【作者】 万毅；

【导师】 徐勇勇；

【作者基本信息】 第四军医大学 ， 流行病与卫生统计学， 2009， 博士

【摘要】 随着信息化的发展,循证医学证据资源产生、评价与利用的标准化、规范化越来越备受重视。如何为广大临床实践工作者提供便捷高效的医学证据支持,促进循证医学证据信息的评价、利用和资源共享是循证医学实践的关键环节和有待解决的问题。对于医学研究证据信息资源来说,利用本体可以提供一种智能化的检索机制,可以突破传统信息检索技术的弊端,真正满足用户的信息需求和表达习惯,增强检索系统的处理能力,在减轻终端检索用户认知负担的同时,把循证医学证据评价信息从基于关键词的层面提升到基于概念或知识的层面,提高循证医学证据及评价的信息检索和利用,最终实现“智能化”的证据评价信息检索和利用。本文综合运用了文献调查分析、网络信息检索与比较、案例分析、综合归纳、专家访谈等多种研究方法,以不同类型医学研究证据信息资源为实例,对基于本体的知识组织这一课题进行了深入的研究,运用图书馆学、情报学的相关知识,以为用户服务的角度出发,构建具有循证医学特色的基于本体的证据评价知识组织研究体系模型框架。研究通过构建循证医学证据评价的语义模型,研究医学证据评价知识表示方式。在循证医学证据评价基本要求的指导下,应用本体工程原理,对医学证据评价相关项目进行全面的知识体系分析,整理和收集相关的资料信息,并对它们之间的关系进行分析,建立循证医学证据评价的语义模型和系统评价实例,并通过系统评价及相关研究实例对模型的适用性进行初步测试和评价。本研究为循证医学证据评价提供了知识表示的一种规范形式,对广大医学工作者和对医学感兴趣的用户提供了循证医学全面而专业化的证据资源评价描述信息。在模型的知识体系构建中,本研究从医学证据评价的各个关键环节对证据质量和现状进行了全面的分析,同时参考国际通用规范和标准等资料,从研究证据的来源、类型、结构、相关性、科学性和可靠性等各方面确定了不同研究设计类型的医学研究证据评价的知识体系,并以此为基础构建循证医学证据评价的语义模型。知识本体的建立以公认的循证医学证据评价理论为依据,遵循本体开发的基本原则和斯坦福大学医学院本体开发的七步法,采用知识本体编辑工具Protégé构建语义模型的层级结构,定义概念与概念之间的关系以及语义槽的专业词汇和其他元数据的规范描述。医学证据评价本体收纳了常用的证据评价相关指标和项目,并通过对各项指标和项目的应用环境的描述,建立了指标概念之间的逻辑关系。本研究构建了包含证据评价基本要素的语义模型,创建了7个类、27个子类和87个语义槽(属性)和元数据,并实现了模型的结构化和形式化表达。该模型根据不同的研究设计类型又分为以下几个部分:系统评价与Meta分析评估(QUORUM)语义模型;随机对照试验质量评估(CONSORT)语义模型;观察性研究(STROBE)评价语义模型;诊断试验(STARD)标准语义模型;临床指南(AGREE)评估语义模型等。研究还通过系统评价实例的创建对模型重用和完善,同时为临床实践提供了有效证据。系统评价实例以房颤患者抗凝控制指标的选择及其与不良事件的关系为研究问题,具有重要的临床实践意义。研究检索MEDLINE,EMBASE和Cochrane Library截至2008年关于房颤患者口服抗凝剂治疗研究中,报告抗凝控制指标的文献。相关指标包括:在抗凝目标值范围内的时间百分比(percentage of time in therapeutic range,TTR),在抗凝目标值范围内的国际标准化比值百分比(percentage of INRs in range)和临床不良事件(主要有严重出血和血栓栓塞事件)发生率等。从检索到的全部1571篇文献中,根据纳入、排除标准和语义模型证据质量评价结果,入选文献38篇,共计47个研究组。为了使抗凝控制收益最大化,患者需要定期监测和合理调整抗凝治疗以保持国际标准化比值(INR)位于参考值范围之内。结果显示,TTR仅增加7%就可以使严重出血下降1例每百人年。基于本研究的分析结果,抗凝治疗要达到收益的最大化和有效降低不良事件发生率,最好使TTR达到70%～80%的目标。在口服抗凝剂治疗的房颤患者中,TTR和抗凝目标值范围内的INR百分比能有效反映抗凝控制质量。提高TTR和抗凝目标值范围内的INR百分比水平能有效降低严重出血和血栓栓塞不良事件发生率。系统评价实例的研究过程和结果为循证医学证据评价语义模型的改善和适用性评估提供了良好的环境。通过系统评价及相关不同研究类型实例的模型重用,本研究调整和完善了模型的结构和属性,验证了模型的清晰一致性和适用性;同时,为系统评价实例研究过程提供了医学文献证据质量评价的形式化表达和结果推论的参考。本研究循证医学证据评价语义模型的开发过程均根据相应的国际上通用的研究标准或规范来构建,结合实例来完善和评估,具有权威性、清晰性、一致性和适用性;同时,模型设计过程中考虑到了将来修订和发展的需要,设定了相应的类、语义槽和元数据,使模型具有可扩展性能。本研究还对本体技术在循证医学证据评价的知识运用与表达上和在与系统的交互性上进行了实验性的探索。虽然没有达到检索的智能推理,但研究基于医学研究证据评价本体的概念检索,能提供以循证医学证据评价内在联系为基础的检索,提高检索的全面程度,同时通过对检索结果的适当交互控制,提高准确程度。研究结果促进了循证医学证据评价的信息化和循证医学临床实践的开展。综上所述,建立通用的循证医学证据评价的语义模型和基于该模型的各种研究设计证据评价模型,对于医学研究证据信息资源的有效管理和开发利用,对于循证医学从临床问题、证据开发与评价到临床实践应用的转化具有重要意义。更多还原

【Abstract】 With the development of information, the standardization and normalization of evidence resources, evaluation and utilization of evidence-based medicine are more and more being focused upon. One of the key issues to practice evidence-based medicine is how to provide convenient and efficient medical evidence to support clinical practitioners and to promote evaluation, utilization and sharing of resources of medical evidence information.As for evidence resources on medical research information, the use of ontology can provide an intelligent search mechanism and can break through the disadvantages of traditional information retrieval technology which really meet the user’s needs for information and habits of expression. It can enhance ability to deal with retrieval systems, and reduce the terminal retrieve user’s cognitive burden at the same time, which elevate evidence information from keywords-based level to concepts or knowledge-based level. It can also improve the retrieval and use of medical evidence and evaluation information, eventually achieve intelligent information retrieval and use of medical evidence appraisal.We integrated used different research methods, which included literature survey analysis, web information retrieval and comparison, case analysis, synthesis, and expert interviews method. We used different types of information resources for medical research evidence as real examples to study the ontology-based knowledge organization. We used the relevant knowledge of Library and Information Science, from customer service point of view, to construct a knowledge organization model framework of evidence appraisal research system which has the characteristics of evidence-based medicine.We studied on the knowledge expression way for medical evidence appraisal through the construction of evidence appraisal semantic model for evidence-based medicine. The comprehensive demand analysis was performed on the basis of the basic requirements of critical appraisal in evidence-based medicine and under the guidance of the application of ontology engineering principles. Then the related data and information was collected, analysis of the relationship between them was performed and the semantic model and the examples of systematic review were established. A preliminary test and evaluation on applicability of the model was studied through the systematic review and other research examples. The study provided a standardized form for knowledge express of critical appraisal information in evidence-based medicine. It also provided a comprehensive and professional evaluation of the evidence described in information resources for broad masses of medical staff and interested users.During the demand analysis for model, the study comprehensive analyzed the quality and status of medical evidence on every key aspect, taking into account international norms and standards, identified the demands from various aspects, including evidence sources, types, structure, relationship, accuracy and reliability, and built the critical appraisal semantic model.The construction of knowledge ontology was based on the recognized theory of evidence based medicine, followed the basic principles of ontology development and the seven-step method on ontology development provided by School of Medicine, Stanford University, used Protégéontology editing tool to build the hierarchical structure of semantic model, to define the relationship between the concepts, the professional vocabulary of the semantic slots and standard description of other metadata. The critical appraisal ontology of medical evidence included the commonly used indicators and projects, and set up a logical relationship between concepts through the description of the application environment of the indicators and projects. The semantic model contained the basic elements of the critical appraisal of medical evidences. There were 7 categories, 27 sub-categories and 87 semantic slots (properties) and the metadata. The implementation of the model structured and formal expression was accomplished. The model was based on different types of study designs which can be divided into the following sections: The Systematic Review and Meta-analyses (QUORUM) semantic model; The Randomized Controlled Trials (CONSORT) semantic model; The Observational studies in Epidemiology (STROBE) semantic model; The Diagnostic Accuracy (STARD) semantic model; The Clinical guideline (AGREE) semantic model.The study reused and improved the model through the systematic review and other real examples, and at meantime provided an effective evidence for clinical practice. The example of systematic review was focused on the relationship between INR control measures and adverse effects of oral anticoagulation in atrial fibrillation, which had important meaning for clinical practice. We searched MEDLINE, EMBASE and Cochrane Library to 2008 for studies reporting INR control measures as percentage of time in therapeutic range (TTR), percentage of INRs in range and clinical outcomes in patients with AF receiving warfarin therapy. Outcomes analyzed were major haemorrhage and thromboembolic events. From the total 1571 retrieved articles, there were 38 articles meeting the including criteria which included 47 study groups. To obtain optimal benefits of anticoagulation control, individuals need to be maintained within their INR reference range, which requires regular monitoring and appropriate modification of treatment. This study systematically examined the relationship between INR control measurements and the prediction of adverse events in patients with atrial fibrillation on oral anticoagulation. For instance, a small increment in TTR, of 7%, can lead to a reduction in major haemorrhage by 1 event per 100 patient years of treatment. Based on our analysis anticoagulation services ought to aim for a TTR between 70% and 80% to optimize benefits and reduce harms for the patients. TTR and percentage of INRs in range can predict INR control effectively. The results support the relationship between INR control and clinical outcomes of major haemorrhage and thromboembolic events.The process and outcome of systematic review example provided a good environment for the improvement and applicability evaluation of the semantic model. Through the model reuse of different types of studies, the structure and property of the model was adjusted and improved, and the consistency and applicability of the model was verified.The development process of the semantic model was in accordance with corresponding international research standards or norms. The model was also assessed and adjusted with real examples, which made the model have authority, clarity, consistency and applicability. At the same time, the design process of the model took into account the future revision and development requirement, set the corresponding category, semantic slot and metadata, which made the model with expandability.This study also experimentally explored system interoperation of the model. Although there had not achieved intelligent inference retrieval, but the research of ontology-based concept retrieval could provide retrieval based on the intrinsic link of critical appraisal of medical evidences, which improved the overall degree of retrieval, and improved the accuracy through appropriate control on the interaction of search results. The research outcome promoted the informationization of critical appraisal of evidence-based medicine and development of carry out evidence-based practice.On the conclusion, the development of a common critical appraisal semantic model on evidence-based medicine has great importance on effective management, development and utilization of information resources of medical evidences. It also has great importance on the transformation from clinical problem, evidence development and appraisal to clinical practice.更多还原