Methods for Development of Optimized Complex Expert Systems

专家系统（即模拟人类专家决策能力的计算机系统）可以支持许多领域的决策，包括医学，金融和航空。在我们的数字时代，收集的数据比以往任何时候都多。对于试图管理复杂（多种）慢性疾病的临床医生和患者来说，这在医疗保健中尤其如此。新的健康数据来源，包括使用家用医疗设备进行的连续生理测量，现在可以作为决策的因素。此外，现在可以利用计算机访问实验室结果和药物列表，用于患者和临床医生的专家系统。虽然所有这些数据都可能是相关的，但人们只能吸收有限的信息用于决策。报告的研究表明，支持临床决策的专家系统可以提高医疗服务提供者的性能，特别是用于诊断的目的，但他们的使用与患者的研究很少。此外，还没有很好地研究开发专家系统的过程，该专家系统结合了患者数据的各种和连续的来源，例如来自患者家中的生理数据。这项研究是及时的，不仅因为大量新数据源的激增，而且因为最近意识到患有多种慢性病的人的自我护理不良和次优临床管理的重大影响（5%的人口消耗> 50%的医疗保健资金）。然而，在医疗保健中使用的专家系统必须在确保安全性方面达到特别高的严格水平，因为专家系统的不适当建议可能会产生严重的负面后果。这项发现补助金将支持研究调查和应用专家系统进行复杂的决策，其中包括新兴的数据来源。这将包括调查：1）方法来编程和可视化复杂的专家系统2）不同类型的逻辑在专家系统中使用的适用性，3）嵌入用户偏好到专家系统的安全性和实用性，以及4）的作用和方法，将新的数据源集成到专家系统的慢性病管理。特别是，这项发现补助金将资助研究生研究三个具体项目：1）调查开发优化的患者专家系统的过程，2）调查开发优化的临床医生专家系统的过程，以及3）确定如何将患者健康记录（PHR）数据纳入专家系统。 ** 许多过程、技术和见解可以应用于其他领域。例如，可以开发专家系统，将个人支出、储蓄和投资方面不断增长的可用数据纳入其中，以支持他们的财务决策。未来计划的研究包括调查日益复杂的医疗保健专家系统（即，纳入其他数据源，如基因组学）和人工智能的其他领域（例如，预测健康结果或急性事件）。创建优化专家系统的过程可以帮助管理许多部门的“大数据挑战”。

Expert systems (ie, computer systems that emulate the decision-making ability of a human expert) can support decision-making in numerous fields including medicine, finance, and aviation. In our digital age, more data is being collected than ever before. This is particularly true in healthcare for both clinicians and patients who try to manage complex (multiple) chronic conditions. New sources of health data, including continuous physiological measurements taken with home medical devices can now be factored in for decision-making. In addition, computerized access to lab results and medication lists can now be leveraged for use in expert systems for both patients and clinicians. While all these data may be relevant, people have the ability to assimilate only a limited amount of information for decision-making.**Reported studies have indicated that expert systems supporting clinical decision-making can improve the performance of healthcare providers especially for purposes of diagnosing, but research on their use with patients are few. In addition, the processes to develop expert systems that incorporate varied and continuous sources of patient data, such as physiological data from the patient's home, have not been well studied. This research is timely not only because of the proliferation of novel and large data sources, but also because of the recent awareness of the significant impact of poor self-care and suboptimal clinical management of people with multiple chronic conditions (5% of the population consumes > 50% of all dollars devoted to healthcare). However, expert systems used in healthcare must meet an especially high level of rigor in terms of ensuring safety, due to the potential serious negative consequences of inappropriate recommendations from the expert systems.**This Discovery Grant would support research investigating and applying expert systems for complex decision-making that incorporates emerging sources of data. This will include investigation into: 1) methods to program and visualize complex expert systems 2) the applicability of different types of logic to be used in expert systems, 3) the safety and utility of embedding user preferences into expert systems, and 4) the role and methods to integrate novel data sources into expert systems for chronic disease management. In particular, this Discovery Grant will fund graduate students to research three specific projects:1) investigate the processes to develop optimized patient expert systems, 2) investigate the processes to develop optimized clinician expert systems, and 3) determine how to incorporate patient health record (PHR) data into expert systems. **Many of the processes, techniques, and insights could be applied to other fields. For example, expert systems could be developed that incorporate the growing available data on an individual's spending, savings, and investments to support their financial decision-making. Future planned research includes investigation into increasingly complex healthcare expert systems (ie, incorporating additional sources of data such as genomics) and other areas of artificial intelligence (eg, predictions of health outcomes or acute events). A process to create optimized expert systems can help manage the "big data challenge" in many sectors.