Introduction
Due to Brazilian population aging, prevalence of aortic stenosis, and limited number of
scores in literature, it is essential to develop risk scores adapted to our reality and
created in the specific context of this disease.
Methods
This is an observational historical cohort study with analysis of 802 aortic stenosis
patients who underwent valve replacement at Hospital São Lucas, Pontifícia
Universidade Católica do Rio Grande do Sul, from 1996 to 2018. With the aid of logistic
regression, a weighted risk score was constructed based on the magnitude of the coeficients
β of the logistic equation. Two performance statistics were obtained: area under the
receiver operating characteristic curve and the chi-square (χ2) of Hosmer-Lemeshow
goodness-of-fit with Pearson’s correlation coeficient between the observed events and
predicted as a model calibration estimate.
Results
The risk predictors that composed the score were valve replacement surgery combined with
coronary artery bypass grafting, prior renal failure, New York Heart Association class III/IV
heart failure, age > 70 years, and ejection fraction < 50%. The receiver operating
characteristic curve area was 0.77 (95% confidence interval: 0.72-0.82); regarding the model
calibration estimated between observed/predicted mortality, Hosmer-Lemeshow test χ2 =
3,70 (
P
=0.594) and Pearson’s coeficient r = 0.98
(
P
<0.001).
Conclusion
We propose the creation of a simple score, adapted to the Brazilian reality, with good
performance and which can be validated in other institutions.