Abstract:
Introduction: The use of antibiotics leads to antibiotic resistance (ABR). Different methods have been used to predict and control ABR. In recent years, artificial intelligence (AI) has been explored to improve antibiotic (AB) prescribing, and thereby control and reduce ABR. This review explores whether the use of AI can improve antibiotic prescribing for human patients. Methods: Observational studies that use AI to improve antibiotic prescribing were retrieved for this review. There were no restrictions on the time, setting or language. References of the included studies were checked for additional eligible studies. Two independent authors screened the studies for inclusion and assessed the risk of bias of the included studies using the National Institute of Health (NIH) Quality Assessment Tool for observational cohort studies. Results: Out of 3692 records, fifteen studies were eligible for full-text screening. Five studies were included in this review, and a narrative synthesis was carried out to assess their findings. All of the studies used supervised machine learning (ML) models as a subfield of AI, such as logistic regression, random forest, gradient boosting decision trees, support vector machines and K-nearest neighbours. Each study showed a positive contribution of ML in improving antibiotic prescribing, either by reducing antibiotic prescriptions or predicting inappropriate prescriptions. However, none of the studies reported the engagement of AB prescribers in developing their ML models, nor their feedback on the user-friendliness and reliability of the models in different healthcare settings. Conclusion: The use of ML methods may improve antibiotic prescribing in both primary and secondary settings. None of the studies evaluated the implementation process of their models in clinical practices. Prospero Registration: (CRD42022329049).
摘要:
简介:使用抗生素会导致抗生素耐药性(ABR)。已经使用不同的方法来预测和控制ABR。近年来,已经探索了人工智能(AI)来改善抗生素(AB)处方,从而控制和降低ABR。这篇综述探讨了人工智能的使用是否可以改善人类患者的抗生素处方。方法:本综述检索了使用AI改善抗生素处方的观察性研究。时间没有限制,设置或语言。检查纳入研究的参考是否有其他合格研究。两名独立作者筛选了纳入研究,并使用美国国立卫生研究院(NIH)质量评估工具进行观察性队列研究,评估了纳入研究的偏倚风险。结果:在3692条记录中,15项研究符合全文筛选条件.这篇综述包括了五项研究,并进行了叙事综合以评估他们的发现。所有研究都使用监督机器学习(ML)模型作为AI的子领域,如逻辑回归,随机森林,梯度增强决策树,支持向量机和K近邻。每一项研究都显示了ML在改善抗生素处方方面的积极贡献,通过减少抗生素处方或预测不适当的处方。然而,没有一项研究报告AB处方者参与开发他们的ML模型,他们对不同医疗保健环境中模型的用户友好性和可靠性的反馈也是如此。结论:ML方法的使用可以改善主要和次要环境中的抗生素处方。没有一项研究评估了其模型在临床实践中的实施过程。Prospero注册:(CRD42022329049)。