Algoritmi Research Center, University of Minho, Campus of Gualtar, Braga, 4710, Portugal.
Centro Hospitalar Universitário do Porto, Porto, 4099, Portugal.
Sci Rep. 2022 Jul 22;12(1):12549. doi: 10.1038/s41598-022-15968-z.
Nowadays, we are facing the worldwide pandemic caused by COVID-19. The complexity and momentum of monitoring patients infected with this virus calls for the usage of agile and scalable data structure methodologies. OpenEHR is a healthcare standard that is attracting a lot of attention in recent years due to its comprehensive and robust architecture. The importance of an open, standardized and adaptable approach to clinical data lies in extracting value to generate useful knowledge that really can help healthcare professionals make an assertive decision. This importance is even more accentuated when facing a pandemic context. Thus, in this study, a system for tracking symptoms and health conditions of suspected or confirmed SARS-CoV-2 patients from a Portuguese hospital was developed using openEHR. All data on the evolutionary status of patients in home care as well as the results of their COVID-19 test were used to train different ML algorithms, with the aim of developing a predictive model capable of identifying COVID-19 infections according to the severity of symptoms identified by patients. The CRISP-DM methodology was used to conduct this research. The results obtained were promising, with the best model achieving an accuracy of 96.25%, a precision of 99.91%, a sensitivity of 92.58%, a specificity of 99.92%, and an AUC of 0.963, using the Decision Tree algorithm and the Split Validation method. Hence, in the future, after further testing, the predictive model could be implemented in clinical decision support systems.
如今，我们正面临着由 COVID-19 引发的全球大流行。监测感染这种病毒的患者的复杂性和速度要求使用灵活且可扩展的数据结构方法。OpenEHR 是近年来备受关注的医疗保健标准，因其全面而强大的架构而受到关注。采用开放、标准化和可适应的临床数据方法的重要性在于提取价值以生成有用的知识，这些知识真正可以帮助医疗保健专业人员做出果断的决策。在面对大流行背景时，这种重要性更加突出。因此，在这项研究中，使用 OpenEHR 开发了一个从葡萄牙医院跟踪疑似或确诊 SARS-CoV-2 患者症状和健康状况的系统。所有在家中接受护理的患者的进化状态数据以及他们的 COVID-19 检测结果都用于训练不同的 ML 算法，目的是开发一种能够根据患者识别出的症状严重程度识别 COVID-19 感染的预测模型。使用 CRISP-DM 方法论进行了这项研究。结果令人鼓舞，最佳模型使用决策树算法和拆分验证方法，达到了 96.25%的准确率、99.91%的精度、92.58%的灵敏度、99.92%的特异性和 0.963 的 AUC。因此，在未来，经过进一步测试后，预测模型可以在临床决策支持系统中实施。
