University of Leeds, Leeds, United Kingdom.
University of Leeds, Leeds, United Kingdom; Royal Centre for Defence Medicine, Research & Clinical Innovation (RCI), ICT Centre, Vincent Drive, Birmingham, United Kingdom.
Artif Intell Med. 2024 Dec;158:102999. doi: 10.1016/j.artmed.2024.102999. Epub 2024 Oct 23.
Using graph theory, Electronic Health Records (EHRs) can be represented graphically to exploit the relational dependencies of the multiple information formats to improve Machine Learning (ML) prediction models. In this systematic qualitative review, we explore the question: How are graphs used on EHRs, to predict diagnosis and health outcomes?
The search strategy identified studies that used patient-level graph representations of EHRs to utilise ML to predict health outcomes and diagnoses. We conducted our search on MEDLINE, Web of Science and Scopus.
832 studies were identified by the search strategy, of which 27 studies were selected for data extraction. Following data extraction, 18 studies used ML with patient-level graph-based representations of EHRs to predict health outcomes and diagnoses. Models ranged from traditional ML to neural network-based models. MIMIC-III was the most used dataset (n = 6, where n is the number of occurrences), followed by National Health Insurance Research Database (NHIRD) (n = 4) and eICU Collaborative Research Database (eICU) (n = 4). The most predicted health outcomes were mortality (n = 9; 21%), hospital readmission (n = 9; 21%), and treatment success (n = 4; 9%). Model performances ranged across outcomes, mortality prediction (Area Under the Receiver Operating Characteristic (AUROC): 72.1 - 91.6; Area Under Precision-Recall Curve (AUPRC): 34.8 - 81.3) and readmission prediction (AUROC: 63.7 - 85.8; AUPRC 39.86 - 84.7). Only one paper had a low Risk of Bias (RoB) that applied to our research question (4%).
Graph-based representations using EHRs, for individual health outcomes and diagnoses requires further research before we can see the results applied clinically. The use of graph representations appears to improve EHR representation and predictive performance compared to baseline ML methods in multiple fields of medicine.
利用图论，电子健康记录（EHR）可以以图形方式表示，以利用多种信息格式的关系依赖性来改进机器学习（ML）预测模型。在这个系统的定性综述中，我们探讨了一个问题：如何在 EHR 上使用图形来预测诊断和健康结果？
搜索策略确定了使用 EHR 患者水平图形表示来利用 ML 预测健康结果和诊断的研究。我们在 MEDLINE、Web of Science 和 Scopus 上进行了搜索。
通过搜索策略共确定了 832 项研究，其中 27 项研究被选作数据提取。在数据提取后，18 项研究使用基于患者水平图形的 EHR ML 来预测健康结果和诊断。模型范围从传统 ML 到基于神经网络的模型。MIMIC-III 是使用最多的数据集（n = 6，其中 n 是出现的次数），其次是国家健康保险研究数据库（NHIRD）（n = 4）和 eICU 协作研究数据库（eICU）（n = 4）。预测最多的健康结果是死亡率（n = 9；21%）、住院再入院（n = 9；21%）和治疗成功（n = 4；9%）。模型性能因结果而异，死亡率预测（接收者操作特征曲线下的面积（AUROC）：72.1-91.6；精度-召回率曲线下的面积（AUPRC）：34.8-81.3）和再入院预测（AUROC：63.7-85.8；AUPRC 39.86-84.7）。只有一篇论文的风险偏倚（RoB）较低，适用于我们的研究问题（4%）。
使用 EHR 进行基于图形的个体健康结果和诊断表示需要进一步研究，然后我们才能看到临床应用的结果。在多个医学领域中，与基线 ML 方法相比，图形表示的使用似乎可以提高 EHR 表示和预测性能。
