Page 150 - Read Online

Dokko et al. Vessel Plus 2022;6:53 https://dx.doi.org/10.20517/2574-1209.2022.11 Page 5 of 15

Chronic kidney disease < 0.01
Stage 3 11.57% 6.12% 20.00%
Stage 4 1.24% 0.68% 2.11%
ESRD 5.79% 4.76% 7.37%
CKD, with dialysis 3.31% 2.04% 5.26% 0.27
CKD, without dialysis 23.55% 14.97% 36.84% < 0.01
Obesity 22.73% 18.37% 29.47% 0.04
Iron deficiency anemia 13.22% 12.93% 13.68% 0.86
Rheumatoid arthritis/collagen vascular diseases 4.96% 5.44% 4.21% 0.77
Fluid and electrolyte disorders 1.65% 2.04% 1.05% 0.65
Pulmonary hypertension 20.66% 16.33% 27.37% 0.04
Thrombocytopenia 28.51% 33.33% 21.05% 0.04
Previous internal cardioverter-defibrillator 3.31% 0.68% 7.37% 0.01
Hypothyroidism 13.64% 12.24% 15.79% 0.43
Intra-aortic balloon pump 2.89% 2.72% 3.16% 1.00
*For categorical variables, P-values were based on chi-squared test with exact P-value from Monte Carlo simulation; for continuous variables,
P-value was based on Welch’s t-test. Note: For categorical variables, column percentages were reported; for continuous variables, mean +/- std
were reported. CHF: Congestive heart failure; CAD: coronary artery disease; COPD: chronic obstructive pulmonary disease; MI: myocardial
infarction; CKD: chronic kidney disease; r-SAVR: repeat surgical aortic valve replacements; ViV-TAVR: valve-in-valve transcatheter aortic valve
replacements.

the demographics and risk factors to select each model’s eligible variables. Using a clinical conceptual
framework, domain-related variables were evaluated for potential collinearity. These standard multivariable
model eligible screening steps were used for both propensity score matching to identify predictors of
POAF/AFL, as well as to identify the POAF/AFL impact upon risk-adjusted MM and READMIT.
Additionally, Firth bias correction was used in all MM and READMIT models to correct semi-separation
issues due to data sparsity.

In the context of each patient’s unique risk factor profile, patients’ propensity scores were calculated by
applying the POAF/AFL multivariable model’s algorithm; the basis for these propensity score calculations
was the final POAF/AFL multivariate model that identified patient risk characteristics predictive of
POAF/AFL. These patient specific POAF propensity scores were used as model eligible variables for both
the MM and READMIT models. To evaluate each model’s predictive power and calibration, the
performance metrics (i.e., C-index and Hosmer-Lemeshow test) were reported [Tables 4-6].

Based on protocol-driven hypotheses, statistical significance was pre-established at P < 0.05 for the
co-primary outcomes and P < 0.01 for the secondary and tertiary outcomes. The final models’ statistically
significant variables and their odds ratios with 95% confidence intervals have been reported; however, all
P-values have been shared to facilitate independent interpretation.

RESULTS
Baseline characteristics of r-SAVR and ViV-TAVR study population
From the SPARCS Database, 74,675 first-time SAVR/TAVR procedures were recorded, of whom 242
patients underwent r-AVR procedures from January 2005 to November 2018 - 147 r-SAVR and 95
ViV-TAVR patients [Figure 1].

Patients who underwent r-SAVR were significantly younger than patients who underwent ViV-TAVR, with
a mean age of 61.8 ± 13.9 years and 74.0 ± 11.6 (P < 0.01.), respectively, and are shown in [Table 1]. Across
the three major insurance categories (i.e., Commercial insurance, Medicaid/other insurance, and Medicare