Househ Mowafa, Aldosari Bakheet
Department of Health Informatics, College of Public Health & Health Informatics, King Saud Bin Abdul-Aziz University for Health Sciences, Riyadh, Saudi Arabia.
Stud Health Technol Inform. 2017;238:80-83.
From the mid-1990s, data mining methods have been used to explore and find patterns and relationships in healthcare data. During the 1990s and early 2000's, data mining was a topic of great interest to healthcare researchers, as data mining showed some promise in the use of its predictive techniques to help model the healthcare system and improve the delivery of healthcare services. However, it was soon discovered that mining healthcare data had many challenges relating to the veracity of healthcare data and limitations around predictive modelling leading to failures of data mining projects. As the Big Data movement has gained momentum over the past few years, there has been a reemergence of interest in the use of data mining techniques and methods to analyze healthcare generated Big Data. Much has been written on the positive impacts of data mining on healthcare practice relating to issues of best practice, fraud detection, chronic disease management, and general healthcare decision making. Little has been written about the limitations and challenges of data mining use in healthcare. In this review paper, we explore some of the limitations and challenges in the use of data mining techniques in healthcare. Our results show that the limitations of data mining in healthcare include reliability of medical data, data sharing between healthcare organizations, inappropriate modelling leading to inaccurate predictions. We conclude that there are many pitfalls in the use of data mining in healthcare and more work is needed to show evidence of its utility in facilitating healthcare decision-making for healthcare providers, managers, and policy makers and more evidence is needed on data mining's overall impact on healthcare services and patient care.
从20世纪90年代中期开始，数据挖掘方法就被用于探索和发现医疗保健数据中的模式及关系。在20世纪90年代和21世纪初，数据挖掘是医疗保健研究人员非常感兴趣的一个话题，因为数据挖掘在运用其预测技术来帮助构建医疗保健系统模型和改善医疗保健服务的提供方面展现出了一些前景。然而，人们很快发现挖掘医疗保健数据存在许多与医疗保健数据的准确性相关的挑战，以及预测建模方面的局限性，这导致了数据挖掘项目的失败。随着大数据运动在过去几年中蓬勃发展，人们对使用数据挖掘技术和方法来分析医疗保健领域产生的大数据的兴趣再度兴起。关于数据挖掘对医疗保健实践在最佳实践、欺诈检测、慢性病管理和一般医疗保健决策等问题上的积极影响，已经有很多论述。但关于数据挖掘在医疗保健领域应用的局限性和挑战的论述却很少。在这篇综述论文中，我们探讨了在医疗保健领域使用数据挖掘技术时的一些局限性和挑战。我们的研究结果表明，数据挖掘在医疗保健领域的局限性包括医疗数据的可靠性、医疗保健组织之间的数据共享、不恰当的建模导致预测不准确。我们得出结论，在医疗保健领域使用数据挖掘存在许多陷阱，需要更多工作来证明其在促进医疗保健提供者、管理人员和政策制定者进行医疗保健决策方面的效用，并且需要更多证据来证明数据挖掘对医疗保健服务和患者护理的总体影响。
