Study design and setting

A retrospective cross-sectional analysis was conducted at a tertiary-level hospital in Johannesburg, South Africa. Epidemiological data for all colistin-resistant A. baumannii and K. pneumoniae isolates from 1 January 2022 to 30 June 2024 were analyzed. For each isolate, patient demographic and isolate information such as age, sex, ward, specimen type, organism identification (ID), antibiotic susceptibility profile and colistin broth microdilution (BMD) results were obtained from the Corporate Data Warehouse of the National Health Laboratory Services (NHLS).

Data processing and statistical analysis

Prior to analysis, duplicate entries were identified and removed using Microsoft Excel v2502 (Microsoft Corporation, USA). If duplicates represented the same organism, the earliest sterile site specimen information was used. In cases where both A. baumannii and K. pneumoniae were recorded under the same identifier, the first sterile site specimen for each species was retained.

Cleaned epidemiological data were then binary coded and imported into Stata v18 (StataCorp LLC, College Station, TX, USA). Sex, age group, ward category and sample type were treated as categorical variables. Associations between ward category and colistin resistance were evaluated using the χ² test, with statistical significance defined as a two-tailed pvalue < 0.05.

Ward classification

To facilitate comparison, clinical wards were grouped into broader categories (Supplementary Table 1). The following were the categories: adult medical, neonatal units, pediatric medical, pediatric surgical, trauma, adult medical multidisciplinary intensive care units (ICUs), adult oncology, general surgery, other surgical specialities and gynaecology.

Resistance categorization

The on-site 24-hour NHLS Microbiology Laboratory routinely assessed susceptibility to 5 antibiotic classes: β-lactams (including aminopenicillins, cephalosporins, carbapenems, β-lactam/β-lactamase inhibitor combinations), aminoglycosides, antifolates, fluoroquinolones, and tetracyclines (Supplementary Table 2). All specimens were processed using standard microbiological protocols. Species identification was performed using either the Vitek®2 system or the Vitek MS matrix-assisted laser desorption/ionization time-of-flight mass spectrometry platform (MALDI-TOF MS) (bioMérieux Inc., Marcy l’Étoile, France). Antimicrobial susceptibility testing was conducted using the Vitek®2 AST N255 panel (bioMérieux Inc.).

Isolates were classified according to internationally accepted definitions of MDR and XDR. Multidrug resistance was defined as “non-susceptibility to at least one agent in three or more antimicrobial categories” and XDR was defined as “non-susceptibility to at least one agent in all but two or fewer antimicrobial categories (ie, bacterial isolates remain susceptible to only one or two categories)”. ^{Reference Magiorakos, Srinivasan and Carey7} Isolates meeting the criteria for MDR or XDR were referred to the NHLS Infection Control Services Laboratory for colistin susceptibility testing by the BMD method. During the study period, 2 900 A. baumannii and K. pneumoniae isolates were processed, of which 830 underwent colistin BMD testing. Minimum inhibitory concentration (MIC) values were interpreted according to the Clinical and Laboratory Standards Institute (CLSI) guidelines (M100, v32 to v34; 2022–2024), ⁸ with MIC ≤ 2 µg/mL classified as intermediately susceptible and MIC ≥ 4 µg/mL as resistant. No susceptible breakpoints exist for colistin for either species according to the CLSI standards.