文献速递医学人工智能与信息学流行病学与生物统计基础与转化研究

机器学习自评工具提升糖尿病前期与糖尿病检出

Enhancing prediabetes and diabetes detection through a machine learning-enabled self-assessment approach.
PubMed 文献 · J Clin Epidemiol · 2026-04-06

研究基于1988-2018年美国NHANES数据(N=17,458)开发并外部验证机器学习自评系统MEDWACS,仅用7项易获取参数预测糖尿病前期或糖尿病风险。内部验证ROCAUC为0.804,外部验证美国0.773、韩国0.780,校准良好,决策曲线分析显示临床效用优于现有筛查指南。

要点

  • MEDWACS仅需年龄、腰围、收缩压、性别、大腿长、臂围、BMI七项自评参数。
  • 外部验证ROCAUC美国0.773、韩国0.780,支持跨人群泛化。
  • 决策曲线分析显示其临床效用优于既定筛查指南。

核心结论

  • MEDWACS是一种经外部验证的无创机器学习风险分层工具,用七项易获取参数识别可能存在的糖尿病前期或糖尿病个体。
  • 该工具可促进及时临床评估,有望减轻公共卫生负担。

研究设计与样本

  • 分析1988-2018年NHANES数据(N=17,458),开发预测糖尿病前期/糖尿病风险的机器学习模型。
  • 复合结局定义为空腹血糖≥100 mg/dL或HbA1c≥5.7%(≥39 mmol/mol)。
  • 采用Boruta算法筛选关键预测因子,比较多种机器学习模型,最佳者(神经网络)构成MEDWACS。
  • 外部验证使用NHANES 2021-2023(N=3,043)和韩国NHANES 2023(N=5,492)。

主要发现

  • 最终七参数MEDWACS模型包含年龄、腰围、收缩压、性别、大腿长、臂围和体重指数。
  • 内部验证ROCAUC为0.804(95%CI 0.792-0.816),亚组表现稳健。
  • 外部验证ROCAUC美国为0.773(95%CI 0.756-0.790),韩国为0.780(95%CI 0.768-0.792),校准良好。
  • 决策曲线分析显示MEDWACS临床效用优于既定筛查指南。

关键数值与推荐

  • 预测结局复合标准:空腹血糖≥100 mg/dL或HbA1c≥5.7%(≥39 mmol/mol)。
  • MEDWACS七项自评参数:年龄、腰围、收缩压、性别、大腿长、臂围、体重指数。
  • 内部验证ROCAUC 0.804(95%CI 0.792-0.816);外部验证美国0.773(0.756-0.790)、韩国0.780(0.768-0.792)。

临床怎么用

  • MEDWACS可作为一种免费在线工具,供个体在家自评糖尿病前期/糖尿病风险,无需实验室检测。
  • 高风险结果可提示个体及时寻求临床评估,但不替代医生诊断。
  • 该工具可用于临床环境辅助风险分层。

以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。

原文

OBJECTIVES: Reliable, accessible, noninvasive self-assessment screening for prediabetes/diabetes is lacking, leading to missed opportunities for early intervention. We aimed to develop and externally validate a machine learning (ML)-derived self-assessment system to predict the likelihood of prevalent prediabetes or diabetes using easily accessible health parameters. STUDY DESIGN AND SETTING: We analyzed 30 years (1988-2018) of National Health and Nutrition Examination Survey (NHANES) data (N = 17,458). ML models predicting prediabetes/diabetes risk (composite outcome: fasting plasma glucose ≥100 mg/dL or HbA1c ≥ 5.7% [≥39 mmol/mol]) were developed using multimodal data. The Boruta algorithm identified key predictors. Multiple ML models were compared; the best performer (neural network) formed the Machineborne Early Diabetic Warning And Control System (MEDWACS). A final set of seven easily accessible parameters suitable for self-assessment was selected. Performance was assessed via receiver operating characteristic curve (ROCAUC) and calibration. External validation used NHANES 2021-2023 (N = 3043) and Korea NHANES 2023 (N = 5492). RESULTS: The final seven-parameter MEDWACS model included age, waist circumference, systolic blood pressure, gender, upper leg length, arm circumference, and body mass index. Internally, MEDWACS achieved ROCAUC 0.804 (95% confidence interval, 0.792-0.816) with robust subpopulation performance. External validation confirmed strong performance and generalizability (ROCAUCs: US 0.773 [0.756-0.790], Korea 0.780 [0.768-0.792]) and good calibration. Interpretability analysis identified key drivers. Decision curve analysis showed that MEDWACS had superior clinical utility compared to the established screening guidelines. An online tool was developed to facilitate home-based self-assessment and clinical use. CONCLUSION: MEDWACS provides a validated, noninvasive ML risk stratification tool using seven accessible parameters to identify individuals likely having prevalent prediabetes or diabetes. It can aid in prompting timely clinical evaluations, potentially reducing the public health burden. PLAIN LANGUAGE SUMMARY: Many people have prediabetes or type 2 diabetes and don't know it because early detection often requires clinic visits and blood tests. These can be significant barriers for some individuals. To address this, we developed a free, easy-to-use online tool called MEDWACS (Machineborne Early Diabetic Warning And Control System). Using 30 years of health data from the US National Health and Nutrition Examination Survey (NHANES), we used advanced computer learning methods to build a system that predicts a person's risk of having prediabetes or diabetes. The tool only needs seven simple pieces of information that can be measured at home, such as age, body mass index, waist circumference, and blood pressure, with no laboratory tests required. We confirmed that the tool is accurate using data from people in both the United States and South Korea. Our analysis showed that MEDWACS is more effective at identifying high-risk individuals than the current screening guidelines recommended by major US health organizations. MEDWACS can serve as a simple, no-cost first step for people to check their own risk. It is not a replacement for a doctor's diagnosis, but a high-risk result can empower individuals to seek timely medical evaluation. By making early risk detection more accessible, we hope that MEDWACS can help more people get diagnosed and treated sooner, potentially reducing the serious health complications of diabetes.

来源:J Clin Epidemiol,查看原文。版权归原作者所有。

引用

  1. Yoo D, Maggiore U, Jolliet O, et al. Enhancing prediabetes and diabetes detection through a machine learning-enabled self-assessment approach. J Clin Epidemiol. 2026. PMID: 41950976. doi:10.1016/j.jclinepi.2026.112266

供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。