Seven-Year Surveillance and AI-Based Forecasting of Antimicrobial Resistance in Pediatric Escherichia coli Infections in Northern China (2018–2024)

Background: infections are a major cause of morbidity, and increasing antimicrobial resistance (AMR) complicates empirical treatment. Long-term local surveillance combined with trend forecasting may support rational antimicrobial use and stewardship in northern China. Methods: -tests, linear regression, and autoregressive integrated moving average (ARIMA) and long short-term memory (LSTM) models to evaluate temporal trends and generate exploratory forecasts for 2025-2027. Results: Nearly half of isolates were from children 50% in 2018-2020 and around 40% thereafter, while carbapenems and amikacin maintained very low resistance rates. Both ARIMA and LSTM models suggested a modest further decline in the overall resistance indicator through 2027, with LSTM showing slightly better fit and lower prediction errors. Conclusion: isolates in our center exhibited high resistance to several common oral agents but encouraging declines in ESBL prevalence and cephalosporin resistance, likely reflecting local antimicrobial stewardship. Exploratory AI-based time-series models may help anticipate resistance trajectories and support pediatric antibiotic policies and stewardship, although forecasts from a short annual series should be interpreted cautiously.