|
| 老年2型糖尿病合并骨质疏松患者骨折预测模型构建 |
| Construction of a fracture prediction model in patients with type 2 diabetes mellitus complicated with osteoporosis |
| |
| DOI：10.3969/j.issn.1006-7108.2025.07.008 |
| 中文关键词: 2型糖尿病 骨质疏松 骨折 贝叶斯网络 预测模型 |
| 英文关键词:type 2 diabetes mellitus osteoporosis fracture Bayesian network predictive model |
| 基金项目:广西哲学社会科学研究项目（23FRK003）；2024年右江民族医学院研究生创新计划项目（YZCXJH2023012） |
|
| 摘要点击次数: 975 |
| 全文下载次数: 0 |
| 中文摘要: |
| 目的 基于贝叶斯网络模型探讨2型糖尿病合并骨质疏松患者骨折发生的影响因素及因素间的交互关系。方法 将2022年12月至2024年4月于广西某三甲医院接受治疗的236例糖尿病合并骨质疏松患者作为研究对象，根据是否发生骨折将患者分为骨折组（71例）和非骨折组（165例）。比较两组患者临床资料，采用单因素及多因素回归分析2型糖尿病合并骨质疏松患者骨折发生的影响因素。通过R软件等构建贝叶斯模型，并进行模型的推理预测，对模型效能进行验证。结果 根据Logistic回归分析筛选出9个变量作为网络节点，构建一个含10个节点，12条有向边的2型糖尿病合并骨质疏松患者骨折发生影响因素的贝叶斯网络模型，并获得各节点的条件概率。结果显示，跌倒、腰椎T值、股骨颈T值、25(OH)D、甘油三酯、体质量指数等是骨折发生的独立危险因素。模型AUC值为0.879（95%CI：0.765～0.993，P＜0.001），敏感度为81.3%，特异度为75%。结论 本研究基于贝叶斯网络构建的骨折预测模型具有良好的预测能力，模型通过揭示各因素间的复杂交互关系，更准确地评估了骨折风险，为制定个性化医疗防治提供参考依据。 |
| 英文摘要: |
| Objective To explore the influence of the fracture in type 2 diabetes mellitus (T2DM) patients with osteoporosis based on Bayesian network model. Methods A total of 236 patients with T2DM and osteoporosis treated in our hospital from December 2022 to April 2024 were studied. They were divided into fracture group (71 cases) and non-fracture group (165 cases). The clinical data between the two groups were compared. Univariate and multivariate regression was used to analyze the influencing factors of fracture occurrence in patients with T2DM and osteoporosis. The Bayesian network model was built with R software. The inference of the model was predicted to verify the efficiency of the model. Results With the logistic regression analysis, 9 variables were selected as network nodes. A Bayesian network model with 10 nodes and 12 directed edges with T2DM and osteoporosis was constructed. The conditional probability of each node was obtained. The results showed that T values of the lumbar spine and the femoral neck, 25OHD, TG, BMI, and falls were independent risk factors for fracture occurrence (P<0.05). Model AUC value was 0.879 (95% CI: 0.765-0.993, P<0.001), with a sensitivity of 81.3% and a specificity of 75%, respectively. Conclusion The constructed fracture prediction model based on Bayesian network is good in prediction. The model reveals complex relationship among the factors and more accurately assesses the fracture risk. This provides reference basis for making personalized prevention scheme. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
| 关闭 |