Background: Antimicrobials are used by poultry farmers in Vietnam as a tool to treat and prevent infectious diseases. We aimed to determine the fraction of disease episodes likely to remain untreated due to the administration of antimicrobials on non-susceptible pathogens in chicken flocks in the Mekong Delta of Vietnam. Weekly data on antimicrobial use and clinical signs were collected from 88 randomly chosen chicken flocks over 124 full production cycles (i.e. time between restocking flocks with day-old chicks and sale for slaughter). A naïve Bayes model was trained to infer the probabilities of disease episodes having been caused by each of 24 pathogens, given the observed clinical sign profile, and expert knowledge on their relative incidence. Results: A total of 224 disease episodes were observed, of which 44.8% were attributed to viruses (95% CI 31.1-58.4%), 54.6% (CI 40.4-68.7%) to bacteria, and 0.6% (CI 0-1.7%) to a protozoan (Eimeria spp.). Antimicrobials were more frequently administered on weeks with disease than on weeks without disease (43.3% vs. 17.8%; p < 0.001). A median of 2 [IQR 0-4] antimicrobials were used by episode. The choice of specific antimicrobials was independent on whether the flocks had disease clinical signs or not. Antimicrobials were not used in 30.3% of the episodes. The overall probability that episodes were not effectively treated was 74.2, and 53.7% when discounting cases where the inferred aetiology is viral. Considering only episodes where antimicrobials were given, these probabilities were 57.4 and 23.8% respectively. Conclusions: This study highlights untargeted use of antimicrobials on small-scale Vietnamese chicken farms, as well as the limitations of antimicrobials as effective tools to control infectious diseases.