Improving triaging from primary care into secondary care using heterogeneous data-driven hybrid machine learning
Improving triaging from primary care into secondary care using heterogeneous data-driven hybrid machine learning
复制标题
DOI：
10.1016/j.dss.2022.113899
复制
发表时间：
2023-01-31
影响因子：
7.5
通讯作者：
Chan，Antoni T. Y.
中科院分区：
文献类型：
--
作者：
Wang，Bing;Li，Weizi;Chan，Antoni T. Y.
从初级护理到二级护理的有效和快速分诊在为患者提供及时治疗和管理日益增长的医疗资源需求方面发挥着关键作用。现有的从初级保健到二级保健的分诊方法是劳动密集型过程，涉及手动审查来自多个来源的转诊数据，并可能导致转诊到治疗时间较长。目前还没有研究使用机器学习方法来自动分析异类数据，包括推荐信，以识别支持初级到二级护理分诊的规则。本文提出了一种包含自然语言处理(NLP)的异质数据驱动的混合机器学习模型，以提高医院分诊的效率。在具有可解释的风险分层的分诊点上，所提出的模型在识别非炎症性疾病(NIC)和炎症性关节炎(IA)患者时的精确度为0.83，召回率为0.82，F1-Score为0.83，准确度为0.82，AUC为0.90。我们的模型在英国一家大型二级护理医院的真实世界试验中试用，以比较我们的模型和临床医生之间的转诊准确性和节省的时间，并评估其用户的可接受性。我们的模型的准确率和召回率分别为0.83和0.81，而临床医生的准确率和召回率分别为0.80和0.78。研究还表明，我们的模型启用的决策支持每周可以为临床医生节省8小时来评估转诊评估。这篇论文是第一次使用机器学习来简化从初级护理到二级护理的医院分诊。
Effective and rapid triaging from primary care into secondary care plays a pivotal role in providing patients with timely treatment and managing increasing demands for healthcare resources. Existing triaging methods from primary care to secondary care are labor-intensive processes that involve manually reviewing referral data from multiple sources and can cause long referral to treatment time. There has been no research using machine learning methods that automatically analyzes heterogeneous data including referral letters to recognize regularities to support the primary to secondary care triage. In this paper, we propose a heterogeneous data-driven hybrid machine learning model including Natural Language Processing (NLP) to improve hospital triage efficiency at the point of triage. The proposed model achieved a precision of 0.83, recall of 0.82, F1-Score of 0.83, accuracy of 0.82, AUC of 0.90 in identifying patients with non-inflammatory conditions (NIC) and inflammatory arthritis (IA) at the point of triage with explainable risk stratifications. Our model is piloted in a real-world trial in a large secondary care hospital in the UK to compare referral accuracy and time saved between our model and clinicians, and evaluate its acceptability by users. Our model achieved precision and recall of 0.83 and 0.81, compared with the precision and recall of 0.80 and 0.78 by clinicians. The research also shows that our model enabled decision support can save clinicians 8 h per week in assessing the referral assessment. This paper is the first study to streamline hospital triage from primary care to secondary care using machine learning.