Developing Data Science Approaches to Improve Paediatric Critical Care Patient Flows and its Related Health Economics Benefits in Scotland

儿科危重护理(PCC)是为危重儿童提供医疗保健的一项基本服务。在苏格兰，PCC服务由两个儿科重症监护病房(PCCU)提供；一个在爱丁堡，一个在格拉斯哥。近年来，随着PCCU床位危机的报道，对PCCU床位的需求一直在增加。在重症监护病房用完可用床位的情况下，患者必须被转移到有可用床位的病房，或者紧急护理必须被推迟，直接影响到患者的护理。对PCCU病床需求增加的原因包括季节性呼吸道感染、病人出院延迟以及PCC病人特征的变化。病床使用率和人手的数据通常是从PCCU收集的，以及个别病人的水平数据，包括入院、出院和生理数据。这为开发强大的数据驱动工具提供了独特的机会，这些工具可用于优化和改善PCC中的患者流程，并最终改善患者护理。通过PCCU模拟患者流动的能力可以提供对运营瓶颈和低效的宝贵见解。医疗保健中的离散事件仿真(DES)、系统动力学(SD)和基于代理的系统(ABS)等仿真模型已经建立得很好，一旦开发出来就具有公认的好处[1][2][3]。与通常严重依赖专家意见作为改进方法的其他方法不同，准确和具有代表性的模拟提供了在集成到医院系统之前测试可能的改进方案的机会。近年来，机器学习(ML)与大数据集的可获得性和计算性能的提高相结合，给许多领域带来了革命性的变化。在医疗保健领域，这一点也不例外，使用ML预测患者护理轨迹成为一个日益活跃的研究领域[4][5]。随着PCCU在单个患者层面上常规获得的丰富数据，开发机器学习模型可以根据患者的生理表型预测患者可能如何过渡到医院的护理轨迹，从而可以用于规划未来的护理和改善患者流。这个项目将开发一个模拟模型，可以用于模拟PCCU中的患者流，识别瓶颈，然后通过模拟建议的场景来测试改进方案。其次，将开发机器学习方法，允许从常规收集的临床数据中预测患者护理轨迹，以帮助规划患者护理。最后，它将包括彻底的卫生经济学分析，以充分量化这种数据驱动的方法所带来的影响。参考文献[1]库苏姆·S·马修斯和ELISA F·龙。改善重症监护病人流量和床位使用的概念框架。[2]赛义德·莫希丁、约翰·巴斯比、耶琳娜·萨沃维奇、艾莉森·理查兹、凯特·诺斯通、威廉·霍林沃斯、珍妮·L·多诺万和克里斯托斯·瓦西拉基斯。英国急诊科内的病人流动：计算机模拟模拟方法使用的系统回顾。《英国医学杂志》，7(5)，2017年。[3]爱德华多·卡布雷拉、马内尔·塔博阿达、马路易莎·伊格莱西亚斯、弗朗西斯科·埃佩尔德和埃米利奥·卢克。基于代理的模拟优化医疗急救部门。Procedia Computer Science，4：1880-1889,2011.[4]Trang Pham，Truyen Tran，Dinh Phung和Svetha Venkatesh。从医疗记录预测医疗保健轨迹：一种深度学习方法。生物医学信息学杂志，2017，69：218-229。[5]徐洪腾，吴维昌，沙米姆·奈马提，查宏源。通过相互纠正过程的辨别性学习进行病人流量预测。《IEEE知识与数据工程学报》，29(1)：157-171,2016。

Paediatric Critical Care (PCC) is an essential service providing health care to critically ill children. In Scotland the PCC service is delivered by two Paediatric Critical Care Units (PCCU); one in Edinburgh, and one in Glasgow. In recent years there has been increasing demand for PCCU beds with a PCC bed crisis being reported. In cases where the critical care units run out of available beds, patients must be diverted to a unit with an available bed or urgent care must be delayed directly impacting patient care. Reasons for the increasing demand for PCCU beds range from seasonal respiratory infections to delays in patient discharges as well as changes in PCC patient characteristics. Bed usage and staffing data is routinely collected from the PCCUs as well as individual patient level data including admission, discharge and physiological data. This gives a unique opportunity to develop powerful data driven tools which could be used to optimise and improve patient flow and ultimately patient care in the PCC. The ability to simulate patient flow through a PCCU could provide valuable insights into operational bottlenecks and inefficiencies. Simulation models such as Discrete Event Simulation (DES), System Dynamics (SD) and Agent Based Systems (ABS) within healthcare are well established with proven benefits once developed [1][2][3]. An accurate and representative simulation gives the opportunity to test possible scenarios for improvement before integration into hospital systems unlike other approaches which typically rely heavily on expert opinion as an approach towards improvement. In recent years Machine Learning (ML) coupled with the availability of large datasets and increase in computational performance has revolutionised many domains. In healthcare this is no different, with the prediction of patient care trajectories using ML becoming an increasingly active area of research [4][5]. With the wealth of data routinely available from the PCCUs at an individual patient level, the development of machine learning models which can predict a probable care trajectory for how a patient may transition through the hospital based on a patient physiological phenotype could be employed to plan future care and improve patient flow.This project will develop a simulation model which can be used to simulate patient flow in the PCCUs, identify bottlenecks and then test solutions for improvement through simulating the proposed scenarios. Secondarily machine learning methods will be developed which allow for the prediction of patient care trajectories from routinely collected clinical data to aid in the planning of patient care. Finally, it will include a thorough health economic analysis to fully quantify the impact this data driven approach brings.References[1] Kusum S Mathews and Elisa F Long. A conceptual framework for improving critical care patient flow and bed use. Annals of the American Thoracic Society, 12(6):886-894, 2015.[2] Syed Mohiuddin, John Busby, Jelena Savovic, Alison Richards, Kate Northstone, William Hollingworth, Jenny L Donovan, and Christos Vasilakis. Patient flow within UK emergency departments: a systematic review of the use of computer simulation modelling methods. BMJ open, 7(5), 2017.[3] Eduardo Cabrera, Manel Taboada, Ma Luisa Iglesias, Francisco Epelde, and Emilio Luque.Optimization of healthcare emergency departments by agent-based simulation. Procedia computer science, 4:1880-1889, 2011.[4] Trang Pham, Truyen Tran, Dinh Phung, and Svetha Venkatesh. Predicting healthcare trajectories from medical records: A deep learning approach. Journal of biomedical informatics, 69:218-229, 2017.[5] Hongteng Xu, Weichang Wu, Shamim Nemati, and Hongyuan Zha. Patient flow prediction via discriminative learning of mutually-correcting processes. IEEE transactions on Knowledge and Data Engineering, 29(1):157-171, 2016.