The aim of this study was to compare in-hospital cost and outcomes between transcatheter aortic valve implantation (TAVI) and surgical aortic valve replacement (SAVR). TAVI is an effective treatment option in patients with symptomatic aortic stenosis who are at high risk for traditional SAVR. Several studies using trial data or outside United States registry data have addressed TAVI cost issues, although there is a paucity of cost data involving commercial cases in the United States. Using Agency for Healthcare Research and Quality Healthcare Cost and Utilization Project Nationwide Inpatient Sample files, a propensity score–matched analysis of all commercial TAVI and SAVR cases performed in 2011 was conducted. Overall hospital cost and length of stay, as well as procedural complications, were compared between the 2 matched cohorts: 595 TAVI patients were matched to 1,785 SAVR patients in a 1:3 ratio. There was no difference in mean ($181,912 vs $196,298) or median ($152,993 vs $155,974) hospital cost between TAVI and SAVR (p = 0.60). The TAVI group had significantly shorter lengths of hospital stay than the SAVR group (mean 9.76 vs 12.01 days, p <0.001). There was no difference in postprocedural in-hospital death or stroke, but TAVI patients were more likely to have bleeding complications, to have vascular complications, and to require pacemakers. In conclusion, when analyzing in-hospital cost of commercial TAVI and SAVR cases using the Nationwide Inpatient Sample data set, TAVI is an economically satisfactory alternative to SAVR and results in an approximately 2-day shorter length of stay during the index hospitalization.
Transcatheter aortic valve implantation (TAVI) is an effective treatment for patients with symptomatic aortic valve stenosis who are at high risk or are ineligible for traditional surgical aortic valve replacement (SAVR). The US Food and Drug Administration approved the use of the Edwards SAPIEN transcatheter valve with a transfemoral (TF) approach in inoperable patients in November 2011. The Society of Thoracic Surgeons and American College of Cardiology Transcatheter Valve Therapy (TVT) registry reported that 8,075 commercial TAVI cases were performed in the United States in the 17 months after approval. As the indication for TAVI continues to broaden, many physicians have expressed opinions regarding cost given the increased penetration and use of this relatively expensive technology. Studies limited to trial data have addressed the cost-effectiveness of TAVI, with respect to SAVR (in high-risk operable patients) and standard therapy (in inoperable patients). Most published cost analyses involve Markov modeling using the Placement of Aortic Transcatheter Valve (PARTNERS) data. However, cost data from commercial cases performed in the United States are relatively limited. The 2011 Agency for Healthcare Research and Quality Healthcare Cost and Utilization Project Nationwide Inpatient Sample (NIS) contains the first available United States data containing inclusive cost information on TAVI admissions. Using this comprehensive source, we compared cost data between TAVI and SAVR in a propensity-matched model. We also compared in-hospital complication rates as defined by the NIS data set.
Methods
Data were obtained from the NIS files in 2011. The NIS is a 20% stratified sample of all nonfederal United States hospitals. The NIS contains deidentified information for 38,590,733 discharges from 1,049 hospitals and 46 states in 2011. Discharges are weighted on the basis of the sampling scheme to permit inferences for a nationally representative population. Each record in the NIS includes all procedure and diagnosis International Classification of Diseases codes recorded for each patient’s hospital discharge.
Hospitalizations identifying aortic valve replacement or implantation were selected using the respective International Classification of Diseases, Ninth Revision, Clinical Modification procedure codes for SAVR (35.21 and 35.22) and TF TAVI (35.05) in any of the 15 procedure fields. Patient-level and hospital-level variables were included as baseline characteristics. The Agency for Healthcare Research and Quality co-morbidity measures, based on the Elixhauser methods, were used to identify co-morbid conditions. The main end points analyzed were actual total in-hospital cost, defined in the NIS as total charges, and length of hospital stay (LOS). Additional in-hospital patient-level end points (postprocedural stroke, myocardial infarction, major bleeding, vascular complications, and permanent pacemaker [PPM] placement) were determined from International Classification of Diseases, Ninth Revision, billing codes in the NIS database.
We compared baseline characteristics between the TAVI and SAVR groups for the entire unmatched 2011 NIS cohort of patients. Given the likelihood of significant differences between these 2 groups, and to adjust for selection bias for TAVI versus SAVR, a propensity score was calculated using a multilevel model including all covariates used in the primary analysis. After PROC GLIMMIX estimation of propensity score, we used PROC LOGISTIC to analyze the matched data set created by PROC GLIMMIX. The discrimination of the propensity model for the receipt of TAVI was evaluated using the C statistic. Using previously described methods, each TAVI case was matched with 3 SAVR controls for statistical efficiency by predicted propensity scores, and the success of the propensity score–matched model was assessed by examining the standardized differences in baseline characteristics before and after matching.
Using the matched cohort, we used Wilcoxon’s 2-sample test to determine the difference in total hospital cost and in LOS between the TAVI and SAVR groups. Using Mantel-Haenszel statistics, we determined the relative risk for postprocedural complications between the 2 matched groups. We also analyzed the raw data with respect to risk factors and procedural complications in our matched TAVI group in an attempt to analyze how closely this matched group mimics trial and registry data. In addition, using Wilcoxon’s rank sum tests, we compared differences in cost according to hospital size, geographic region, and type (academic vs nonacademic). All statistical analyses were conducted using SAS version 9.2 (SAS Institute, Inc., Cary, North Carolina).
Results
Of 38,590,733 discharges in the 2011 NIS data set, 76,361 patients underwent SAVR and 887 underwent TAVI. TAVI patients were more likely to be older, to be treated at large academic centers, and to have a higher incidence of co-morbidities, including lung disease, anemia, renal failure, ischemic heart disease, and coronary revascularization. Of the 887 TAVI patients, 595 were suitable for propensity matching in a 1:3 ratio to 1,785 SAVR patients. The C statistic of our propensity model was 0.96. Table 1 lists baseline characteristics and standardized differences of the 2 groups before and after propensity matching, with most variables having absolute standardized differences <10. The most notable variables with standardized differences >10 were diabetes (higher prevalence in TAVI arm), age (older in the SAVR by a mean of 1 year), and presence of peripheral vascular disease (higher in the SAVR arm).
| Baseline Characteristics | Before PS Matching | After PS Matching | ||||||
|---|---|---|---|---|---|---|---|---|
| SAVR | TAVI | SD | p-value | SAVR | TAVI | SD | p-value | |
| Number | 76,361 | 887 | 1,785 | 595 | ||||
| Age (years) | 68.7 | 83.0 | 126.8 | <0.001 | 82.3 | 81.3 | -13.1 | 0.25 |
| Female | 37.9 % | 44.7 % | 13.8 | 0.06 | 47.3 % | 45.4 % | -3.8 | 0.75 |
| White race | 82.8 % | 82.5 % | 0.8 | 0.98 | 86.6 % | 81.5 % | -14.0 | 0.18 |
| Payer Status | <0.001 | 0.62 | ||||||
| Medicare | 65.6 % | 94.7 % | 78.4 | 96.1 % | 97.5 % | 8.0 | ||
| Medicaid | 4.1 % | 1.1 % | -18.9 | 0.6 % | 0.8 % | 2.4 | ||
| Private/HMO | 26.4 % | 3.7 % | -66.9 | 2.8 % | 1.7 % | -7.4 | ||
| Hospital Size | <0.001 | 0.38 | ||||||
| Small | 6.3 % | 0 % | -36.7 | 0 % | 0 % | 0 | ||
| Medium | 16.2 % | 6.3 % | -31.7 | 5.3 % | 7.6 % | 9.4 | ||
| Large | 77.6 % | 93.7 % | 47.2 | 94.7 % | 92.4 % | -9.4 | ||
| Region | <0.001 | 0.92 | ||||||
| Northeast | 25.3 % | 35.3 % | 21.9 | 42.9 % | 44.5 % | 3.2 | ||
| Midwest | 20.8 % | 21.6 % | 2.0 | 15.1 % | 16.0 % | 2.5 | ||
| South | 32.7 % | 33.7 % | 2.1 | 32.8 % | 29.4 % | -7.4 | ||
| West | 21.1 % | 9.5 % | -32.7 | 9.2 % | 10.1 % | 3.2 | ||
| Teaching | 70.6 % | 95.8 % | 71.6 | <0.001 | 96.4 % | 94.1 % | -10.8 | 0.30 |
| Anemia | 21.0 % | 34.7 % | 30.9 | <0.001 | 30.8 % | 31.9 % | 2.4 | 0.82 |
| Collagen Vascular Disease | 2.9 % | 3.7 % | 4.5 | 0.51 | 4.2 % | 1.7 % | -14.8 | 0.26 |
| Chronic Lung disorder | 20.6 % | 40.5 % | 44.2 | <0.001 | 38.1 % | 36.1 % | -4.1 | 0.74 |
| Coagulopathy | 30.5 % | 23.2 % | -16.5 | 0.03 | 21.9 % | 23.5 % | 3.8 | 0.70 |
| Diabetes mellitus, uncomplicated | 24.2 % | 27.4 % | 7.3 | 0.31 | 26.3 % | 31.1 % | 10.6 | 0.34 |
| Diabetes mellitus, complicated | 4.6 % | 3.7 % | -4.5 | 0.72 | 4.5 % | 4.2 % | -1.5 | 0.99 |
| Hypertension | 69.4 % | 83.2 % | 32.9 | <0.001 | 83.5 % | 81.5 % | -5.3 | 0.67 |
| Liver disease | 1.8 % | 5.3 % | 19.0 | 0.003 | 4.8 % | 5.9 % | 4.9 | 0.63 |
| Neurological disease | 4.6 % | 5.3 % | 3.2 | 0.60 | 5.3 % | 5.0 % | -1.4 | 0.99 |
| Obesity | 16.7 % | 12.1 % | -13.1 | 0.10 | 10.4 % | 12.6 % | 6.9 | 0.50 |
| Peripheral vascular disease | 20.3 % | 24.7 % | 10.6 | 0.15 | 24.9 % | 20.2 % | -11.3 | 0.32 |
| Chronic renal failure | 17.0 % | 43.7 % | 60.7 | <0.001 | 35.3 % | 37.8 % | 5.2 | 0.66 |
| Angina pectoris | 3.5 % | 3.2 % | -1.7 | 0.99 | 3.9 % | 1.7 % | -13.3 | 0.38 |
| Chronic ischemic heart disease | 55.3 % | 67.4 % | 25.0 | <0.001 | 65.3 % | 64.7 % | -1.3 | 0.91 |
| Hemodialysis | 1.2 % | 1.1 % | -0.9 | 0.99 | 0.8 % | 0.8 % | 0 | 0.99 |
| Prior CABG | 5.5 % | 25.8 % | 58.2 | <0.001 | 19.1 % | 21.9 % | 6.9 | 0.51 |
| Prior PCI | 8.0 % | 21.6 % | 39.0 | <0.001 | 16.5 % | 19.3 % | 7.3 | 0.49 |
Table 2 lists cost and LOS analysis of the propensity-matched groups. There was no difference in mean ($181,912 vs $196,298) or median ($152,993 vs $155,974) hospital cost between TAVI and SAVR, respectively (p = 0.60). The TAVI group did, however, have a significantly shorter LOS than the SAVR group (mean 9.76 vs 12.01 days, median 7 vs 9 days, respectively, p <0.001).
| SAVR | TAVI | p-value | |
|---|---|---|---|
| Median LOS (days) | 9.0 | 7.0 | <0.001 |
| Mean LOS and STD (days) | 12.0 ± 8.2 | 9.8 ± 7.5 | |
| Median Charge ($) | 155,974 | 152,992 | 0.60 |
| Mean Charge and STD ($) | 196,298 ± 155,033 | 181,912 ± 123,535 |
Subgroup analysis of costs of the propensity-matched groups according to hospital specifics (as defined by the NIS ) are listed in Table 3 . TAVI and SAVR costs were highest in western hospitals, nonacademic centers, and smaller institutions. However, there was no difference between TAVI and SAVR within each hospital subgroup, including geographic location, hospital size, and type (academic vs nonacademic).
| Hospital Specifics | Modality | n | Cost ($) | p-value | |
|---|---|---|---|---|---|
| Geographic Locale | Northeast | TAVI | 53 | 150,298 | 0.16 |
| SAVR | 153 | 157,677 | |||
| Midwest | TAVI | 19 | 139,138 | 0.99 | |
| SAVR | 54 | 134,318 | |||
| South | TAVI | 35 | 150,549 | 0.91 | |
| SAVR | 117 | 143,632 | |||
| West | TAVI | 12 | 281,555 | 0.58 | |
| SAVR | 33 | 278,191 | |||
| Type | Academic | TAVI | 112 | 150,549 | 0.52 |
| SAVR | 344 | 151,253 | |||
| Non-academic | TAVI | 7 | 213,469 | 0.69 | |
| SAVR | 13 | 255,595 | |||
| Size ∗ | Medium | TAVI | 9 | 222265 | 0.56 |
| SAVR | 19 | 249431 | |||
| Large | TAVI | 110 | 150298 | 0.52 | |
| SAVR | 338 | 149621 | |||
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
