2.4. Statistical analyses
Propensity scores (PS) were used to correct potential systematic differences between AF (n=12,797) and no-AF groups (n=289,432) (Figure 1). Each patient’s PS to estimate the risk of developing AF was calculated and adjusted for the covariates in a logistic regression analysis model. PS matching was performed on logit-transformed PS-matched to the nearest neighbor in a 1:1 fashion with a caliper of 0.1, without using replacement.
The details of methods are presented in method of supplementary material.