高龄膝骨关节炎患者跌倒的危险因素分析及风险预测模型建立

Risk factors and a risk forecasting model for falls in aged patients with knee osteoarthritis

目的:分析高龄膝骨关节炎患者跌倒的危险因素,建立跌倒风险预测模型。方法:以2020年1月至2022年4月在遂宁市中心医院就诊的年龄≥65岁的膝骨关节炎患者为研究对象。根据6个月内是否有排除外力因素的跌倒史进行分组,有跌倒史者归入跌倒组,无跌倒史者归入未跌倒组。收集2组患者的年龄、性别、病程、体质量指数(body mass index,BMI)、患病侧别、其他部位疼痛情况、合并其他慢性病情况、其他部位关节置换史、经常爬楼梯史、膝关节外伤史、Kellgren-Lawrence分级等一般资料。采用西安大略和麦克马斯特大学骨关节炎指数(Western Ontario and McMaster Universities osteoarthritis index,WOMAC)评价患膝疼痛、僵硬和功能受限程度,采用Lysholm膝关节评分评价膝关节功能。进行下肢三维运动学和动力学测试,收集包括时空参数、运动学参数和动力学参数在内的各项数据。比较2组患者的一般资料、临床评价结果及下肢各项力学参数,进行单因素分析。根据单因素分析结果采用Logistic回归分析高龄膝骨关节炎患者跌倒的危险因素,构建跌倒风险预测模型,并对模型进行评价。结果:①单因素分析结果。共纳入140例患者,跌倒组78例,未跌倒组62例。2组患者病程、其他部位疼痛情况、其他部位关节置换史、经常爬楼梯史、膝关节外伤史、Kellgren-Lawrence分级的组间比较,差异均有统计学意义。跌倒组患者WOMAC评分中疼痛、僵硬、功能障碍评分均高于未跌倒组,对侧腿离地期、步时长于未跌倒组,步长短于未跌倒组,步速慢于未跌倒组,足跟着地期踝背伸角、支撑期最大踝跖屈角、最大踝背伸角、最大踝跖屈角、足跟着地期膝伸角、支撑期最大膝伸角均大于未跌倒组,支撑期最大膝屈角、最大膝屈角、最大髋伸角、髋屈力矩峰值、髋伸力矩峰值、膝屈力矩峰值、膝伸力矩峰值、踝跖屈力矩峰值均小于未跌倒组。②高龄膝骨关节炎患者跌倒的危险因素分析及风险预测模型评价结果。Logistic回归分析结果显示,病程长、经常爬楼梯、疼痛重、足跟着地期踝背伸角增大、足跟着地期膝伸角增大、最大膝屈角减小、膝屈力矩峰值减小、踝跖屈力矩峰值减小均为高龄膝骨关节炎患者跌倒的独立危险因素[OR=5.230,95%CI(3.232,7.021),P=0.004;OR=4.367,95%CI(2.648,5.953),P=0.003;OR=4.252,95%CI(2.159,6.231),P=0.003;OR=3.473,95%CI(2.982,4.028),P=0.021;OR=6.977,95%CI(3.667,8.964),P=0.001;OR=3.989,95%CI(1.667,5.264),P=0.010;OR=7.051,95%CI(4.267,8.164),P=0.001;OR=4.675,95%CI(2.563,6.798),P=0.008]。跌倒风险预测列线图模型显示,病程>6年、经常爬楼梯、WOMAC疼痛评分>48分、足跟着地期踝背伸角>9.25°、足跟着地期膝伸角>2.35°、最大膝屈角<45°、膝屈力矩峰值<0.65 N·m·kg^(-1)、踝跖屈力矩峰值<0.90 N·m·kg^(-1)时,高龄膝骨关节炎患者跌倒风险的预测值总分为409分,患者发生跌倒的概率为71.90%。受试者操作特征(receiver operating characteristics,ROC)曲线分析结果显示,预测模型区分度较高,训练集ROC曲线下面积为0.698(P=0.000),灵敏性71.43%,特异性65.89%;验证集ROC曲线下面积为0.699(P=0.000),灵敏性78.26%,特异性63.18%。Hosmer-Lemeshow拟合优度检验结果显示,预测模型拟合优度好(训练集:χ^(2)=0.748,P=0.504;验证集:χ^(2)=1.328,P=0.263)。临床决策曲线分析结果显示,训练集阈概率在9%~88%时净获益率高;验证集阈概率在11%~92%时净获益率高。结论:病程长、经常爬楼梯、疼痛重、足跟着地期踝背伸角增大、足跟着地期膝伸角增大、最大膝屈角减小、膝屈力矩峰值减小、踝跖屈力矩峰值减小均为高龄膝骨关节炎患者跌倒的独立危险因素;根据这些危险因素建立的跌倒风险预测模型,对于高龄膝骨关节炎患者的跌倒风险具有一定的预测价值。

Objective:To analyze the risk factors for falls in aged patients with knee osteoarthritis(KOA),and to build a fall risk prediction model.Methods:The patients aged≥65 years who were treated in the Suining Central Hospital for KOA from January 2020 to April 2022 were selected as the subjects.The patients with and without the history of fall caused by non-external force factors within the past 6 months were assigned into a fall group and a non-fall group,respectively.The general information,including age,gender,disease course,body mass index(BMI),affected side,pain in other sites,combined with other chronic diseases,history of joint replacement in other parts,history of frequent stair climbing,history of knee trauma,Kellgren-Lawrence classification,was collected.The affected knee pain degree,stiffness and functional limitation degree were evaluated by using Western Ontario and McMaster Universities osteoarthritis index(WOMAC),and the knee function was assessed by using Lysholm knee score;furthermore,the three-dimensional kinematic and kinetic tests on the lower limbs were conducted,and the temporal-spatial parameters,kinematic parameters,and kinetic parameters were measured and extracted.Moreover,the general information,clinical evaluation results,and lower limb mechanical parameters were compared between the 2 groups,and based on the single factor analysis results,the risk factors for falls among aged KOA patients were analyzed by logistic regression,and then a fall risk prediction model was constructed and evaluated.Results:①One hundred and forty patients were included in the final analysis,78 ones in the fall group,and 62 ones in the non-fall group.The differences in disease course,pain in other sites,history of joint replacement in other parts,history of frequent stair climbing,history of knee trauma,and Kellgren-Lawrence classification were statistically significant between the 2 groups.The scores of pain,stiffness and dysfunction in WOMAC scores were higher,the contralateral leg off-ground time and step time were longer,the step length was shorter,and the step speed was slower in fall group compared to non-fall group;in addition,the ankle dorsal extension angle at the heel strike phase,maximum ankle plantarflexion angle at the support phase,maximum ankle dorsal extension angle,maximum ankle plantarflexion angle,knee extension angle at the heel strike phase,and the maximum knee extension angle at the support phase were all greater in fall group compared with those of non-fall group;while,the maximum knee flexion angle at the support phase,maximum knee flexion angle,maximum hip extension angle,peak hip flexion moment,peak hip extension moment,peak knee flexion moment,peak knee extension moment,and peak ankle plantarflexion moment were all smaller in fall group compared with those of non-fall group.②The results of logistic regression analysis revealed that a long disease course,frequent stair climbing,severe pain,increased ankle dorsal extension angle at the heel strike phase,increased knee extension angle at the heel strike phase,decreased maximum knee flexion angle,decreased peak knee flexion moment,and decreased peak ankle plantarflexion moment were all the independent risk factors for falls among the aged KOA patients(OR=5.230,95%CI(3.232,7.021),P=0.004;OR=4.367,95%CI(2.648,5.953),P=0.003;OR=4.252,95%CI(2.159,6.231),P=0.003;OR=3.473,95%CI(2.982,4.028),P=0.021;OR=6.977,95%CI(3.667,8.964),P=0.001;OR=3.989,95%CI(1.667,5.264),P=0.010;OR=7.051,95%CI(4.267,8.164),P=0.001;OR=4.675,95%CI(2.563,6.798),P=0.008).The fall risk prediction nomogram model showcased that a total score of 409 points for fall risk and a probability of 71.90%for falls were predicted in the aged KOA patients when the disease course>6 years,frequent stair climbing,WOMAC pain score>48 points,ankle dorsal extension angle>9.25 degrees at the heel strike phase,knee extension angle>2.35 degrees at the heel strike phase,maximum knee flexion angle<45 degrees,peak knee flexion moment<0.65 N·m/kg,and peak ankle plantarflexion moment<0.90 N·m/kg.The results of receiver operating characteristics(ROC)curve analysis showed that the risk forecasting model had high discrimination performance,with the area under ROC curve,sensitivity,and specificity as 0.698(P=0.000),71.43%,and 65.89%,respectively,in training set,and 0.699(P=0.000),78.26%,and 63.18%,respectively,in validation set.The results of Hosmer-Lemeshow goodness-of-fit(GOF)test showed that the model had a good GOF(training set:χ^(2)=0.748,P=0.504;validation set:χ^(2)=1.328,P=0.263).The results of clinical decision curve analysis indicated a high net benefit rate when threshold probability ranged from 9%to 88%in the training set,and 11%to 92%in the validation set.Conclusion:A long disease course,frequent stair climbing,severe pain,increased ankle dorsal extension angle at the heel strike phase,increased knee extension angle at the heel strike phase,decreased maximum knee flexion angle,decreased peak knee flexion moment,and decreased peak ankle plantarflexion moment are all the independent risk factors for falls in the aged KOA patients.The model established based on the above risk factors has a certain applied value in forecasting the risk for falls in the aged KOA patients.