As of Phase III, the Kor-GLASS network includes 13 university hospitals; among these, five function as collection centers, seven as analysis centers (five of which also serve as collection centers), and one as the QCC. Within the Kor-GLASS framework, clinical isolates obtained at collection centers are transferred to designated analysis centers for standardized identification, AST, and AMR gene analysis. Isolates are shipped in 10% skim milk preservation medium prepared by dissolving commercial skim milk powder (MBCell, KisanBio, Seoul, South Korea) in distilled water, to ensure consistency during transport and processing. As of Phase III, seven analysis centers are actively participating in the network, collectively covering 15 target pathogens. Each analysis center specializes in one or more predefined organism groups, including Staphylococcus aureus, Enterococcus faecalis, Enterococcus faecium, Neisseria gonorrhoeae, Salmonella spp., Shigella spp., Streptococcus pneumoniae, Haemophilus spp., Neisseria meningitidis, Acinetobacter spp., Pseudomonas aeruginosa, Candida spp., Clostridioides difficile, Escherichia coli, and Klebsiella pneumoniae. The bacterial species, AST panels, and AMR genes assigned to each analysis center are summarized in Supplementary Table S1.

Kor-GLASS was developed in alignment with the WHO GLASS framework, including surveillance principles, target pathogens, and antimicrobial-pathogen combinations [5]. All participating laboratories performed laboratory procedures in accordance with the standardized Kor-GLASS operating procedures. These procedures incorporated Clinical and Laboratory Standards Institute (CLSI)-based laboratory methods and interpretive criteria, including guidance on disk diffusion [6], broth microdilution [7], AST breakpoints [8, 9], and microbial identification by targeted DNA sequencing [10]. Bacterial species identification was primarily conducted using matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS). When species-level identification was not achieved or confirmatory identification was required, gene sequencing methods including 16S rRNA and rpoB sequencing for bacterial isolates and internal transcribed spacer (ITS) sequencing for Candida spp. were performed. AST was carried out using either the broth microdilution method or the disk diffusion method, depending on the bacterial species and antimicrobial agents tested. For Clostridioides difficile, AST was performed using the agar dilution method, which differs substantially from routine AST methods used for other target species, and was not included in IPT by the QCC. AMR genes were analyzed using polymerase chain reaction (PCR)-based methods or sequencing techniques.

For clinical breakpoints, CLSI M100 30th edition [8] was applied throughout Phase II (2020–2022), whereas CLSI M100 33rd edition [9] was applied throughout Phase III (2023–2025), to maintain consistency within each surveillance phase.

Within the Kor-GLASS framework, IPT refers to an interlaboratory comparison and retesting process in which the QCC independently retests a subset of routine clinical isolates submitted by analysis centers. IPT was generally conducted on a monthly basis. For organism groups with a limited number of isolates, including Salmonella spp., Streptococcus pneumoniae, Candida spp., and Haemophilus spp., IPT was performed at a reduced frequency.

IPT was performed by comparing AST results between the analysis centers and the QCC. Approximately 5% of isolates from each analysis center were randomly selected during each IPT round; when the total number of isolates was limited, a minimum of five isolates was selected. The QCC independently performed AST on the selected and transferred isolates, and the results were compared with those generated by the corresponding analysis center.

For practical purposes within the Kor-GLASS scheme, both susceptible–resistant (S–R) and resistant–susceptible (R–S) discordances were counted together as major errors, without separately applying the conventional very major error category. Minor errors were defined as discrepancies involving intermediate (I) interpretations (S–I or I–R). Categorical agreement (CA) was defined as concordant classification into susceptible, intermediate, or resistant categories between the two centers. An IPT result was considered acceptable when the major error rate was <3% and the categorical agreement was ≥90%. Antifungal susceptibility testing for Candida spp. was assessed using CA, essential agreement (within ±2 dilutions of the minimum inhibitory concentration), and total agreement, in accordance with established antifungal susceptibility testing metrics [11].

In cases involving major errors or failure to meet the acceptance criteria, the analysis center was required to submit a corrective action report, including an investigation of the underlying causes and implementation of appropriate corrective measures.

In this study, IPT results obtained during 2020–2024 (Phase II and the ongoing Phase III) were analyzed.

Within the Kor-GLASS framework, EQA refers to an assessment of the performance of participating centers based on test results for pre-characterized strains distributed by the QCC. EQA was conducted quarterly by distributing pre-characterized strains from the QCC to collection centers and analysis centers. Each EQA round typically consisted of five strains and was designed to evaluate performance in species identification, AST, and AMR gene analysis, according to the role of each participating center. Collection centers performed species identification only. Analysis centers performed species identification for all distributed strains, while AST and AMR gene analysis were conducted exclusively for strains belonging to the organism groups assigned to each analysis center.

The results of species identification, AST, and AMR gene analysis were independently evaluated and scored as percentage agreement. An overall score of ≥90% was required to achieve a pass status. In cases where the acceptance criteria were not met, the participating center was required to investigate the causes of discordant results and submit a corrective action report detailing the identified issues and implemented remedial measures.

In this study, EQA results obtained during 2020–2024 (Phase II and the ongoing Phase III) were analyzed.

Following the addition of ceftazidime-avibactam to the AST panel in the Kor-GLASS Phase III, its analytical performance was further evaluated as part of routine IPT. AST for ceftazidime-avibactam was performed using the disk diffusion method for Escherichia coli and Klebsiella pneumoniae. Results generated at the designated analysis center were compared with those independently obtained by the QCC to assess CA and differences in zone diameter measurements.

To support the introduction of Haemophilus spp. as an additional target bacterial species within the Kor-GLASS Phase III analytical framework, an interlaboratory validation was conducted prior to formal implementation. An existing analysis center, previously designated for Salmonella spp., Shigella spp., Streptococcus pneumoniae, and Neisseria meningitidis, participated in this validation as a candidate center for the expansion of its analytical scope to include Haemophilus spp.

AST results for validation isolates of Haemophilus spp. generated by the candidate analysis center were compared with results independently generated by two separate teams within the National Reference Laboratory (NRL teams I and II), and the QCC. Following three pilot rounds, a formal validation round was conducted, in which predefined acceptance criteria (major error rate <3% and categorical agreement ≥90%, based on pairwise comparisons against the QCC) were met, thereby confirming the suitability of the expanded analytical scope.

In addition to IPT and EQA, the QCC undertook supplementary quality assurance activities to support sustained standardization, coordination, and capacity building within the Kor-GLASS network and the national One Health framework. These activities included regular communication with Kor-GLASS participating institutions, as well as quality control support for laboratories participating in the Korean One Health project that perform AST of non-human-derived isolates. The QCC also provided technical consultation and educational support. Educational and feedback activities were delivered through online practical lectures for laboratory technicians, quarterly Kor-GLASS operational meetings, and feedback accompanying IPT and EQA reports, with more focused support provided when new centers, target organisms, or antimicrobial agents were introduced. In addition, the QCC deposited reference strains to support standardized laboratory practice. All QCC activities related to culture, preservation, IPT, and EQA for antimicrobial-resistant organisms were conducted under an ISO 9001:2015-certified quality management system.

Interlaboratory agreement of ceftazidime-avibactam disk diffusion zone diameter measurements between the analysis center and the QCC was evaluated using Bland-Altman analysis [12] with Microsoft Excel (Microsoft Corp., Redmond, WA, USA). The mean zone diameter and interlaboratory difference (analysis center - QCC) were calculated for each paired measurement. The mean difference, standard deviation of the differences, and upper and lower 95% limits of agreement were then calculated. The results were visualized as a Bland-Altman plot using the scatter plot function. Exact McNemar test was performed using SPSS software version 27 (IBM Corp., Armonk, NY, USA) to assess directional discordance between susceptible/resistant classifications.

A representative IPT report is presented in Figure 2, and the aggregated IPT results are summarized in Table 1. During Phase II (2020–2022), a total of 180 IPT reports were generated based on 2,381 isolates, whereas 122 IPT reports were issued from 1,291 isolates during Phase III (2023–2024; ongoing). For Candida spp., during Phase II, 63 isolates were evaluated, yielding CA, essential agreement, and total agreement rates of 99.8%, 96.8%, and 98.3%, respectively; in Phase III, 49 isolates were assessed, with corresponding agreement rates of 99.7%, 100%, and 99.9%. Shigella spp., Neisseria meningitidis, and Neisseria gonorrhoeae were not included in the IPT due to extremely limited availability of eligible clinical isolates. Across all species tested, the overall categorical agreement consistently exceeded 97% throughout the study period, and this performance was maintained across the transition from Phase II to Phase III.

Figure 3 shows a representative EQA report selected from the actual reports issued during the study period, whereas Table 2 summarizes the overall EQA results of all participating centers. During the 5-year period, 226 EQA reports were generated from 12 participating centers, consisting of five collection centers and seven analysis centers; one collection center joined the EQA program from the third quarter of 2023. Overall, all participating centers demonstrating sustained compliance with predefined quality criteria, with no EQA failures observed.

The IPT pass rate was 100% during Phase II (2020–2022) and 96.2% during Phase III (2023–2024; ongoing). The EQA pass rate was 100% throughout the study period. For IPT and EQA rounds that did not meet the pass criteria, as well as for rounds that passed but contained major errors, appropriate corrective actions were implemented according to the identified causes (Table 3). During Phase II, a total of 26 major errors were identified, with reading-related errors and near-breakpoint discrepancies accounting for the majority of cases. These errors were primarily associated with difficulties in zone interpretation, disk placement, and narrow interpretive breakpoints. In contrast, the number of major errors decreased to four cases during Phase III, all of which were attributable to data entry errors or sample-related issues. Notably, no reading-related errors were observed during Phase III following targeted educational interventions, indicating a substantial improvement in interpretive consistency.

Following the introduction of ceftazidime-avibactam into the AST panel in Phase III (2023–), disk diffusion results were analyzed for a total of 870 Enterobacterales isolates. Complete categorical agreement was observed for 455 Escherichia coli isolates, whereas three discordant results (0.7%) were identified among 415 Klebsiella pneumoniae isolates (Table 4). Exact McNemar testing showed no statistically significant directional difference in susceptible and resistant classifications of K. pneumoniae between the analysis center and the QCC (p = 0.25). All discordant cases represented near-breakpoint discrepancies, in which zone diameters measured at the analysis center were 19–20 mm (interpreted as resistant), while corresponding measurements at the QCC ranged from 22 to 26 mm (interpreted as susceptible), based on the breakpoint definition (≤20 mm, resistant; ≥21 mm, susceptible) [9]. The mean zone diameters and distribution ranges for susceptible and resistant isolates were highly comparable between the two centers across both species. In addition, Bland–Altman analysis demonstrated a small interlaboratory bias in zone diameter measurements between the analysis center and the QCC (mean difference [analysis center - QCC], −0.81 mm) (Figure 4).

Interlaboratory validation results of AST for Haemophilus spp. are summarized in Table 5. Following completion of the three pilot rounds, a formal validation round was performed, in which all predefined acceptance criteria were met based on pairwise comparisons against the QCC. These results confirmed the suitability of expanding the analytical scope of the existing analysis center to include Haemophilus spp.

During the study period, the QCC provided quality control support and educational programs to six additional laboratories participating in the Korean One Health project. Network coordination was maintained through quarterly Kor-GLASS meetings, and annual educational programs were conducted to reinforce standardized practices across sectors.

Furthermore, seven rare and epidemiologically significant strains, including multi-carbapenemase-producing Enterobacterales, were fully characterized and deposited as reference isolates in the National Culture Collection for Pathogens. Throughout the study period, the QCC maintained its ISO 9001:2015-certified quality management system.

National AMR surveillance depends on standardized laboratory testing and robust quality assurance systems.

A centralized, QCC-led framework sustained high IPT and EQA performance.