高血压合并脑梗死老年患者潜在用药安全风险及Lasso-logistic回归分析

Potential medication safety risks and Lasso-logistic regression analysis in hypertension complicated with cerebral infarction elderly patients

目的探讨高血压合并脑梗死老年患者潜在不适当用药(PIM)情况并建立影响因素Lasso-logistic回归列线图模型。方法纳入2020年1月至12月高血压合并脑梗死多重用药治疗的出院患者105例,以《中国老年人潜在不适当用药判断标准》评价PIM情况,Lasso-logistic回归分析PIM风险影响因素,建立PIM风险影响因素列线图模型,评价模型区分度和准确度。结果105例患者PIM为85例(81.0%),涉及26种药品、184例次,排第1位是氯吡格雷48例次(26.1%)。Lasso-logistic回归提示,用药种数(P=0.003,OR=1.412,95%CI:0.582~2.868)、糖尿病(P<0.001,OR=0.081,95%CI:-3.899~-1.129)对PIM影响显著。ROC曲线下面积(AUC)为85.3%(95%CI:0.772~0.933),提示模型精准度和区分度良好;Bootstrap法(重采样次数B=1000)内部验证提示模型预测能力一般。结论基于Lasso-logistic回归构建的PIM风险影响因素列线图能较好识别PIM风险,高血压合并脑梗死多重用药治疗住院老年患者PIM风险高,有必要减少用药种数以降低PIM风险。

AIM To investigate the potentially inappropriate medications(PIM)in elderly patients with hypertension combined with cerebral infarction and to develop a nomogram model based on Lasso-logistic regression of the influencing factors.METHODS A total of 105 patients with hypertension combined with cerebral infarction and multidrug therapy discharged from January to December 2020 were included,and the PIM was evaluated by the criteria of potentially inappropriate medications for older adults in China.The PIM influencing factors were analyzed by Lasso-logistic regression,Nomogram model of factors influencing the PIM risk was developed,and the discrimination and accuracy of the model were evaluated.RESULTS The incidence of PIM in 105 patients of hypertension complicated with cerebral infarction was 85 cases(81.0%)and 184 times,involving 26 drugs.The 1st drug was clopidogrel with 48 times(26.1%).A statistically significant effect of the number of drugs(P=0.003,OR=1.412,95%CI:0.582 to 2.868)and diabetes(P<0.001,OR=0.081,95%CI:-3.899 to-1.129)on PIM was suggested by the Lasso-logistic regression.Of the model,a good accuracy and discrimination was suggested by the area under the ROC curve(AUC),which was 85.3%(95%CI:0.772 to 0.933),and an acceptable predictive power was suggested by internal validation with Bootstrap method(B=1000).CONCLUSION The nomogram model based on Lasso-logistic regression can better identify the risk of PIM.Elderly patients with hypertension complicated with cerebral infarction and multi-drug therapy have a high risk of PIM,and it is necessary to reduce the number of medications to reduce the risk of PIM.