Data Mining of Population Health-sub-health-disease Based on Dynamic System Theory
试验速览
- 阶段
- 不适用
- 状态
- 进行中（未招募）
- 入组人数
- 380,000
- 试验地点
- 1
- 主要终点
- Discriminative Accuracy for Next Diagnosis
研究概览
简要总结
This study aims to explore the dynamic evolution patterns of population health, sub-health, and disease states through dynamic system theory and big data mining methods, providing scientific evidence for personalized prevention and health management.
详细描述
Specific objectives include: (1) Identifying individual health, sub-health, and disease states using unsupervised system modeling techniques, while investigating their mutual transformation pathways. (2) Identifying key indicators determining state transitions, clarifying their mechanisms and interactions. (3) Developing dynamic system models to simulate state transition trajectories under multivariate influences, predicting individual probabilities of progression from health to sub-health or disease. (4) Creating interpretable health prediction tools based on modeling results to support precision interventions. The ultimate goal is to establish a scientifically validated yet implementable health state modeling system, offering quantifiable tools for early intervention and personalized health management to reduce chronic disease incidence and healthcare burdens.
研究设计
- 研究类型
- Observational
- 观察模型
- Cohort
- 时间视角
- Retrospective
入排标准
- 性别
- All
- 接受健康志愿者
- 是
入选标准
- •Participants must have completed at least two consecutive physical examinations at the Physical Examination Center of Beijing Friendship Hospital, Capital Medical University, between June 2007 and August 2025, with a minimum interval of 6 months between adjacent records.
- •Data records should be relatively complete, with missing rates for key research variables (e.g., core biochemical indicators, demographic information, and essential questionnaire items) ≤30%.
- •Participants must have no prior history of severe organic diseases prior to their first study inclusion (as documented in medical records, primarily including: malignant tumors (non-curable/end-stage), severe cardiac insufficiency (NYHA Class III-IV), end-stage renal disease (CKD Stage 5), decompensated cirrhosis, or significant functional impairment caused by sequelae of severe cerebrovascular disease).
排除标准
- •- Individuals with a severe lack of basic data (such as unique identification, key demographic information, and core indicators of detection) or who cannot be effectively anonymized.
研究组 & 干预措施
Health Data Science Database of Beijing Friendship Hospital
- Participants must have completed at least two consecutive physical examinations at the Physical Examination Center of Beijing Friendship Hospital, Capital Medical University, between June 2007 and August 2025, with a minimum interval of 6 months between adjacent records.
- Data records should be relatively complete, with missing rates for key research variables (e.g., core biochemical indicators, demographic information, and essential questionnaire items) ≤30%.
- Participants must have no prior history of severe organic diseases prior to their first study inclusion (as documented in medical records, primarily including: malignant tumors (non-curable/end-stage), severe cardiac insufficiency (NYHA Class III-IV), end-stage renal disease (CKD Stage 5), decompensated cirrhosis, or significant functional impairment caused by sequelae of severe cerebrovascular disease).
干预措施: No intervention will be applied. (Other)
结局指标
主要结局
Discriminative Accuracy for Next Diagnosis
时间窗: Evaluate on an internal validation dataset. This dataset contains individual historical data up to January 1, 2018, based on which the model predicts the next diagnostic event that will occur immediately. Calculate AUC for diseases with over 1000 ICD-10
Age- and Sex-stratified Area Under the Receiver Operating Characteristic Curve, AUC
Long-term Predictive Accuracy
时间窗: Evaluate the AUC values of disease occurrence in the 1st, 2nd, 3rd, 5th, and 10th year after prediction on the internal validation dataset.
AUC stratified by age and gender, assessing the risk of disease occurrence within specific time intervals (1 year, 2 years,..., 10 years) after prediction. This indicator measures the decay of a model's predictive ability over time.
Trajectory-level Predictive Accuracy
时间窗: On the validation subset, evaluate the accuracy of disease event predictions for each year from the simulation starting point (60 years old) to the following 1 to 20 years.
The proportion of correctly predicted disease events. In each simulated future year, match the generated disease events with the actual disease events that occur in individuals, and calculate the success rate (%) of the matching.
次要结局
未报告次要终点
研究者
Lv, Han
Professor
Beijing Friendship Hospital