top of page

Obstructive sleep apnea predicts 10-year cardiovascular disease–related mortality in the Sleep Heart Health Study: a machine learning approach

In this study, we developed a random forest model on a population with OSA to assess its ability to calculate the risk of CVD mortality after 10 years and compared it with the Framingham Risk Score. The random forest model performed the best with a history, lipid levels, blood pressure, and use of antihypertensive medications. In contrast, the most important features in our approach had little overlap with the Framingham Risk Score with the exception of age and history of hypertension. Thus, it appears that in an OSA population, factors predicting 10-year CVD mortality are different than in a general adult population.

We found AHI is an important predictor for assessing the risk of 10-year CVD mortality. By applying mutual information analysis, we found it was more informative for predicting CVD mortality than diabetes, high density lipoprotein, and cholesterol, which are commonly used features, although it was less useful than FVC and FEV 1 . Additionally, the random forest model learned to use AHI as an important feature to build decision trees and showed a high discrimination ability. We also found severe OSA patients have a higher risk than mild OSA patients on 10-year CVD mortality.

0 views0 comments


bottom of page