**Background:** Antibiotic overuse and misuse are major drivers of antimicrobial resistance, particularly in low- and middle-income countries. In China, primary care institutions serve over half of all outpatient visits, yet antibiotic prescribing practices in these settings, especially in underdeveloped regions, are poorly characterized. This study aimed to describe trends and patterns of antibiotic prescriptions in primary care institutions in Guizhou Province, a relatively underdeveloped area in southwest China, from 2017 to 2022.
**Methods:** A retrospective analysis was conducted using electronic health records from 25 primary care institutions (township hospitals and community health service centers) in Guizhou Province. A total of 941,924 systemic antibiotic prescriptions (after excluding topical antibiotics and tuberculosis treatments) were included. Prescriptions were classified as appropriate or inappropriate based on Chinese national guidelines (2015 Guiding Principles for Clinical Antibiotic Use), CDC guidelines, and expert opinion. Inappropriate use was further categorized into unnecessary use, incorrect spectrum of antibiotics, and combined use of antibiotics. Generalized estimation equations (GEE) were used to identify factors associated with inappropriate prescribing. Holt-Winters and seasonal autoregressive integrated moving average (SARIMA) models were developed to predict the number of inappropriate antibiotic prescriptions, with model performance compared using R², BIC, RMSE, MAPE, and MSE.
**Key Results:** The overall number of antibiotic prescriptions and the inappropriate antibiotic rate (AIR) showed a general downward trend from 2017 to 2022. The average AIR across the study period was 66.19% (61.04% due to incorrect spectrum, 5.15% due to unnecessary use). Diseases of the respiratory system accounted for 70.62% of all antibiotic prescriptions, with acute upper respiratory infections of multiple and unspecified sites (J06) representing 52.04% of these cases. Penicillins were the most commonly prescribed antibiotic class (64.44%). The highest AIRs were observed for diseases of the musculoskeletal system and connective tissue (100%) and symptoms, signs, and abnormal clinical findings (100%). In multivariate GEE analysis, factors significantly associated with inappropriate antibiotic prescribing included: physician age >35 years (OR 1.12 for 35–45 years, 1.38 for 46–64 years vs. 26–34 years), associate chief physician title (reference), work experience >11 years (OR 1.40 for 11–20 years, 1.51 for 31–40 years vs. 6–10 years), male physician sex (OR 1.07), patient age 0–5 years (reference), injection route (OR 0.50 for oral vs. injection), and out-of-pocket payment (OR 0.93 for insurance vs. out-of-pocket). The SARIMA model (ARIMA (0,1,0) (1,1,0)~12~) outperformed the Holt-Winters simple seasonal model in predicting the number of inappropriate antibiotic prescriptions for 2022, with a relative error of 0.35% vs. 27.65%.
**Clinical Implications:** Despite a declining trend, inappropriate antibiotic prescribing remains highly prevalent in primary care institutions in underdeveloped regions of China, with over two-thirds of prescriptions deemed inappropriate. Respiratory infections, particularly viral upper respiratory infections, are the most common targets of unnecessary antibiotic use. The study identifies key physician and patient characteristics associated with inappropriate prescribing, such as older physician age, higher professional title, longer work experience, and younger patient age. These findings underscore the need for targeted educational interventions and antibiotic stewardship programs in primary care settings, especially for physicians with entrenched prescribing habits. The superior predictive performance of the SARIMA model suggests it can be a useful tool for monitoring and forecasting antibiotic use trends to inform policy and intervention planning.