目的 构建老年女性骨质疏松症(OP)患者髋关节骨折的风险预测决策树模型，并验证其效能。方法 分析郑州大学第一附属医院2020年1月至2023年6月收治的315例老年女性OP患者的临床资料，采用随机数字表分为建模组(210例)和测试组(105例)。建模组根据是否发生髋关节骨折分为发生组(41例)和未发生组(169例)，比较两者的资料，Logistic回归分析影响因素，构建决策树模型;绘制受试者工作特征(ROC)曲线验证模型的效能。结果 髋关节骨折发生率为19。68％，建模组19。52％、测试组20。00％;OP病程[比值比(OR)=2。123，95％可信区间(CI):1。143～3。945]、合并糖尿病(OR=2。018，95％CI:1。116～3。647)、子宫切除手术史(OR=1。584，95％CI:1。048～2。395)、平均每天运动时间(OR=0。541，95％CI:0。327～0。895)、平均每天光照时间(OR=0。547，95％CI:0。379～0。791)、骨折风险评估工具(FRAX)(OR=2。261，95％CI:1。241～4。120)、未遵医嘱抗 OP 治疗(OR=1。817，95％CI:1。206～2。736)、外力打击(OR=4。607，95％CI:1。193～6。143)均是髋关节骨折的影响因素(P＜0。05);构建的决策树模型筛选出上述8个解释变量，且外力打击与髋关节骨折最具相关性(x2=10。488，P＜0。05);决策树模型预测测试组髋关节骨折发生的灵敏度和特异度分别为85。71％、88。10％，曲线下面积(AUC)(95％CI)为0。973(0。949～0。998)。结论 OP病程、合并糖尿病、子宫切除手术史、平均每天运动时间、平均每天光照时间、FRAX、未遵医嘱抗OP治疗、外力打击均是老年女性OP患者髋关节骨折的影响因素，基于此构建的决策树模型效能高。
Prediction of hip fractures in elderly women with osteoporosis
Objective To construct a risk prediction decision tree model for hip fractures in elderly women with osteoporosis(OP)and verify its efficacy.Methods An analysis was conducted on the clinical data of 315 elder women with OP admitted to the hospital from January 2020 to June 2023.A random num-ber table was used to divide the patients into a modeling group(n=210)and a testing group(n=105).The modeling group was divided into two subgroups based on the occurrence of hip fracture:the occurrence subgroup(n=41)and the non-occurrence subgroup(n=169).The data of the two groups were compared,and the factors were analyzed using Logistic regression,then a decision tree model was constructed.Re-ceiver operating characteristic(ROC)curve was drawn to validate the model efficiency.Results The inci-dence of hip joint fractures was 19.68％,of which modeling and testing groups were 19.52％and 20.00％respectively.Course of OP[odds ratio(OR)=2.123,95％confidence interval(CI):1.143-3.945),di-abetes(OR=2.018,95％CI:1.116-3.647),history of hysterectomy(OR=1.584,95％CI:1.048-2.395),average daily exercise time(OR=0.541,95％CI:0.327-0.895),average daily light time(OR=0.547,95％CI:0.379-0.791),fracture risk assessment(FRAX)(OR=2.261,95％CI:1.241-4.120),and failure to follow the doctor's advice on anti-OP treatment(OR=1.817,95％CI:1.26-2.736)and external impact(OR=4.607,95％CI:1.193-6.143)were contributing factors to hip joint fractures(P＜0.05).The decision tree model constructed selected the above 8 explanatory variables,and external force impact was the most correlated with hip joint fractures(x2=10.488,P＜0.05).The sensi-tivity and specificity of the decision tree model in predicting the occurrence of hip joint fractures in the tes-ting group were 85.71％and 88.10％,respectively.The area under the curve(AUC)(95％CI)was 0.973(0.949-0.998).CConclusion Course of OP,diabetes,history of hysterectomy,average daily ex-ercise time,average daily light time,FRAX,failure to follow the doctor's advice on anti-OP treatment and external force blow are all influencing factors of hip fracture in elder women with OP,and the decision tree model constructed based on this has high efficiency.
NETL
NSTL
万方数据
