| 王琦,舒适,鲁婵婵,等.基于LASSO回归建立与验证中风痰瘀滞络证诊断模型的探索[J].浙江中医药大学学报,2023,47(4):416-421. |
| 基于LASSO回归建立与验证中风痰瘀滞络证诊断模型的探索 |
| An Exploration to Establish and Verify A Diagnostic Model of Phlegm and Blood Stasis Blocked Collateral Syndrome of Wind-stroke Based on LASSO Regression |
| DOI：10.16466/j.issn1005-5509.2023.04.015 |
| 中文关键词: 中风 中医分型 痰瘀滞络证 LASSO回归 Logistic回归 预测模型 列线图 互联网医疗 |
| 英文关键词: wind-stroke TCM syndrome types phlegm and blood stasis blocked collateral syndrome LASSO regression Logistic regression prediction model nomogram telemedicine |
| 基金项目:浦东新区优秀青年医学人才项目(PWRq2020-53、PWRq2021-16)；上海市浦东新区临床中医特色学科建设资助项目(PDZY-2018-0612)；上海市浦东新区临床高原学科项目(PWYgy2021-11) |
|
| 摘要点击次数: 1043 |
| 全文下载次数: 674 |
| 中文摘要: |
| [目的] 探索并建立一个中风痰瘀滞络证的临床预测诊断模型，可应用于互联网医疗等现代大环境。[方法] 对2021年6月至2022年2月上海市浦东新区浦南医院中医科收治的新发脑卒中有偏瘫后遗症的患者305例进行病史采集，给予中医证型诊断。将病例以7:3的比例区分为训练集及验证集，运用R语言进行LASSO回归筛选疾病研究因子，以二元Logistic回归建立模型，并予区分度、校准度验证，以列线图形式展示模型。[结果] 30个研究因子经过筛选后留下9个因子建立模型，模型区分度:训练集曲线下面积(area under curve，AUC)0.942，95%可信区间(0.906，0.979)；验证集AUC 0.951，95%可信区间(0.895，1.000)。校准度:Hosmer-Lemeshow指数(H-L)训练集P=0.47，验证集P=0.39。模型以列线图进行可视化展示。[结论] 该诊断模型有较好的诊断效能，可在多种无法进行望诊、切诊的情况下给予辨证参考。 |
| 英文摘要: |
| [Objective] To explore and develop a clinical predictive diagnostic model of phlegm and blood stasis blocking collateral syndrome of wind-stroke， which is applied potentially to telemedicine， a modern remote clinical service. [Methods] From June 2021 to February 2022， 305 new-onset wind-stroke patients with sequela of hemiplegia were collected and differentiated for traditional Chinese medicine(TCM) syndromes through TCM Department of Pu‘nan Hospital. Patients were divided into either training or verifying group with ratio of 7:3. R Language and LASSO regression were applied to screen disease research factors. The binary Logistic regression was used to establish the diagnostic model as well as verify the classification and calibration. The model was displayed in nomogram. [Results] The diagnostic model was established by nine research factors， which were filtrated from thirty candidates. The report model classification in training group indicated the area under curve(AUC) was 0.942 with 95% confidence interval(0.906，0.979)； and AUC was 0.951 with 95% confidence interval(0.895，1.000) in verifying group. In terms of Hosmer-Lemeshow index(H-L) of calibration，P=0.47 was the result of training group， whereas P=0.39 was the outcome of verifying group. The model was displayed in nomogram visually. [Conclusion] This diagnostic model was proved to be effective. It can be used as a reference for syndrome differentiation when face-to-face inspection and palpitation are not available. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |