医学信息挖掘和统计分析
随着数字化医疗和个性化医疗在卫生领域的普及，医疗数据在医院管理中的作用也随之提高，从现代医院管理战略出发，医院的管理逐渐趋向科学化，精细化，而科学化精细化的管理主要体现在医院借助信息系统对临床诊疗、科研、医院管理方面数据的整合和利用，数据共享成为可能，也使得健康医疗机构积累起海量健康数据。如何从这些复杂的信息中提取出有价值的信息成为必须要解决的问题之一。最常用的数据挖掘方法有关联规则、决策树、人工神经网络和聚类分析，根据实际情况选择适当的数据挖掘方法能够更好地揭示医疗数据间的关系和规律，通过疾病间的关联规则、药物的相互作用、基因测序和患者的病史等上下文环境对疾病的发展过程、阶段以及治疗效果进行预测，同时也可根据现实情况进行治疗方案的及时调整，达到疾病的精准诊断、精准治疗，实现患者病情的个性化治疗的目的。
医疗行业中的数据具有自身独特性，涉密性、增量性、偏差性是医疗数据自身所带的特点。由于医院收集患者就医数据不可避免的会涉及患者隐私信息，因此数据具有涉密行；医疗业务中的图片数据、影像数据、文字数据等，一旦生成不能删除，因此具有增量性；人工在录入医疗业务数据时难免出现误差，此类数据与正式数据产生冲突，因此医疗数据具有偏差性。对涉密的，增量的，偏差的医疗数据，采用统计学技术对医疗数据挖掘的热点方向和医疗相关领域进行分析，可以促使其有效应用于诊断、治疗及预后评估等医疗实践中的各环节，有助于从海量数据中提取有价值的知识和规则，从而为疾病的诊断和治疗、医院的决策管理和科研服务提供科学合理的依据，为高效分析、利用医疗数据提供新的方法。
医学信息挖掘与统计特色学科是将传统的模拟式数据转化为数字式数据的工具和桥梁，对高维度，非线性，非高斯的数据，采用统计挖掘方法，可以提供更好的预测精度。医学信息挖掘与统计特色学科的驱动方式是从假设型驱动的科学问题思维方式变为数据型驱动的挖掘问题方式，目前面对的主要问题有医学数据的多样化、分析过程的复杂化和标准体系不完善。通过对医院整体信息资源的整合，在为信息建设的更高层面的建设，医院收益颇多。
第一、更好的了解医院的现有资源情况，特别是对病人与病种的分析，实现医院以患者为核心的医疗管理和服务方式。
第二、在更好的了解医院的基本状况和优势下，对医院的市场定位和实现该市场定位做了很多的决策支持作用。
第三、医院信息孤岛的消除，管理者可以在一个平台看所有数据，并且数据间相互关联，相互为对方揭示原因与结果。实现真正的全局化管理和自上而下的监控。
第四、建立起医院的综合评价系统，对人员、科室的绩效考核轻松而客观，留住好的人力资源，增加医院的竞争实力。
第五、提高医疗质量的监督和水平，从而帮助提高医院知名度。还可以未雨绸缪地做一些管理的准备工作。
With the popularity of digital health care and customized health care in health field, heath care data is playing a more and more important role in hospital management. Modern hospital management strategy emphasizes on scientific and detailed management, which is demonstrated on the fact that hospitals are using information system to integrate and apply the clinical diagnosis and treatment data, research and development data, hospital management data to enable sharing of data and accumulation of tremendous health data for health and medical care organizations. One of the critical problems is how to extract valuable information from this complicated information pool. The most common digital extraction methods include: rule of association, decision tree, artificial nerve network, and cluster analysis. Choosing a proper digital extraction method based on actual situation will help us to reveal the relation and law of health care data, to forecast disease development process, stages and treatment effects with the reference of the context including association rule between diseases, interaction of medicines, gene sequencing, and disease history of the patient. It will also help to adjust the treatment plan according to actual situation in a timely manner, allowing for accurate diagnose and treatment of diseases and achieving the goal of customized treatment for patients.
Health care data has its own uniqueness, and it is confidential, incremental and deflective. It is confidential because hospitals are collecting data of the patients, which inevitably involves the patent’s privacy. It is incremental because none of the picture, video or literal data can be deleted once generated. And it is deflective because there is human error when entering the medical care business data, which is a conflict of formal data. In regard to the confidential, incremental and deflective medical data, we employ statistical technology to analyze the medical care data extraction trend and relevant field so that the data will be properly used in medical care processes such as diagnose, treatment, and prognosis evaluation. It will eventually assist us to extract valuable knowledge and rules from the tremendous data pool, provide scientific and logical basis for diagnose and treatment of disease, decision management in hospitals and research and development services, as well as new methods for efficient analysis and application of medical care data.
The medical care information extraction and data collection discipline is a tool and a bridge to transmit traditional simulative data into digital data. It employs statistical extraction method to deal with high dimensional data, non-liner data and non-Gaussian data, enabling better and more accurate forecast. The driving engine of the medical care information extraction and data collection discipline works with the transmission from assumption-driven scientific way of thinking to data-driven problem extraction method, and the challenges are diversity of medical care data, and complicity of the analysis process, and incomplete standard system. The hospital will benefit from it in regard to integration of hospital overall information resources and higher level establishment of information.
First, with better understanding of the hospital existing resources, especially analysis of patients and diseases, the hospital will achieve the goal of patient-oriented medical care management and service method.
Second, with better understanding of the hospital basis and advantages, the hospital has more supporting information for decision-making regarding its market positioning and achievement of the position.
Third, the removal of isolated hospital information enables managers to review all the data, association between data, and causal and consequence effect of data on a platform, fulfilling actual overall management and supervision from top to bottom.
Fourth, establish a hospital overall evaluation system to make personnel and departmental performance review easy and objective, keeping good human resources and enhancing the hospital competitive strength.
Fifth, Improve supervision of the hospital quality and improve the hospital popularity. It will also make preparation for future management in advance.
