基于图模型的脑卒中功能变量关系结构探析

关键词：脑卒中  国际功能、残疾和健康分类  图建模  网络分析

Keywords:stroke  international classification of functioning, disability and health  graph modeling  network analysis

摘要：

      摘要 目的：为国际功能、残疾和健康分类（ICF）在脑卒中康复中的应用提供新思路和新依据。 方法：选取100例脑卒中患者组成的便利样本，访谈法评定国际版脑卒中ICF综合核心组合的166个条目。以条目为节点，条目间的风险相关性为连线，构建图模型。采用R软件（3.2.2版）建模，用Pajek64（5.02版）进行网络分析和可视化。 结果：在图模型总体内，存在主组元结构，并且可以从中提取具有稳定路径结构的3-核结构。主组元中还可以提取具有牢固连接关系的主岛屿结构。其中“d450步行”既属于通用组合，又属于简要核心组合，并且占据了重要位置。简要核心组合条目在总图中的子网络，以及这些条目在各级限定值上的频数分布，可以提供相互参照和相互补充的功能变量知识。 结论：ICF综合核心组合的图模型，可反映脑卒中功能变量间的复杂关系结构。该组合的频数分布，可反映卒中功能变量的属性特征。从关系结构和属性特征两个不同视角进行分析，可以为ICF的康复应用提供综合策略。

Abstract:

      Abstract Objective: To provide new thinking and evidence about applications of International Classification of Functioning, Disability and Health (ICF) for persons with stroke. Method: A convenient sample of 100 stroke persons participated in this study. Investigation included a set of 166 ICF categories in the comprehensive core set for stroke. Graphical model was composed of categories as nodes and risk correlations as lines. The study used R software (Version 2.2.2) for graphical modeling and Pajek64 (version 5.02) for network analysis and visualization. Result: There is a main component in the global structure of graphical model. The main component contains a 3-core with robust connectivity and a main island with solid connections. In the main island, a critically localized category, d450 (Walking), belongs to both generic core set and brief core set. A network of brief core set was extracted from the graphical model. For the same set of categories, this study mapped their frequency distribution on ICF scores. By comparing the brief core set network and their frequency distribution, complementary knowledge of functional variables emerged from the findings. Conclusion: Graphical model of comprehensive core set for stroke can reveal complex relational features embedded in functional variables. Frequency distribution of the same set makes it possible to interpret attribute features of functional variables. By analyzing different perspectives from relational and property data, these techniques may offer an integrated strategy for the applications of ICF.