基于多维项目反应理论计算机自适应模型的颈椎病针灸疗效评价方法

To begin with the research on quantification of the traditional Chinese treatment on cervical spondylosis (CS) patients through patient-reported outcomes (PRO), the project intends to introduce the definitive new instruments, which would improve the reliability, validity, and precision of PRO. Multidimensional item response theory(MIRT) measurement model reflects the relationship between the items and the mental, physical and social dimensions which define "Health" as consisting of ,accordingly MIRT allows the reduction and improvement of items and assembles domains of items which are unidimensional and not excessively redundant. Computerized adaptive testing（CAT） selects the most informative remaining item in a domain until a desired degree of precision and achieves suitable the PRO scale matched with each specific patient, so as to reduce the burden of patients ,save test time and require fewer patients in a trial to achieve the same statistical power obtained..In the analysis of the PRO scale, combined with the traditional Chinese medicine theories about symptomatic and pathology classification，the weights of evaluation each healthy domain tends to be obtained through information entropy, analytic hierarchy process, norm and optimized with particle swarm optimization algorithm，consequently a clinical outcome evaluation model would be established based on fuzzy comprehensive evaluation method and uncertainty knowledge reasoning, with the purpose of providing clinical outcome evaluation of Chinese medicine treatment on CS with more scientific and reasonable measures. With the support of above model, a nonlinear prediction model and the control model of inverse problem between Chinese medicine treatment of CS strategies and clinical curative effect would be further built based on multi-object genetic algorithm and neural network, in order to provide scientific reference for the optimization of the Chinese medicine treatment strategy and the development of acupuncture treatment of CS.

颈椎病是以退行性病理改变为基础的疾患，其多样性、复杂性对诊断提出的更高要求。本研究是在颈椎病患者报告的结局指标（PRO）研究的基础上，提出基于多维项目反应理论的计算机自适应测试模型，分析患者生存质量在心理、生理、社会关系等各领域与项目反应之间的关系，据此筛选，获取更可靠、有效的结局指标项目，实现不同病情患者匹配一套适合其健康状况的PRO量表，从而达到减少受试患者数目、节省测验时间的目的。.在对病人的PRO量表的测试结果分析中，结合中医的症候分类、西医分型，针对不同病患人群，通过信息熵等优化算法获取并优化相关评价指标的权重，确立基于多目标区域遗传算法与神经网络结合的非线性颈椎病针灸治疗策略，建立基于模糊评价和不确定性知识推理结合的临床疗效综合评价模型。从而为颈椎病针灸疗效评价提供更为科学、合理的衡量标准。

本课题针对颈椎病中最常见的神经根型，首先收集大量文献作循证研究以及临床案例经验分析，建立了针灸治疗颈椎病的临床疗效评价条目池；建立专家调研问卷网络移动系统，进行专家意见的收集和反馈，从而条目得到进一步筛选；通过语言调整与五级李克特量化，初步建立针灸治疗神经根型颈椎病临床疗效评价的PRO量表；利用信度分析、效度分析、因子分析、多维项目反应理论以及结构方程分析等多元统计方法和机器学习方法，建立可靠的针灸治疗颈椎病临床疗效评价度量指标体系。接着，通过纸质和电话抽样，利用数据库技术进行量表数据的采集；利用同向化、规范化以及提取主成分特征的方式，进行量表数据的预处理；最终，利用神经网络技术建立有效的针灸治疗颈椎病临床疗效的综合评价模型与预测模型，反应了相应疗效指标、生存质量指标要求在某一期望范围时的中医治疗策略的相应变化，为控制治疗策略与疗效、生存质量的一致性，为了解与把握颈椎病病患的治疗过程与规律提供决策依据。