Abstract: Objective  To explore the influencing factors of early depressive symptoms after acute ischemic stroke and construct a prediction model.
Methods  A retrospective cohort analysis was conducted. The study included patients with acute ischemic stroke who were admitted to the Affiliated Hospital of Shandong Second Medical University from July 2023 to July 2024, had their first onset of the disease, were diagnosed by cranial MRI, were admitted within 48 hours of onset, had a mini-mental state examination score≥15 points, and had complete clinical and follow-up data. The baseline data, assessment of early depressive symptoms after acute ischemic stroke, laboratory examination, stroke-related scale scores, and imaging data were collected. Based on the Hamilton depression scale (HAMD) scores of outpatient follow-up at (30±3) days after stroke, patients were divided into the depressive symptom group and the non-depressive symptom group. R 4.3.0 software was used to compare the data differences between the two groups. Multivariate logistic regression analysis was used to screen the independent predictors of early depressive symptoms after acute ischemic stroke. Backward stepwise regression was applied to determine the final variables for constructing the prediction model. The model’s performance was evaluated using the ROC curve.
Results  A total of 211 patients with acute ischemic stroke were ultimately included in this study, with 110 patients (52.13%) in the non-depressive symptom group and 101 patients (47.87%) in the depressive symptom group. Compared with the non-depressive symptom group, the depressive symptom group had lower LDL-C and red blood cell count, but higher admission C-reactive protein, admission NIHSS score, mRS score, Hamilton anxiety scale (HAMA) score, HAMD score, and discharge NIHSS score (all P＜0.05). Imaging analysis showed that the depressive symptom group had a higher proportion of multiple infarctions and infarctions in the frontal lobe and thalamus (all P＜0.05). Multivariate logistic regression analysis showed that admission C-reactive protein, admission NIHSS score, mRS score, HAMD score, frontal lobe infarction, and multiple infarctions were independent risk factors for early depressive symptoms after acute ischemic stroke (all P＜0.05). Discharge NIHSS score was an independent protective factor, while red blood cell count and HAMA score were potential protective and risk factors, respectively. The prediction model constructed based on these factors demonstrated an AUC of 0.923 (95%CI 0.888-0.959). The Hosmer-Lemeshow test yielded a χ2 value of 8.244 (P=0.41). At the optimal cutoff value of 0.763 determined by the maximum Youden index, the model achieved a sensitivity of 88.1% and a specificity of 88.2%. 
Conclusions  The admission C-reactive protein, admission NIHSS score, mRS score, HAMD score, discharge NIHSS score, frontal lobe infarction, and multiple infarctions are important influencing factors for early depressive symptoms after acute ischemic stroke. The prediction model constructed based on these factors has good clinical discrimination efficiency.

Key words: Post-stroke depression; Hamilton depression scale; National Institutes of Health stroke scale; Frontal lobe infarction; Prediction model

摘要： 目的 探究急性缺血性卒中后早期抑郁症状的影响因素并构建预测模型。
方法 采用回顾性队列分析，纳入2023年7月—2024年7月山东第二医科大学附属医院首次发病、经头颅MRI检查确诊、发病48 h内入院、简易精神状态检查评分≥15分且相关临床及随访评分完整的急性缺血性卒中患者。收集患者的基线资料、急性缺血性卒中后抑郁症状评估、实验室检查、卒中相关量表评分及影像学资料，以卒中后（30±3）d门诊随访的汉密尔顿抑郁量表（Hamilton depression scale，HAMD）评分为依据，将患者分为抑郁症状组和非抑郁症状组。使用R 4.3.0软件对比两组间资料的差异，采用多因素logistic回归分析筛选急性缺血性卒中后早期抑郁症状的独立预测因子。使用向后逐步回归法确定最终变量构建预测模型，并通过ROC曲线评估模型效能。
结果 本研究最终纳入211例急性缺血性卒中患者，非抑郁症状组110例（52.13%），抑郁症状组101例（47.87%）。与非抑郁症状组相比，抑郁症状组LDL-C和红细胞计数均较低，入院C反应蛋白、入院NIHSS评分、mRS评分、汉密尔顿焦虑量表（Hamilton anxiety scale，HAMA）评分、HAMD评分和出院NIHSS评分均较高（均P＜0.05）。影像学分析显示，抑郁症状组多发梗死的比例及额叶、丘脑梗死比例均更高（均P＜0.05）。多因素logistic回归分析表明，入院C反应蛋白、入院NIHSS评分、mRS评分、HAMD评分、额叶梗死和多发梗死是急性缺血性卒中后早期抑郁症状的独立危险因素（均P＜0.05），出院NIHSS评分为独立保护因素，红细胞计数和HAMA评分则分别是潜在保护因素和潜在危险因素。基于此构建的预测模型AUC为0.923（95%CI 0.888～0.959），Hosmer-Lemeshow检验χ2=8.244，P=0.41。该模型约登指数最大化的截断值为0.763，此时敏感度为88.1%，特异度为88.2%。
结论 入院C反应蛋白、入院NIHSS评分、mRS评分、HAMD评分、出院NIHSS评分、额叶梗死和多发梗死是急性缺血性卒中后早期抑郁症状的重要影响因素，基于此构建的预测模型具有良好的临床判别效能。

关键词: 卒中后抑郁; 汉密尔顿抑郁量表; 美国国立卫生研究院卒中量表; 额叶梗死; 预测模型