Objectives Although multiple studies indicate that associations of antimicrobial resistance (AMR) traits do not occur randomly in multidrug resistant (MDR) bacteria, clinically relevant antibiotic resistance patterns remain relatively unexplored. Machine learning approaches have proved usefulness to analyse MDR phenotypes in large datasets [1]. Here, we aim to identify and quantify resistance associations within isolates from community E. coli acquired urinary tract infections (CA-E.coli-UTI) using association rule mining. Methods We analysed antibiograms of isolates from CA-E.coli-UTI collected from 2018 to 2021 by the French national surveillance system of antimicrobial resistance in the community. An association rule approach was used to assess separately extended-spectrum beta-lactamase producing (ESBL) E. coli isolates and non-ESBL E. coli isolates [1]. We described MDR patterns and quantified the associations of resistance traits within these patterns using two quality measures: expected support, which reflects pattern frequency, and conditional lift, which assesses the association strength. We selected patterns in which these measures were significantly higher than expected by chance (p-value ≤ 0.05), and we visualized them as graphical networks of AMR traits. Results The yearly number of CA-E.coli-UTI isolates increased from 362,744 in 2018 to 578,423 in 2021. For 2018, analyses were conducted on 10,204 ESBL and 352,540 non-ESBL E. coli isolates. In ESBL E. coli (Figure 1A), more MDR patterns were selected compared to non-ESBL E. coli (124 vs 27 patterns). In particular, we found frequent associations between resistance to fluroquinolones, third generation cephalosporins and penicillins (amoxicillin and ticarcillin). By contrast, in non-ESBL E. coli, associations were generally stronger (Figure 1B), notably amoxicillin and ticarcillin resistance associated with ofloxacin, nalidixic acid and sulfamethoxazole-trimethoprim resistance. We further investigated temporal trends in MDR patterns using data from the following years, as well as their spatial distribution across France. Conclusion Using association rule mining allowed to discover and quantify associations among AMR resistance traits in CA-E.coli-UTI. While the MDR patterns we described are consistent with current knowledge, this approach may also inform clinical decision by providing insight on the strength and frequency of emerging resistance associations.