Effects of a feedback intervention on antibiotic prescription control in primary care institutions based on depth graph neural network technology: a cluster randomized cross-over controlled trial

Background Overuse and misuse of antibiotics are major factors in the development of antibiotic resistance in primary care institutions of rural China. In this study, the effectiveness of an artificial intelligence (AI)-based, automatic, and confidential antibiotic feedback intervention was evaluated to determine whether it could reduce antibiotic prescribing rates and avoid inappropriate prescribing behaviors by physicians. Methods A randomized, cross-over, cluster-controlled trial was conducted in 77 primary care institutions of Guizhou Province, China. All institutions were randomly divided into two groups and given either a 3-month intervention followed by a 3-month period without any intervention or vice versa. The intervention consisted of 3 feedback measures: a real-time warning pop-up message of inappropriate antibiotic prescriptions on the prescribing physician’s computer screen, a 10-day antibiotic prescription feedback, and distribution of educational brochures. The primary and secondary outcomes are the 10-day antibiotic prescription rate and 10-day inappropriate antibiotic prescription rate. Results There were 37 primary care institutions with 160 physicians in group 1 (intervention followed by control) and 40 primary care institutions with 168 physicians in group 2 (control followed by intervention). There were no significant differences in antibiotic prescription rates (32.1% vs 35.6%) and inappropriate antibiotic prescription rates (69.1% vs 72.0%) between the two groups at baseline ( p = 0.085, p = 0.072). After 3 months (cross-over point), antibiotic prescription rates and inappropriate antibiotic prescription rates decreased significantly faster in group 1 (11.9% vs 12.3%, p < 0.001) compared to group 2 (4.5% vs 3.1%, p < 0.001). At the end point, the decreases in antibiotic prescription rates were significantly lower in group 1 compared to group 2 (2.6% vs 11.7%, p < 0.001). During the same period, the inappropriate antibiotic prescription rates decreased in group 2 (15.9%, p < 0.001) while the rates increased in group 1 (7.3%, p < 0.001). The characteristics of physicians did not significantly affect the rate of antibiotic or inappropriate antibiotic prescription rates. Conclusion The conclusion is that artificial intelligence based real-time pop-up of prescription inappropriate warning, the 10-day prescription information feedback intervention, and the distribution of educational brochures can effectively reduce the rate of antibiotic prescription and inappropriate rate. Trial registration ISRCTN, ID: [ISRCTN13817256][1]. Registered on 11 January 2020 ### Competing Interest Statement The authors have declared no competing interest. ### Clinical Trial ISRCTN, ID: [ISRCTN13817256][1]. ### Clinical Protocols ### Funding Statement The study was funded by the National Natural Science Foundation of China Grant on ‘Research on feedback intervention mode of antibiotic prescription control in primary medical institutions based on the depth graph neural network technology’ (71964009) and the Science and Technology Fund Project of Guizhou Provincial Health Commission Grant on “Application Research of Deep Learning Technology in Rational Evaluation and Intervention of Antibiotic Prescription” (gzwjkj2019-1-218). Corresponding author YC is the project leader. The funders had a role in the study which we should acknowledge. Specifically, all funders provided travel expenses during the data collection process, as well as the expert’s expenses for providing guidance on the study design, technological guidance and data analysis. ### Author Declarations I confirm all relevant ethical guidelines have been followed, and any necessary IRB and/or ethics committee approvals have been obtained. Yes The details of the IRB/oversight body that provided approval or exemption for the research described are given below: The trial was approved by the Human Trial Ethics (Appendix 2) Committee of Guizhou Medical University (Certificate No.: 2019 (148)) in Dec. 27, 2019, and the protocol was published on January 7th, 2022. I confirm that all necessary patient/participant consent has been obtained and the appropriate institutional forms have been archived, and that any patient/participant/sample identifiers included were not known to anyone (e.g., hospital staff, patients or participants themselves) outside the research group so cannot be used to identify individuals. Yes I understand that all clinical trials and any other prospective interventional studies must be registered with an ICMJE-approved registry, such as ClinicalTrials.gov. I confirm that any such study reported in the manuscript has been registered and the trial registration ID is provided (note: if posting a prospective study registered retrospectively, please provide a statement in the trial ID field explaining why the study was not registered in advance). Yes I have followed all appropriate research reporting guidelines and uploaded the relevant EQUATOR Network research reporting checklist(s) and other pertinent material as supplementary files, if applicable. Yes All relevant data are within the manuscript and its Supporting Information files. * CDSS : Clinical Decision Support System HIS : Hospital Information System DGNN : Depth Graph Neural Network technology AI : artificial intelligence LWTC : Lianke Weixin Technology Co., LTD.。 ICGPHC : Information Center Guizhou Provincial Health Commission ICD-10 : International Classification of Diseases 10th Edition APR : Antibiotic prescription rate IAPR : Inappropriate antibiotic prescription rate [1]: /external-ref?link_type=ISRCTN&access_num=ISRCTN13817256