Machine learning models can use demographic and lifestyle data, physical examination results, and laboratory values to effectively predict a patient’s risk of sleep problems, according to a new study published this week in an open-access journal. and, pro swan Samuel Y. Huang, Virginia Commonwealth University School of Medicine and Alexander A. Huang, Northwestern Feinberg University School of Medicine, USA
The prevalence of diagnosed sleep disorders among American patients has increased significantly over the past decade. Sleep disorders are a significant risk factor for diabetes, heart disease, obesity and depression, so it is important to better understand and reverse this trend.
In a new study, researchers used the machine learning model XGBoost to analyze public data on 7,929 US patients who completed the National Health and Nutrition Examination Survey. Data included 684 variables per patient and included demographic, diet, exercise, and mental health questionnaire responses, as well as clinical and physical examination information.
Overall, 2,302 patients in the study were diagnosed with sleep disorders by their physicians. XGBoost was able to predict the risk of sleep disorder diagnosis with high accuracy (AUROC=0.87, sensitivity=0.74, specificity=0.77) using a total of 64 variables in the full dataset. The greatest predictors of sleep disturbance based on machine learning models were depression, weight, age and waist circumference.
The authors conclude that machine learning methods may be an effective first step to screen patients for sleep disorder risk without relying on physician judgment or bias.
Samuel Y. Huang adds: Fifty-seven other variables are associated with insomnia, but the contribution of each is visualized in a highly predictive model. ”
For more information:
using machine learning to identify risk factors for insomnia, pro swan (2023). DOI: 10.1371/journal.pone.0282622
Journal information:
pro swan
Courtesy of the Public Science Library