Machine learning-based prediction model for myocardial ischemia under high altitude exposure: a cohort study

• Published in: Scientific reports 2024-01, Vol.14 (1), p.686-686, Article 686
• Main Authors: Chen, Yu, Zhang, Xin, Ye, Qing, Cao, Ning, Li, Shao-Ying, Yu, Jie, Zhao, Sheng-Tao, Zhang, Juan, Xu, Xin-Ming, Shi, Yan-Kun, Yang, Li-Xia
• Format: Article
• Language: English
• Subjects: 631/114; 692/1537; 692/4019; Altitude; Aspartate aminotransferase; Cardiovascular diseases; Cohort analysis; Datasets; Globulins; High-altitude environments; Humanities and Social Sciences; Ischemia; Learning algorithms; Leukocytes (eosinophilic); Machine learning; multidisciplinary; Myocardial ischemia; Prediction models; Questionnaires; Risk factors; Science; Science (multidisciplinary)
• ISSN: 2045-2322; 2045-2322
• DOI: 10.1038/s41598-024-51202-8

High altitude exposure increases the risk of myocardial ischemia (MI) and subsequent cardiovascular death. Machine learning techniques have been used to develop cardiovascular disease prediction models, but no reports exist for high altitude induced myocardial ischemia. Our objective was to establish a machine learning-based MI prediction model and identify key risk factors. Using a prospective cohort study, a predictive model was developed and validated for high-altitude MI. We consolidated the health examination and self-reported electronic questionnaire data (collected between January and June 2022 in 920th Joint Logistic Support Force Hospital of china) of soldiers undergoing high-altitude training, along with the health examination and second self-reported electronic questionnaire data (collected between December 2022 and January 2023) subsequent to their completion on the plateau, into a unified dataset. Participants were subsequently allocated to either the training or test dataset in a 3:1 ratio using random assignment. A predictive model based on clinical features, physical examination, and laboratory results was designed using the training dataset, and the model's performance was evaluated using the area under the receiver operating characteristic curve score (AUC) in the test dataset. Using the training dataset (n = 2141), we developed a myocardial ischemia prediction model with high accuracy (AUC = 0.86) when validated on the test dataset (n = 714). The model was based on five laboratory results: Eosinophils percentage (Eos.Per), Globulin (G), Ca, Glucose (GLU), and Aspartate aminotransferase (AST). Our concise and accurate high-altitude myocardial ischemia incidence prediction model, based on five laboratory results, may be used to identify risks in advance and help individuals and groups prepare before entering high-altitude areas. Further external validation, including female and different age groups, is necessary.