Islam Md Mohaimenul, Poly Tahmina Nasrin, Yang Hsuan-Chia, Li Yu-Chuan Jack
Graduate Institute of Biomedical Informatics, College of Medical Science and Technology, Taipei Medical University, Taipei 110, Taiwan.
International Center for Health Information Technology (ICHIT), Taipei Medical University, Taipei 110, Taiwan.
Diagnostics (Basel). 2021 May 29;11(6):990. doi: 10.3390/diagnostics11060990.
Laboratory tests are performed to make effective clinical decisions. However, inappropriate laboratory test ordering hampers patient care and increases financial burden for healthcare. An automated laboratory test recommendation system can provide rapid and appropriate test selection, potentially improving the workflow to help physicians spend more time treating patients. The main objective of this study was to develop a deep learning-based automated system to recommend appropriate laboratory tests. A retrospective data collection was performed at the National Health Insurance database between 1 January 2013, and 31 December 2013. We included all prescriptions that had at least one laboratory test. A total of 1,463,837 prescriptions from 530,050 unique patients was included in our study. Of these patients, 296,541 were women (55.95%), the range of age was between 1 and 107 years. The deep learning (DL) model achieved a higher area under the receiver operating characteristics curve (AUROC micro = 0.98, and AUROC macro = 0.94). The findings of this study show that the DL model can accurately and efficiently identify laboratory tests. This model can be integrated into existing workflows to reduce under- and over-utilization problems.
进行实验室检查是为了做出有效的临床决策。然而，不恰当的实验室检查医嘱会妨碍患者护理，并增加医疗保健的经济负担。一个自动化的实验室检查推荐系统可以提供快速且恰当的检查选择，有可能改善工作流程，帮助医生有更多时间治疗患者。本研究的主要目的是开发一个基于深度学习的自动化系统来推荐合适的实验室检查。在国民健康保险数据库中于2013年1月1日至2013年12月31日进行了回顾性数据收集。我们纳入了所有至少有一项实验室检查的处方。我们的研究纳入了来自530,050名独特患者的总共1,463,837张处方。在这些患者中，296,541名是女性（55.95%），年龄范围在1岁至107岁之间。深度学习（DL）模型在受试者工作特征曲线下面积（AUROC微值 = 0.98，AUROC宏值 = 0.94）方面表现更高。本研究结果表明，DL模型能够准确且高效地识别实验室检查。该模型可以整合到现有工作流程中，以减少检查使用不足和过度使用的问题。
