摘要

[摘要]
目的 探讨老年抑郁症患者并发认知障碍的风险因素，并构建预测模型。

方法 回顾性分析2023年7月至2025年7月期间，于金华市第二医院接受治疗的154例老年抑郁症患者的临床资料。根据蒙特利尔认知评估量表，将其评分在26分以下的患者归于认知障碍组（n=52），反之归于非认知障碍组（n=102）。收集患者的临床资料、神经影像学检查资料及血清学指标。通过单因素和多因素Logistic回归分析，筛选出老年抑郁症患者并发认知障碍的独立预测因素。使用R软件包绘制列线图模型，采用ROC曲线、校准曲线评估模型的预测效能。

结果 Logistic回归分析显示年龄、受教育年限、汉密尔顿抑郁量表评分、睡眠质量、海马体积、皮层厚度、血红蛋白是老年抑郁症患者并发认知障碍的独立预测因素。基于上述因素建立列线图预测模型，ROC的曲线下面积为0.880，且模型的校准曲线与理想曲线相近。

结论 基于年龄、受教育年限、汉密尔顿抑郁量表评分、睡眠质量、海马体积、皮层厚度和血红蛋白构建了老年抑郁症患者并发认知障碍风险的预测模型，且该模型具有良好的预测性能。

关键词：[关键词] 抑郁症；认知障碍；老年；预测模型

ABSTRACT

[Abstract]
Objective To investigate the risk factors associated with cognitive impairment in elderly individuals diagnosed with depression and to develop a predictive model.

Methods The study involved a retrospective review of clinical data from 154 elderly patients with depression, all of whom were treated at the Second Hospital of Jinhua between July 2023 and July 2025. Patients were subsequently stratified into two groups based on their Montreal Cognitive Assessment scores: a cognitive impairment group (n=52) comprising individuals with scores below 26, and a non-cognitive impairment group (n=102). The clinical data, neuroimaging examination data and serological indicators of patients were collected. Independent predictors of concurrent cognitive impairment in elderly individuals with depression were determined through univariate and multivariate logistic regression. Subsequently, a nomogram model was developed using the R software package, and its predictive accuracy was evaluated via ROC curve analysis and calibration curve assessment.

Results The results of logistic regression analysis indicated that age, years of education, Hamilton Depression Scale score, sleep quality, hippocampal volume, cortical thickness, and hemoglobin were independently associated with cognitive impairment in elderly individuals with depression. These factors were then incorporated into a nomogram prediction model. The model exhibited excellent discrimination, as evidenced by an area under the ROC curve of 0.880, and demonstrated good calibration, with the calibration curve closely approximating the ideal curve.

Conclusion   A nomogram prediction model for identifying elderly patients with depression at risk of concurrent cognitive impairment was constructed, utilizing age, years of education, Hamilton Depression Scale score, sleep quality, hippocampal volume, cortical thickness, and hemoglobin as key variables. The model exhibits promising predictive capabilities.

Key words: [Key words] Depression; Cognitive impairment; Elderly; Predictive model