Predicting Clinical Events Using NLP Analysis of Clinical Notes in Diabetes Patients

“**概述**电子健康记录（EHR）的引入和远离纸质笔记导致了医疗保健数据的激增，其中大部分以自由文本笔记的形式保存，现在可以在不同的医疗保健设置和临床专业之间共享。然而，数据量的增加，以及慢性病和合并症的发病率的增加，意味着临床医生很难在分配的短预约时间内综合这些信息。这就产生了这样的风险：临床医生可能对病人的整体健康状况没有全面的了解，可能会错过重要的症状。这个项目的愿景是建立机器学习模型，这些模型将(1)分析与患者相关的所有临床记录，(2)预测不同临床终点（如心脏病发作或死亡）的风险(3)并将这些信息作为评分或警报呈现给临床医生。临床医生可以使用它来定制咨询，识别高风险患者，并针对特定的临床结果。**目的**本可行性研究将评估开发ML模型并在临床环境中实施的技术可行性。临床和技术专长的合作伙伴关系还将考虑如何将这些技术商业化，以及什么是最合适的商业模式。**Focus**SCI-Diabetes是世界知名的EHR，拥有苏格兰99%糖尿病患者的全面记录。可行性研究将侧重于利用这些数据预测糖尿病患者的不同临床终点。**创新**除了手动点击进入每个笔记-一个耗时的过程-没有办法为临床医生审查病人的整个历史。大多数其他NLP方法旨在从自由文本中提取结构化信息，并将其转换为临床代码（例如识别特定疾病的提及）。该方案不是从自由文本中提取信息，而是使用文本直接预测不同的临床终点。除了分析整个患者病史外，该模型还将受益于能够汇总不同的临床判断，甚至发现疾病进展的新模式。**合作伙伴**主要申请人红星将开发机器学习模型，合作者将是NHSGG&C（在数据、疾病和机器学习模型开发方面的专业知识）、Tactuum（在英国和美国的决策支持工具方面的专业知识）和苏格兰DHI的Ann Wales博士，他也是苏格兰政府知识和决策支持计划的主任。”

"**Overview**The introduction of Electronic Health Records (EHR) and the move away from paper notes has led to a proliferation of healthcare data, much of it held in free-text notes, which can now be shared across different healthcare settings and clinical specialties. However the increase in volume of data, alongside increasing rates of chronic illness and co-morbidities, has meant that clinicians struggle to synthesise this information within the short appointment times allocated. This gives rise to the risk that the clinician may not have a full picture of the overall health of the patient and may miss important symptoms.**Vision**The vision of this project is to build Machine Learning models which will (1) analyse all the clinical notes associated with a patient, (2) predict the risk of different clinical endpoints such as heart attack or death (3) and present this information to the clinician as a score or alert. Clinicians can use this to tailor the consultation, identify high risk patients, and target specific clinical outcomes.**Objectives**This feasibility study will assess the technical feasibility of developing ML models and implementing them in a clinical setting. The collaborative partnership of clinical and technical expertise will also consider how to commercialise such technology and what is the most appropriate business model.**Focus**SCI-Diabetes is a world renowned EHR which has comprehensive records for 99% of diabetes patients in Scotland. The feasibility study will focus on predicting different clinical endpoints for diabetic patients using this data.**Innovation**Other than manually clicking into each note - a time consuming process - there is no way for clinicians to review the entire history of a patient. Most other NLP approaches aim to extract structured information from free text and convert these into clinical codes (such as identifying mentions of specific diseases).Instead of extracting information from free text, this proposal uses the text to directly predict different clinical endpoints. As well as analysing the entire patient history, the model will benefit from being able to aggregate different clinical judgements and even detect new patterns of disease progression.**Partners**The lead applicant Red Star will develop the ML models and collaborators will be NHSGG&C (expertise on the data, disease and development of ML models), Tactuum (expertise in decision support tools in both UK and USA) and Dr Ann Wales from DHI Scotland who is also Director for Scottish Government Knowledge and Decision Support Programme."