基于临床科研共享系统建立膝骨关节炎分级模型

目的:运用有序Logistic回归分析筛选出膝骨关节炎分级的影响因素,建立膝骨关节炎分级模型,为临床膝骨关节炎分级的测评提供工具。方法:对2014年9月18日至2016年4月26日就诊于中国中医科学院望京医院的753例膝骨关节炎病例使用单因素、多因素有序Logistic回归分析,构建模型公式并进行评估。结果:研究发现与膝骨关节炎分级相关的影响因素为13项。其中危险因素包括高龄、肥胖、病程长、反复发作、下肢畸形、股四头肌萎缩,同时VAS评分、WOMAC指数、中医证候量表积分与分级程度呈正相关关系,单膝疼痛、疼痛范围局限、膝关节活动度与分级程度呈负相关关系。本研究构建出膝骨关节炎分级模型,评估检验效能良好(训练集AUC=0.860,测试集AUC=0.879)。结论:研究建立了膝骨关节炎分级模型,且在初步评估下发现具有良好的预测膝骨关节炎分级的能力。

Establishment of grading model of knee osteoarthritis based on clinical research system

Objective: The factors influencing the grade of knee osteoarthritis were screened by ordered Logistic regression analysis. The classification model of knee osteoarthritis was established,which provided tool for the evaluation of clinical classification of knee osteoarthritis.Methods: A total of 753 cases of knee osteoarthritis from Wangjing Hospital,China Academy of Chinese Medical Sciences were treated with single factor and multivariate Logistic regression analysis from September 18,2014 to April 26,2016. The model formula was constructed and evaluated.Results: In this study,it was found that 13 factors associated with knee osteoarthritis grading. Old age,obesity,long course of disease,recurrence,lower limb deformity,quadriceps muscle atrophy,WOMAC index,TCM syndrome score,VAS score and grading degree were positively correlated. One knee pain,limited pain range,knee activity and grading degree were negatively correlated. The knee osteoarthritis grading model was constructed and the evaluation test was effective(Training set AUC=0.860,test set AUC=0.879).Conclusion: In this study,a classification model of knee osteoarthritis was established which has a good ability to predict the classification of knee osteoarthritis under preliminary evaluation.