项目申请人在前期研究中发现基层医疗机构抗生素的无指征、超剂量、超疗程和随意联合用药等滥用情况较为严重。抗生素滥用将增加抗生素耐药性风险，在威胁生命的同时还会增加医疗负担。当前国内外报道了一些抗生素滥用干预措施，但部分措施可行性不高，或干预效果不显著，因此建立有效的抗生素干预机制，促进抗生素的合理使用意义重大。本研究以贵州省基层医疗机构门诊为研究对象，对抗生素处方大数据进行合理性分析，应用“德尔菲法”构建专家论证体系，在此过程中利用深度图神经网络技术进行学习和反馈，形成并不断完善抗生素规范使用的计算机深度学习体系和智能化的理想治疗预案，将该预案整合到基层医院HIS系统中进行预警干预试验，让基层医生在提供诊疗服务的同时，能在线及时有效获得抗生素用药合理性的反馈和建议，形成自动的、智能化的预警干预系统，降低基层医疗机构的抗生素不规范使用比例，为抗生素滥用问题的解决提供可行性方案。

In previous study, the project applicants found that the abuse of antibiotics in the primary care institutions was serious, such as use without indication, use in overdose, use beyond treatment-course and use by random combination. Antibiotic abuse will increase the risk of antibiotic resistance, thus give threaten to our life and increase the medical burden. At present, some interventions measures for antibiotic abuse have been reported in China and abroad, but some of these measures were not feasible, or the intervention effects were not significant. Therefore, it is of great significance to establish an effective antibiotic intervention mode to promote the rational use of antibiotics. This study will take the primary care institutions in Guizhou Province as research object, by analyzing the rationality of the antibiotic prescription based on big data in health information system (HIS) of the primary hospitals, to construct an expert demonstration system using Delphi method. And in this process, to form and improve an ideal treatment plan for antibiotics use by continuously learning and feedback of AI depth map neural network. Then the plan will be integrated into the HIS of primary hospitals for early warning intervention test, thus the physicians will receive timely and effective feedback and suggestions on the rationality of antibiotics while providing medical services. By this automatic and intelligent early warning intervention system, to reduce the proportion of antibiotics used in primary care institutions and provide a feasible reference for the solution of antibiotic abuse problems.