摘要
目的 基于Lasso-Logistic回归分析构建帕金森病患者用药依从性的评分模型。方法 选取2023年7-12月于安徽医科大学附属六安医院慢病门诊开具抗帕金森病药物的290例患者作为研究对象,分为训练集和验证集,收集其临床资料,采用单因素和多因素Lasso-Logistic回归分析筛选帕金森病患者用药依从性的影响因素,并建立评分模型。结果 290例患者中,依从性差的有99例,发生率为34.14%。年龄≥60岁(OR=6.238,95%CI:3.322~11.713)、个人月收入<2000元(OR=2.174,95%CI:1.007~4.691)、药品种类≥5种(OR=3.182,95%CI:1.501~6.745)、有药物不良反应(OR=4.030,95%CI:1.040~15.617)和社会支持量表评分<23分(OR=3.087,95%CI:1.581~6.029)为帕金森病患者用药依从性差的独立危险因素(P<0.05),接受用药宣教(OR=0.125,95%CI:0.031~0.513)为其保护因素(P<0.05)。基于上述因素构建预测模型:Logit(P)=-7.318+年龄×1.831+个人月收入×0.776+药品种类×1.157-接受用药宣教×2.078+药物不良反应×1.394+社会支持量表评分×1.127。模型验证结果显示,训练集和验证集的一致性指数(C-index)为0.945(95%CI:0.928~0.962)和0.897(95%CI:0.865~0.929);训练集和验证集的校准曲线均趋近于理想曲线,Hosmer-Lemeshow检验结果分别为χ^(2)=4.013、P=0.856和χ^(2)=3.303、P=0.770;训练集和验证集的ROC曲线下面积(AUC)为0.931(95%CI:0.897~0.966)和0.812(95%CI:0.719~0.904);决策曲线显示训练集和验证集的阈值概率分别为4%~100%和1%~75%。结论帕金森病患者用药依从性的影响因素较多,本次构建的评分模型对帕金森病患者用药依从性具有良好的预测能力。
Objective To construct a scoring model of medication adherence in Parkinson's disease patients based on Lasso-Logistic regression analysis.Methods A total of 290 patients who were prescribed anti-Parkinsonian drugs in the chronic disease clinic of Lu'an Hospital Affiliated to Anhui Medical University from July to December 2023 were selected as the study subjects,they were divided into the training set and the validation set,and their clinical data were collected;single-factor and multifactorial Lasso-Logistic regression analysis were used to screen for the influencing factors of medication adherence in Parkinson's disease patients and to establish a scoring model.Results Among the 290 patients,99 had poor adherence,with an incidence of 34.14%.Age≥60 years(OR=6.238,95%CI:3.322~11.713),personal monthly income<2000 yuan(OR=2.174,95%CI:1.007~4.691),type of medication≥5(OR=3.182,95%CI:1.501~6.745),adverse drug reaction(OR=4.030.95%CI:1.040~15.617)and social support scale score<23(OR=3.087,95%CI:1.581~6.029)were independent risk factors for poor medication adherence in Parkinson's disease patients(P<0.05),and receiving medication education(OR=0.125,95%CI:0.031~0.513)was a protective factor(P<0.05).A prediction model was constructed based on the above factors:Logit(P)=-7.318+age×1.831+monthly personal income×0.776+type of drug×1.157-receiving medication education×2.078+adverse drug reactions×1.394+score on the social support scale×1.127.The results of the model validation showed that the consistency index(C-index)of the training and the validation sets were 0.945(95%CI:0.928~0.962)and 0.897(95%CI:0.865~0.929),respectively;the calibration curves of the training and validation sets converged to the ideal curves,and the results of the Hosmer-Lemeshow test wereχ^(2)=4.013(P=0.856)andχ^(2)=3.303(P=0.770);the area under the ROC curve(AUC)of the training and validation sets were 0.931(95%CI:0.897~0.966)and 0.812(95%CI:0.719~0.904),respectively;the decision curves showed that the threshold probabilities of the train
作者
李鹏飞
何春远
李增
Li Pengfei;He Chunyuan;Li Zeng(School of Pharmacy,Anhui Medical University,Hefei 230032,China;Department of Pharmacy,Lu'an Hospital Affiliated to Anhui Medical University,Lu'an 237005,China)
出处
《实用药物与临床》
CAS
2024年第12期881-887,共7页
Practical Pharmacy and Clinical Remedies
基金
安徽省自然科学基金项目(2208085MH281)。
