Supplementary file 4_Protective and risk factors in daily life associated with cognitive decline of older adults.docx

BackgroundCognitive decline is a chronic condition which is characterized by a loss of the ability to remember, learn, and pay attention to complex tasks. Many older people are now suffering from cognitive decline, which decreases life quality and leads to disability. This study aimed to identify the risk and protective factors for cognitive decline of the older people from daily life and establish a predictive model using logistic regression.
MethodsWe investigated 3,790 older people with health examination and questionnaires which included information associated with physical condition, lifestyle factors, and cognitive status. Single-factor comparison, principal component analysis with a Manova-Wilk test, multiple linear regression, and logistic regression were performed to filter the risk and protective factors regarding cognitive decline of older individuals. Then a predictive model using logistic regression was established based on the most significant protective and risk factors.
ResultsWe found a significant separation along the coordinate axis between people with normal and declined cognition by principal component analysis, as confirmed by the Manover-Wilk test. Single-factor comparison, multiple linear regression and logistic regression implied that gender, age, hypertension level, height, dietary habit, physical-exercise duration, physical-exercise history, and smoking history could be closely linked with cognitive decline. We also observed significant differences in height, physical exercise duration, physical-exercise years, and smoking years between the male and female of the participants. ROCs of the predictive model by logistic regression were plotted, with AUC values of 0.683 and 0.682, respectively, for the training and testing sets. Although an effective predictive model is thought to have AUC over 0.7, we still believe that the present model is acceptable because the value is close to the threshold.
ConclusionThe protective factors of cognitive decline for older people were male gender, height, keeping moderate exercising, and nicotine stimulation, and the risk factors included age, female gender, vegetarianism and hypertension. Except for the genetic factor, differences in lifestyle, such as smoking and exercise habits, may contribute to the observed differences in cognitive function between genders. The significant results could be utilized in the practice for the early intervention of cognitive decline in aged people.

背景
认知衰退（Cognitive decline）是一种以记忆、学习及复杂任务注意力维持能力受损为特征的慢性病症。当前众多老年人深受认知衰退困扰，该病症不仅会降低生活质量，还会引发残疾。本研究旨在从日常生活场景中筛选老年人认知衰退的风险与保护因素，并构建逻辑回归（Logistic Regression）预测模型。

方法
本研究纳入3790名老年人，通过健康体检与问卷调查收集其身体状况、生活方式及认知状态相关信息。采用单因素比较、结合威尔克检验（Manova-Wilk Test）的主成分分析（Principal Component Analysis）、多重线性回归与逻辑回归，筛选与老年人认知衰退相关的风险及保护因素。随后基于筛选出的最具显著性的保护与风险因素，构建逻辑回归预测模型。

结果
主成分分析显示，认知正常与认知衰退人群在坐标轴上存在显著分离，该结果经威尔克检验验证。单因素比较、多重线性回归与逻辑回归分析结果表明，性别、年龄、高血压水平、身高、饮食习惯、运动时长、运动史与吸烟史均与老年人认知衰退密切相关。本研究同时发现，受试者中男性与女性在身高、运动时长、运动年限与吸烟年限上存在显著差异。本研究绘制了逻辑回归预测模型的受试者工作特征曲线（Receiver Operating Characteristic Curve, ROC），训练集与测试集的曲线下面积（Area Under Curve, AUC）分别为0.683与0.682。尽管业内通常认为有效的预测模型其AUC值需大于0.7，但本模型的AUC值接近该阈值，因此仍具备应用价值。

结论
老年人认知衰退的保护因素包括男性性别、较高身高、规律中等强度运动与尼古丁暴露（即吸烟），而风险因素则包含年龄增长、女性性别、素食习惯与高血压。除遗传因素外，吸烟与运动习惯等生活方式差异，或可解释不同性别间认知功能的现存差异。本研究的显著发现可用于老年人认知衰退的早期干预临床实践。