Yasuda Kosuke, Tomoda Shiori, Suzuki Mayumi, Wada Toshikazu, Fujikawa Toshiyuki, Kikutsuji Toru, Kato Shintaro
NEC Solution Innovators, Ltd., Tokyo, Japan.
Kurashiki Central Hospital Preventive Healthcare Plaza, Okayama, Japan.
AJPM Focus. 2024 Oct 3;3(6):100277. doi: 10.1016/j.focus.2024.100277. eCollection 2024 Dec.
Tools developed to assess individuals' comprehensive health status would be beneficial for personalized prevention and treatment. This study aimed to develop a set of risk prediction models to estimate the risk for multiple diseases such as heart, blood vessel, brain, metabolic, liver, and kidney diseases using health checkup data only.
This is a retrospective study that used health checkup data combined with diagnostic information from electronic health records of Kurashiki Central Hospital in Okayama, Japan. All exposure factors were measured at the first health checkup visit, including demographic characteristics, laboratory test results, lifestyle questionnaires, medication use, and medical history. Primary outcomes were the diagnoses of 15 diseases during the follow-up period. Cox proportional hazard regression was applied to develop risk prediction models for heart, blood vessel, brain, metabolic, liver, and kidney diseases. Area under the curve with 4-year risk assessments were performed to evaluate the models.
From January 2012 to September 2022, a total of 92,174 individuals aged 15-96 years underwent general health checkups. The area under the curve of the models in validation datasets was as follows: atrial fibrillation, 0.81; acute myocardial infarction, 0.81; heart failure, 0.76; cardiomyopathy, 0.72; angina pectoris, 0.70; atherosclerosis, 0.82; hypertension, 0.80; cerebral infarction, 0.77; intracerebral hemorrhage, 0.68; subarachnoid hemorrhage, 0.50; type-2 diabetes mellitus, 0.82; hyperlipidemia, 0.70; alcoholic liver disease, 0.91; liver fibrosis, 0.92; and chronic kidney disease, 0.80.
A set of prediction models to estimate multi-disease risk simultaneously from health checkup results may help to assess comprehensive individual health status and facilitate personalized prevention and early diagnosis.
开发用于评估个体综合健康状况的工具将有助于个性化预防和治疗。本研究旨在开发一套风险预测模型，仅使用健康体检数据来估计多种疾病的风险，如心脏、血管、脑、代谢、肝脏和肾脏疾病。
这是一项回顾性研究，使用了日本冈山县仓敷中央医院的健康体检数据以及电子健康记录中的诊断信息。所有暴露因素均在首次健康体检时测量，包括人口统计学特征、实验室检查结果、生活方式问卷、用药情况和病史。主要结局是随访期间15种疾病的诊断。采用Cox比例风险回归来开发心脏、血管、脑、代谢、肝脏和肾脏疾病的风险预测模型。进行了4年风险评估的曲线下面积分析以评估模型。
2012年1月至2022年9月，共有92174名年龄在15 - 96岁的个体接受了一般健康体检。验证数据集中模型的曲线下面积如下：心房颤动为0.81；急性心肌梗死为0.81；心力衰竭为0.76；心肌病为0.72；心绞痛为0.70；动脉粥样硬化为0.82；高血压为0.80；脑梗死为0.77；脑出血为0.68；蛛网膜下腔出血为0.50；2型糖尿病为0.82；高脂血症为0.70；酒精性肝病为0.91；肝纤维化0.92；慢性肾脏病为0.80。
一套能从健康体检结果中同时估计多种疾病风险预测模型，可能有助于评估个体综合健康状况，并促进个性化预防和早期诊断。
