基于缺血性脑卒中患者日常生活活动能力影响因素的列线图模型构建及验证

关键词：缺血性脑卒中  日常生活活动能力  预测模型  列线图

Keywords:ischemic stroke  activity of daily living  prediction model  nomogram

摘要：

      摘要 目的：探索缺血性脑卒中（IS）患者日常生活活动（ADL）能力障碍发生的影响因素，并构建和验证ADL障碍发生的列线图风险预测模型，为临床尽早实施有效的康复干预提供依据。 方法：以2019年1月至2022年12月在广东省中医院康复科住院的401例IS患者为研究对象，收集可能影响患者ADL障碍发生的临床资料，按照7∶3随机将患者划分为训练集280例与验证集121例。采用单因素和多因素Logistic回归分析IS患者ADL障碍发生的影响因素，应用列线图展示ADL障碍发生的预测风险。采用受试者工作特征曲线下面积（AUC）、特异度、灵敏度及校准曲线分析预测模型的性能。 结果：训练集多因素Logistic回归分析结果显示，年龄、糖尿病、美国国立卫生研究院卒中量表（NIHSS）评分、简化Fugl-Meyer 运动功能评定量表（FMA）评分、血红蛋白（Hb）、血小板计数（PLT）、高密度脂蛋白（HDL）是IS患者ADL障碍发生的独立影响因素。以上述7个变量构建列线图，验证集模型AUC、灵敏度、特异度分别为0.875、82.8%、75.4%。校准曲线显示模型预测概率与实际一致性较好。 结论：本研究构建的列线图风险预测模型可有效预测IS患者ADL障碍发生概率，有利于医务人员尽早开展康复防治干预，减少ADL障碍发生。

Abstract:

      Abstract Objective： To construct and verify a nomogram prediction model basing on the influencing factors of the activity of daily living （ADL） disorders in ischemic stroke （IS），providing evidences for the early clinical implementation of efficient rehabilitation therapies. Method： We retrospectively collected clinical data of 401 IS patients hospitalized in the department of Rehabilitation of Guangdong Provincial Hospital of Chinese Medicine from January 2019 to December 2022，focusing on factors that may influence their ADL. These patients were randomly divided into the training （280 cases） and the validation （121 cases） sets according to 7∶3. We used univariate and multivariate logistic regression to analyze the influencing factors for the occurrence of ADL disorders in IS patients， and established the risk prediction model for the occurrence of ADL disorders by a visualized nomogram. The performance of this prediction model was assessed by the area under the curve（AUC），specificity，sensitivity， and calibration curve. Result： The results of multivariate logistic regression analysis in the training sets showed that age， diabetes mellitus， NIHSS score， FMA score， Hb， PLT， and HDL were independent influencing factors for the occurrence of ADL disorders in IS patients. The nomogram model was constructed with the above 7 factors. The AUC， specificity， and sensitivity of this model were 0.875， 82.8%， and 75.4% in the validation set， respectively， indicating that the model has a good discriminative ability. The calibration curve showed that the model agrees well with the actual predicted probability. Conclusion： The nomogram risk prediction model constructed in this study can effectively predict the probability of ADL disorder in IS patients，aiding medical staff in implementing early rehabilitation intervention as early as possible and reduce the occurrence of ADL disorder.