Institutional Agent Based Models for Healthcare Applications (Simstitution)

我的研究计划的目标是进一步发展基于主体的建模（ABM）技术，结合数据勘探/挖掘/融合。进一步的目标是用真实数据验证abm——作为一种引人注目的方法，医疗保健建模是主要的应用领域。关于数据挖掘的许多想法都是新颖的，因为由于跟踪技术和联机信息的存在以及对这些来源作为ABMs输入的承认，数据变得越来越容易获得。ABM需要在提供数据挖掘机会的技术方面具有强大的背景，了解ABM及其好处，以及统计推断。我帮助开发了用于社区级流行病建模的ABMs，以及用于评估急诊科患者准入政策的机构级模型，以及医院感染和流感样疾病的建模。该应用程序继续开发在机构内使用的ABMs。从长远来看，我正在积极地进行研究，用不同的经验数据来增强ABM。一个原型系统正在开发中，使用智能手机作为近距离定位系统，为基于代理的平台提供真实数据，以改善ED的操作。在第二个例子中，我们正在使用急诊科信息系统数据，作为我们社区和机构一级基于代理的模型的经验输入。在第三个例子中，我将使用gpu来提高我们正在开发的ABMs的性能。我的实验室正在开发这套建模工具，目前有四名研究生和一名暑期学生参与其中。迄今为止的努力表明，HQP培训的机会是巨大的。通过将重点放在数据上，本研究保留了对计算机工程的特殊亲和力，无论是在效用上还是在计算机工程对该研究领域做出独特而重要贡献的潜力上。虽然概述的方法非常适合计算机工程，但所选择的医疗保健建模应用领域也特别重要，因为这些是具有重大公共卫生和卫生政策利益的未解决问题。

The objective of my research program is to further develop agent-based modeling (ABM) techniques in combination with data prospecting/mining/fusion. A further objective is to validate ABMs with real data - as a compelling methodology with healthcare modeling as the main application area. Many of the ideas for data mining are novel, in that data is becoming more readily available as a consequence of the presence of tracking technologies and on-line information as well as the recognition of these sources as input into ABMs. ABMs require a strong background in technologies that offer data mining opportunities, understanding of ABM and its benefits, and of statistical inference. I have helped develop ABMs for community-level epidemic modeling, as well as institution-level models for assessing patient access policy in EDs and modeling of nosocomial infection and influenza-like-illness. This application continues development of ABMs for use within an institution. Over the longer term, I am actively pursuing research to augment ABM with diverse empirical data. A prototype system is being developed, using Smartphones as proximity location systems to provide real data to an agent-based platform for improving ED operations. In a second instance, we are working with Emergency Department Information System data to be used as empirical inputs to both our community- and institution-level agent based models. In a third instance I will be using GPUs to improve the performance of the ABMs we are developing. This suite of modeling tools is under development in my lab, and currently engages four graduate students and one summer student. Efforts to date have demonstrated that the opportunities for the training of HQP are significant. By placing an emphasis on data, this research retains a particular affinity to computer engineering, both in utility and in the potential for computer engineering to make unique and significant contributions to this research area. While the outlined approach is highly suited to computer engineering, the chosen application area of healthcare modeling is also of particular significance as these are unsolved problems with substantial public health and health policy benefits.