Risk Models to Predict the 6-year Risk of Lifestyle-Related Diseases by Health Checkup Data

DOI Web Site

Bibliographic Information

Other Title
  • 健康診断データによる生活習慣病6年以内発症のリスクモデル
  • ケンコウ シンダン データ ニ ヨル セイカツ シュウカンビョウ 6ネン イナイ ハッショウ ノ リスクモデル

Search this article

Abstract

<p>Objective: The objective of this study was to develop risk models for predicting the 6-year risk of lifestyle-related diseases (diabetes, obesity, hypertension, dyslipidemia, and hepatic dysfunction) based on annual health checkup data and medical prescriptions to identify high-risk individuals. </p><p>Methods: The participants were 191,458 individuals who received annual health checkups between 2010 and 2016. Data of 79,414 and 39,778 individuals were used for training and validation of the diabetes risk model, respectively. Risk models, based on random survival forests which were trained using health checkup data and medical prescriptions, were supplementally used to define incident cases. Learning parameters such as the number of trees, maximum depth allowed for a tree, number of variables, and minimum size of terminal nodes were optimized to improve performance. </p><p>Results: The area under the curve (AUC) for predicting the 3-year risk score was 0.963 (95%CI, 0.956–0.969) for diabetes, 0.935 (95%CI, 0.931–0.939) for obesity, 0.864 (95%CI, 0.858–0.871) for systolic hypertension, 0.940 (95%CI, 0.928–0.950) for dyslipidemia (high triglycerides), and 0.852 (95%CI, 0.838–0.864) for hepatic dysfunction (high aspartate aminotransferase). </p><p>Conclusion: Risk models of lifestyle-related diseases were developed to make predictions based on annual health checkup data. These models showed fair to excellent performance, suggesting they are useful for risk stratification.</p>

Journal

Keywords

Details 詳細情報について

Report a problem

Back to top