Journal
HEART RHYTHM
Volume 18, Issue 3, Pages 426-433Publisher
ELSEVIER SCIENCE INC
DOI: 10.1016/j.hrthm.2020.10.022
Keywords
Acquired long QT syndrome; All-cause mortality; Myocardial repolarization; Non-negative matrix factorization; Random survival forests
Categories
Ask authors/readers for more resources
Through the RSF and NMF analyses, it was found that factors such as cancer, potassium, and calcium levels, as well as ECG indicators, can predict all-cause mortality in acquired long QT syndrome patients. Increasing the number of latent features in ECG indices significantly improved the prediction ability of the model.
BACKGROUND Acquired long QT syndrome (aLQTS) is often associated with poor clinical outcomes. OBJECTIVE The purpose of this study was to examine the important predictors of all-cause mortality of aLQTS patients by applying both random survival forest (RSF) and non-negative matrix factorization (NMF) analyses. METHODS Clinical characteristics and manually measured electro-cardiographic (ECG) parameters were initially entered into the RSF model. Subsequently, latent variables identified using NMF were entered into the RSF as additional variables. The primary outcome was all-cause mortality. RESULTS A total of 327 aLQTS patients were included. The RSF model identified 16 predictive factors with positive variable importance values: cancer, potassium, RR interval, calcium, age, JT interval, diabetes mellitus, QRS duration, QTp interval, chronic kidney disease, QTc interval, hypertension, QT interval, female, JTc interval, and cerebral hemorrhage. Increasing the number of latent features between ECG indices, which incorporated from n = 0 to n = 4 by NMF, maximally improved the prediction ability of the RSF-NMF model (C-statistic 0.77 vs 0.89). CONCLUSION Cancer and serum potassium and calcium levels can predict ad-cause mortality of aLQTS patients, as can ECG indicators including JTc and QRS. The present RSF-NMF model significantly improved mortality prediction.
Authors
I am an author on this paper
Click your name to claim this paper and add it to your profile.
Reviews
Recommended
No Data Available