differences when compared to our institutional rates. For one, the complication rates for our institution’s open repair cohort are very low compared to the rates reported in the literature. Additionally, the literature consistently demonstrates a more favorable complication profile in TEVARs compared to open repairs, whereas our institution’s data demonstrates the opposite, with higher complication rates within our TEVAR cohort.

Table IV.

Postoperative Complications and Outcomes

Outcome	Our TEVAR (n=131) % (n)	TEVAR from the literature mean % [range]	Our Open (n=27) % (n)	Open from the literature mean [range]
Stroke Paralysis Major Bleeding Event Myocardial Infarction Respiratory Complication 30-day mortality Reinterventions at 1 year Reinterventions at 3 years	5.3 (7) 2.3 (3) 2.3 (3) 3.0 (4) 14.5 (19) 5.3 (7) 13.7 (18) 17.6 (23)	3 [0-9.5] 2 [0-5] 0.01 [0-0.01] 2.1 [2-2.3] 10 [4.3-16] 3 [0-8.1] 10.8 [4-20] 13.8 [6.6-26]	0 (0) 0 (0) 0 (0) 3.7 (1) 11.1 (3) 3.7 (1) 3.7 (1) 11.1 (3)	5 [2.1-10.3] 4 [1.5-14] 3 [1.4-6.5] 5 [4-6.3] 23 [10.4-44] 7 [2.3-20] 2.9 [0-7] 4.5 [0-10]

Table I displays the values utilized to construct our cost model. Again, the rates were determined from the literature and the marginal cost for each decision point was determined from our institutional cost data. The Monte Carlo Simulation forecasted the cost to the hospital of providing care for TEVAR patients and for open TAA repairs. Forecasts were generated for the index hospitalization, post-intervention year 1, and post-intervention year 3.

The output of the Monte Carlo Simulation predicted that the average cost for index hospitalization for TEVAR would be lower than for open repair ($48,006 vs. $55,109, Figure 1). When modeling costs at 1 year and 3 years post-intervention, the cost gap between the two groups narrowed slightly, but TEVAR remained the less expensive alternative at both time points ($51,885 vs. $57,901 at 1 year, $52,825 vs. $58,426 at 3 years). Of note, these lines never intersected during our 3-year forecast.

Figure 1.

Cost Forecast of TEVAR vs. Open TAA Repair over time. Each forecast reflects the mean surrounded by error bars reflecting the interquartile range.

The estimated cumulative cost of care incurred by the hospital over a 3-year period for each cohort, including a mean estimate (from Table I) around its relevant probability distribution, for all 100,000 iterations of our model are displayed in Figure 2. This figure displays a more detailed view of the output of the Monte Carlo Simulation. Ultimately, these cumulative costs reflect the sum of predicted costs for the index operation, hospitalization, post-operative surveillance, complications, reinterventions, and readmissions in each cohort. While the means and interquartile ranges differ between these two cohorts, there is adequate overlap of each cohort’s probability distribution, which helps account for how our single-institution data deviated from our multi-institutional forecasts.

Figure 2.

Forecasted Cumulative 3-year Costs of TEVAR vs. Open Repair. The x-axis reflects the forecasted cost ($) and the y-axis reflects probability (%). The mean, median, 25^th percentile, and 75^th percentile are indicated for each forecast. Of note, the interquartile range of each forecast does not overlap.

Our sensitivity analysis revealed that the primary cost drivers of TEVAR were (in order) in-hospital cost of an uncomplicated procedure, rate of respiratory complications, and rate of paralysis. Reintervention rate was much lower on the hierarchy. When looking at open repair, in-hospital cost of an uncomplicated procedure was also the strongest driver of cost, although less significant than in TEVARs. Complication rates and complication costs were stronger drivers of total cost in the open group, with rate of respiratory complications, cost of respiratory complications, rate of major bleeding event, and rate of paralysis being the strongest drivers.

Discussion

The present study provides a mid-term cost analysis of TAA repair strategies utilizing Monte Carlo Simulation to forecast cumulative hospital costs up to 3 years post-intervention. In the TAA repair cohort from our institution, there were more complications within the TEVARs compared to the open repairs, which is the inverse of the relationship typically observed in the literature. This finding is likely at least partly explained by the milder preoperative risk profile of our open cohort. There were slightly more complications in our TEVARs compared to TEVARs in the literature, but substantially fewer complications in our open repairs compared to the literature. Additionally, the longest length of stay in our open cohort was 17 days, demonstrating the mild complication profile of this group of patients. Because complications drive hospital costs, it was not surprising that TEVARs had significantly higher costs than open repairs within our cohort. However, this cost disparity was not present when evaluating patients who did not experience any complications, further reinforcing the critical relationship between complications and cost.

In the current health care market, hospitals are reimbursed for caring for more complex patients, even if the patient’s complexity is related to a complication from their procedure. Ideally a hospital would be able to care for a TAA patient without taking a loss. As we move to a reimbursement model in which hospitals are paid a flat rate for taking care of patients with a particular pathology, regardless of complications, maintaining low complication rates will be paramount to maintain financial solvency.

Our Monte Carlo Simulation (using complication rates from the literature and our institution’s cost data) predicted lower hospital costs in TEVARs during their index hospitalization as well as at 1 year and 3 years post-intervention. Considering the differences in our institution’s complication rates compared to the literature, the reversal of this relationship compared to our institution’s raw cost data is not surprising. This discrepancy further emphasizes the caution we should take when drawing conclusions from single institution data, as they may not reflect multi-institutional trends. However, cost modeling with Monte Carlo simulation is one method that may provide more accurate and generalizable information when comparing different treatment modalities, rather than looking at single institution data. Monte Carlo simulation is a well-known analytics technique commonly used in business that utilizes simulations run repetitively to generate probabilistic predictions of outcomes. The primary caveat for Monte Carlo simulation is that the strength and accuracy of the conclusions are only as strong as the assumptions made when defining the model.

When further interrogating our institution’s cost data, there were some interesting findings. The complication that generated the highest increase in hospital cost was major bleeding event. Surprisingly, all 3 major bleeding events were in our TEVAR cohort (2.3% of cases), with none in our open repairs. Alternatively, the increased cost associated with stroke in our data set was quite low. This can partially be attributed to two of our stroke patients dying in the hospital within a week of their intervention, rather than undergoing prolonged hospital stays with many procedures and ICU readmissions, which was the hospital course experienced by our patients with major bleeding events.

In general, TEVARs have higher rates of endovascular and open reinterventions compared to open TAA repairs. However, these rates are not dramatically higher than open reintervention rates (9,12,13,15-18,22). Additionally, the hospital cost of one surveillance chest CT scan is $180 at our institution, so even with frequent imaging in TEVAR patients, the cost of surveillance imaging is quite small compared to the costs for procedural interventions and complications. Consequently, even when forecasted at 3 years, our model predicted lower costs for TEVAR compared to open repair. Our sensitivity analysis further confirmed the relatively small contribution of reinterventions and imaging to total costs, demonstrating that the cost of an uncomplicated index procedure was the primary driver of total cost to the hospital in both TEVARs and open repairs. Therefore, the best way for a hospital to reduce the financial burden of these procedures is to work to reduce the index hospitalization costs, such as by reducing length of stay or negotiating for lower stent graft costs.

Most cost analyses of TAA repairs have only analyzed hospital costs during the index hospitalization. Some of these studies have shown TEVAR to be less expensive due to shorter hospital stays with lower complication rates (10,14), while others have shown no difference between open and endovascular repairs (3,7). One study has evaluated costs beyond the initial hospital stay. Karimi and colleagues evaluated their TAA cost data in a 57 patient, single center cohort for 2 years post-intervention (14). They found that in-hospital and at 2 years post-intervention, TEVAR was the more cost effective option. Our cost model corroborates these findings, presumably with greater validity due to the utilization of data from outside of our single institution.

Several conclusions can be drawn from this study that will impact how we care for TAA patients. Despite the general assumption that the cost of TEVARs over time outweighs the initial cost benefits, our model argues to the contrary. Because surveillance imaging and reinterventions are relatively low drivers of cost, TEVARs maintain an economic advantage in our model. Additionally, our institution’s data demonstrate that with appropriate patient selection, open TAA repair can be performed effectively with low complication rates and low hospital costs. Therefore, at medical centers with surgeons who are proficient at open aortic surgery, this treatment still remains a viable and cost-effective option.

There are select limitations of the present analysis. First, this is a retrospective review of our outcomes and cost data, which comes with the inherent limitations associated with this type of review. Second, follow-up data was inconsistent, as some patients had no documented visits after their hospitalization for their index procedure. Additionally, readmissions and reinterventions were only captured if patients returned to our institution, which likely is not the case 100% of the time. Third, the number of each type of complication within our dataset was relatively small (particularly within our open cohort); thus, our cost estimates associated with each complication is subject to stronger influence by outliers. Fourth, we performed our cost analysis from the perspective of the hospital. However, from the perspective of the entire healthcare delivery system, there are additional costs to the system for patients that experience strokes, paralysis, renal failure requiring dialysis, or require discharge to rehabilitation facilities or skilled nursing facilities (23). Finally, a higher percentage of our open cases were performed earlier in our study window, as TEVAR has become the preferred the first line treatment choice in recent years. Therefore, our open cost data may appear artificially low due to the effects of inflation through time. Because our institutional rates of cost inflation were not obtainable, we were unable normalize our cost data to “2012 dollars”, for instance.

In conclusion, cost modeling with Monte Carlo Simulation demonstrated lower costs for TAA repair using TEVAR compared to open repair at all time points up to 3 years post-intervention. Despite the widespread assumption that TEVAR becomes relatively more expensive through time due to higher reintervention rates and increased surveillance imaging, our model argues against this theory. Because the majority of reinterventions occur within the first 3 years of the index procedure, it is expected that the cost effective dominance of TEVAR will hold through time. While our institutional data suggests that open repair can be performed safely and cost-effectively with appropriate patient selection, TEVAR should remain the preferred option for TAA repair from the standpoint of both patient outcomes and cost effectiveness.