Clinical outcomes of transcatheter aortic valve implantation (TAVI) have significantly improved with the accumulation of operator and institution experience as well as the wide use of newer generation devices. There is limited data on TAVI outcomes compared with surgical aortic valve replacement (SAVR) in contemporary practice in the United States. We queried the 2018 Nationwide Readmission Database of the United States. International Classification Diagnosis code 10 was used to extract TAVI and SAVR admissions. A propensity-matched cohort was created to compare TAVI and SAVR outcomes. A weighted 48,349 TAVI and 24,896 SAVR for aortic stenosis were included and 4.9% of TAVI were performed with an embolic protection device. In propensity-matched cohort (12,708 TAVI and 12,708 SAVR), TAVI conferred lower in-hospital mortality (1.7% vs 3.8%), acute kidney injury (11.3% vs 22.9%), and transfusion rate (5.9% vs. 20.6%) whereas new pacemaker rate was higher in TAVI compared with SAVR (10.5% vs. 7.0%) (all p values < 0.001). Stroke rate was similar between TAVI and SAVR (1.5% vs. 1.5%) (p value = 0.79). The routine discharge was more frequent (66.9% vs 25.8%) and length of stay was shorter (4.8 vs. 9.8 days) in TAVI than SAVR. Hospitalization cost was higher in SAVR than TAVI (51,962 vs 57,754 U.S. dollars) (all p values < 0.001). In-hospital mortality was also lower in TAVI compared with isolated SAVR. TAVI was performed more frequently than SAVR in 2018 in the United States with lower in-hospital mortality of TAVI compared with both SAVR and isolated SAVR.
Case volume of transcatheter aortic valve replacement (TAVI) has recently surpassed that of surgical aortic valve replacement (SAVR) in the United States (US). TAVI is considered less invasive and therefore could offer a perioperative outcome advantage over SAVR, however, previous studies have shown similar short-term mortality between TAVI and SAVR. This could be because that TAVI experiences were still developing. Recent large registry from the US showed that thirty-day mortality of TAVI has decreased significantly from 7.2% in 2011 to 2.5% in 2019 but it remains unknown whether TAVI confers improved short-term mortality over SAVR from large database especially in an era where case selections and managements of TAVI have become more matured. The purpose of our study was to compare TAVI and SAVR outcomes in 2018 from the Nationwide Readmission Database (NRD) of the United State.
Methods
The latest data from NRD 2018 were used from the Agency for Healthcare Research and Quality’s (AHRQ) Healthcare Cost and Utilization Project (HCUP). Details are provided in previous studies. Briefly, the NRD includes a large sample size, which provides sufficient data for analysis across hospital types and the study of readmissions for relatively uncommon disorders and procedures. Discharge data from 28 geographically dispersed states, accounting for 59.7% of the total US resident population and 58.7% of all US hospitalizations is available. Discharge weights are provided in the form of a variable ‘DISCWT’ to obtain national estimates. Unweighted, the NRD contains data from approximately 18 million discharges each year. Weighted, it estimates roughly 35 million discharges. This study was deemed exempt from the Institutional Review Board as the NRD is a publicly available database that contains de-identified patient information.
We used the ICD-PCS (Procedure Coding System) codes of 02RF3JZ, 02RF3KZ, 02RF38Z, and 02RF37Z, to identify all hospitalizations for TAVI and SAVR from 2018. We excluded patients <= 65 years of age, patients with a primary diagnosis of infective endocarditis, and who had both TAVI and SAVR performed during the same hospitalization, and those who underwent transapical TAVI. Figure 1 depicts the flowsheet for the selection including inclusion and exclusion criteria. Further to this we also compared TAVI with isolated SAVR after excluding coronary bypass graft surgery and surgeries on other valves. Because the NRD is a yearly database, all the patients who underwent TAVI in the first 11 months were included in the study so that we could track 30-day readmission outcomes. Readmissions within the 30-days were identified in survivors of the index admission using the ‘nrd_visitlink’ variable. Patients who got readmitted more than once within the designated time-period were counted once for their index readmission. Time to readmission was calculated as the number of days between hospital discharge after index TAVI procedure and the first day of hospital readmission.
For each of the above two cohorts , we extracted baseline patient and hospital characteristics . Patient characteristics included age, sex, race, median household income, and relevant comorbidities such as hypertension, diabetes mellitus, congestive heart failure, chronic lung disease, peripheral vascular disease, chronic kidney disease stage, prior myocardial infarction, prior percutaneous coronary intervention, prior coronary artery bypass grafting, previous valve surgery, prior pacemaker implantation, liver disease, coagulopathy, atrial fibrillation, and obesity. We also gathered data on elective admissions and compared them based on their baseline frailty status divided into low intermediate and high using a validated method for an administrative database . We used the well-validated methodology devised by Quan et al. by utilizing the coding algorithms with ICD-10 for defining the comorbidities. Additionally, we extracted the data on hospital characteristics such as location, teaching status, and bed size . The hospital was considered a teaching facility when it had an American Medical Association approved residency program. It was a member of the Council of Teaching Hospitals and Health Systems. It had full-time equivalent interns and residents to bed ratio of 0.25 or greater.
Our primary outcome of interest was all-cause in-hospital mortality . Other in-hospital outcomes of interest were acute kidney injury, need for transfusion, stroke, and new pacemaker implantation . Finally, we also investigated routine home discharge, hospital length-of-stay, and total hospital cost in US dollars for the indexed intervention. Total hospital charges (the amount hospitals billed for the stay) are reported in the core NRD file, although they do not reflect the actual cost of care. The HCUP provides cost-to-charge ratios filed based on all-payer inpatient costs. This cost information is obtained from the hospital accounting reports collected by the Centers for Medicare and Medicaid Services. Using this information, total hospital costs were calculated by multiplying total hospital charges with the corresponding cost-to-charge ratio.
NRD data design is based on a complex survey design that includes stratification, clustering, and weighting adjustment. We utilized weighting to produce nationally representative unbiased results, variance estimates, and p values. Baseline patient and hospital characteristics, in-hospital procedures, and complications were initially compared between an unmatched population using a test of independence based on the Pearson χ2 statistic. To account for survey design, the Pearson statistic was converted into an F-statistic with noninteger degrees of freedom by using a second-order Rao and Scott correction. Continuous variables were compared between different groups using a t-test, as appropriate. All p values were two-sided, with a conventional significance threshold of p value < 0.05. Categorical variables are expressed as percentages and continuous variables as mean±SD. A propensity score-matched analysis was performed to adjust for potential confounders (including all the variables mentioned in Table 1 . Propensity-score matching was performed in R statistical software using ‘nearest neighbor matching.’ Logistic regression was employed to estimate the distance measure. Matching was performed with a caliper set at 0. 1. Cases were matched with controls without replacement and with common support. All the comparison analyses were repeated in the matched cohort. Data were complete on all covariates except for cost. Missing values were replaced with the dominant category. This approach has been used in prior studies. Statistical analyses were performed using Stata 16.0 (StataCorp. 2019. Stata Statistical Software: Release 16. College Station, TX: StataCorp LLC.) and R (R Development Core Team, Vienna, Austria).
| Unadjusted | P-value | Propensity-matched | p value | |||
|---|---|---|---|---|---|---|
| Variable | TAVI (N = 48,349) | SAVR (N = 24,896) | TAVI (N = 12,708) | SAVR (N = 12,708) | ||
| Age mean (SD) (Years) | 81.0 (6.6) | 74.0 (5.2) | <0.001 | 76.0 (6.2) | 76.0 (5.2) | 0.88 |
| Men | 53.9% | 65.8% | <0.001 | 59.8% | 60.1% | 0.69 |
| Bicuspid aortic valve | 0.4% | 3.6% | <0.001 | 1.1% | 1.1% | 0.68 |
| Hypertension | 90.4% | 86.7% | <0.001 | 88.4% | 88.6% | 0.82 |
| Diabetes mellitus | 38.1% | 37.5% | 0.54 | 41.6% | 40.9% | 0.52 |
| Obesity | 19.2% | 27.1% | <0.001 | 26.6% | 26.1% | 0.67 |
| Prior percutaneous coronary intervention | 23.2% | 11.3% | <0.001 | 15.1% | 15.1% | 0.92 |
| Prior coronary bypass | 16.3% | 5.3% | <0.001 | 8.1% | 7.8% | 0.66 |
| Prior valve replacement | 2.3% | 3.6% | <0.001 | 3.6% | 3.8% | 0.63 |
| Prior myocardial infarction | 12.5% | 8.2% | <0.001 | 10.3% | 9.5% | 0.13 |
| Prior pacemaker | 10.1% | 3.8% | <0.001 | 5.5% | 5.6% | 0.83 |
| Congestive heart failure | 74.2% | 41.5% | <0.001 | 57% | 56.2% | 0.68 |
| Chronic kidney disease III-V | 35% | 21.6% | <0.001 | 27.9% | 27.5% | 0.74 |
| Chronic lung disease | 27.2% | 20.8% | <0.001 | 25.8% | 24.7% | 0.17 |
| Prior stroke | 14.9% | 8.9% | <0.001 | 10.7% | 11% | 0.59 |
| Peripheral vascular disease | 21.7% | 20.3% | 0.54 | 20.8% | 20.5% | 0.8 |
| Liver disease | 2.8% | 3.5% | 0.001 | 4.2% | 4.1% | 0.81 |
| Coagulopathy | 11% | 37.4% | <0.001 | 23.2% | 23.4% | 0.91 |
| Atrial fibrillation | 40.1% | 53.9% | <0.001 | 46.5% | 47.1% | 0.52 |
| Embolic protection device | 4.9% | 0% | 0.004 | 0% | 0% | 0.32 |
| Low frailty | 66% | 53.2% | <0.001 | 57.1% | 57.2% | 0.95 |
| Intermediate frailty | 32.9% | 44.9% | <0.001 | 41.1% | 40.8% | 0.80 |
| High frailty | 1.1% | 1.9% | <0.001 | 1.7% | 1.9% | 0.40 |
| Elective admissions | 83.4% | 78.6% | <0.001 | 79% | 78.8% | 0.93 |
| Teaching hospital | 89% | 83.5% | <0.001 | 85.8% | 86.2% | 0.78 |
| Hospital area | ||||||
| Large metropolitan | 60.3% | 54.6% | 0.003 | 55.9% | 56.9% | 0.70 |
| Small metropolitan | 39% | 43.2% | 0.028 | 42.8% | 41.9% | 0.73 |
| Micropolitan | 0.6% | 2.1% | <0.001 | 1.3% | 1.2% | 0.83 |
| Non-metropolitan | 0% | 0.1% | <0.001 | 0% | 0% | 0.93 |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
