Author + information
- Ioannis A. Kakadiaris,
- Bassam Almogahed,
- Harvey Hecht,
- Christopher Sibley,
- Morteza Naghavi and
- Matthew Budoff
Coronary Artery Calcification (CAC) is a CT measure of subclinical atherosclerosis that may improve cardiovascular risk prediction. Integration of CAC and traditional risk factors (TRF) using machine learning (ML) algorithms has not been described.
6,474 men and women in the Multi-Ethnic Study of Atherosclerosis were classified using a decision tree based algorithm including TRF. Subjects were categorized into low (FRS <10%) and intermediate-high (FRS 10%) groups. Gender-specific and an overall prediction model were generated for each class of events. Endpoints were all coronary heart disease events (CHDA) (angina, revascularization, MI, CH death), hard CH events (MI, CH death) and all cardiovascular events (CVDA) (stroke, angina, revascularization, MI, CH death). Specifically, we developed nine models (CardioRSi) that do not use CAC (base models) and nine models (CardioRS-Ci) that use CAC.
CardioRSi models achieved improved performance over FRS and had notably increased NRI in women. Inclusion of CAC to the base models increased NRI for both genders.
ML algorithms improve CHD and CVD risk prediction compared to existing models, particularly in women. Addition of CAC to the base models improved overall NRI for both genders.
Table: NRI among Different Techniques. A: NRI when comparing CardioRSi predictions to FRS predictions. B: NRI when comparing CardioRS-Ci to FRS plus CAC predictions. C: NRI when comparing CardioRS-Ci to CardioRSi predictions.
Poster Sessions, Expo North
Sunday, March 10, 2013, 3:45 p.m.-4:30 p.m.
Session Title: Acute Coronary Syndromes: Basic IV
Abstract Category: 2. Acute Coronary Syndromes: Basic
Presentation Number: 1256-179
- 2013 American College of Cardiology Foundation