Background
Antimicrobial resistance (AMR) is a major global public health threat. Yet, existing studies lack a systematic review of cross-sectoral AMR governance. This study aims to comprehensively assess China's sectoral responses to AMR from 2017-2024, analyze sectoral and regional governance differences, and quantify correlations between response levels and AMR indicators.
Methods
We used AI (e.g., the collaborative Scrapy web crawling technique) to collect text data on how Chinese government departments at all levels have responded to AMR from 2017 to 2024, including policy papers, news, official WeChat posts, and other public data. A Large Language Model standardized the data for sectoral and regional comparison. Finally, we used a distribution-lag model to evaluate the correlations between provincial response levels and AMR.
Results
A total of 11,291 datapoints mapped China's sectoral AMR governance, with human (54.5%), animal (39.5%), environmental (4.2%), and "One Health" (1.6%) sectors contributing. From 2017 to 2024, the number of relevant texts included increased steadily, peaking in 2020 before declining. The main content types of the included texts were administrative management and policy guidance, at 42.6% and 18.7%, respectively, while basic support (8.7%), industrial innovation (7.3%), academic exchange (5.6%), and international cooperation (0.5%) accounted for relatively low proportions. AMR governance response varied by province, higher response scores correlated with lower prevalence.
Conclusions
Environment sector has begun designing policies, yet monitoring remains challenging. Improving One Health awareness among the public and regional leaders is key to action, with cross-sectoral education promising, especially in rural areas. We call for regional pilots for sharing human-animal surveillance data to forecast AMR better.
Limitations
First, restricted internal access to some data prevented full collection, which to some extent reflects the transparency of governance. Second, despite pre-defined categorization and thorough reviewer communication with third-party adjudication, textual data complexity may still cause minor classification discrepancies among different researchers.
