This study aimed to evaluate the dynamics of antimicrobial susceptibility by analyzing a large dataset of paired and successive bacterial isolates obtained from a single tertiary care center in Korea to determine a reasonable frequency for repeat AST. Our findings provide insights into the stability of antimicrobial susceptibility patterns over time for several common bacterial pathogens, which could help inform local laboratory policies regarding repeat testing.

Our analysis of “paired” isolates demonstrated a high CA for most organisms, including S. aureus, E. faecalis, E. faecium, E. coli, and A. baumannii complex. However, P. aeruginosa exhibited the lowest concordance, even on the same day, indicating a marked variation in MIC. This variability was further highlighted by the substantial differences observed in agreement rates for specific P. aeruginosa—antimicrobial combinations, suggesting potential issues with relying on the AST of a single isolate. For successive isolates, the EMI, CNRR, and combined CNRR+EMI generally increased with longer time intervals (from 1–7 to 31–365 days).

The S. aureus and E. faecalis species demonstrated a low frequency of changes in resistance, with CNRR+EMI rates remaining low even up to 365 days. Similarly, E. faecium exhibited a relatively low CNRR+EMI. These findings are largely consistent with and, in the case of the long-term stability of S. aureus, extend the observations of Sarink et al. [1], who reported a < 10% risk of resistance for S. aureus within 30 days. Our data suggest that this stability persists much longer in a substantial proportion of isolates. Thomson et al. [7] also suggested that repeat testing for S. aureus is not routinely necessary, which aligns with our stability findings.

In contrast, gram-negative bacilli exhibited more pronounced changes over time. For E. coli, the CNRR+EMI increased from 2.4% at 1–7 days to 6.4% at 31–365 days, whereas K. pneumoniae showed a CNRR+EMI of 3.0%, which increased to 9.0% over the longest interval. When considering the upper bound of our 1–7-day interval, these CNRR+EMI rates are broadly comparable to the approximately 10% risk for Enterobacterales within 7 days noted by Sarink et al. [1]. The recommendation by Thomson et al. [7] to retest Enterobacteriaceae after 3 days seems reasonable, given the observed changes in CNRR+EMI rates.

The P. aeruginosa species consistently demonstrated the highest propensity for change, with the lowest EA and CA measured in both paired and successive isolates and the highest EMI rates over time. Moreover, its CNRR+EMI was 4.1% within 1–7 days, which increased to 6.6% at longer intervals. This inherent variability and tendency to acquire resistance in P. aeruginosa have been a consistent theme, supported by Sarink et al. [1], Köck et al. [2], and early work by Thomson et al. [7], all of whom suggest more frequent testing for this pathogen. When examining changes in susceptibility by antibiotic class, β-lactam antibiotics showed more frequent shifts, reflecting higher rates of acquiring resistance. In contrast, colistin exhibited little variation, with minimal changes observed over time.

Our data for A. baumannii complex also showed increasing CNRR+EMI rates over time, highlighting another gram-negative organism that warrants careful consideration for repeat testing. The reasons for these observed changes may be multifactorial, including the selection of preexisting resistant subpopulations, de novo mutations, acquisition of resistance elements, or even the isolation of a new strain, especially over longer periods [11,12].

Focusing on the first 7 days, early antimicrobial susceptibility changes revealed notable differences depending on the pathogen and antibiotic class involved. Although isolates such as E. faecium, S. aureus, and E. coli occasionally crossed the 6% threshold, which represents a statistical doubling of the 3% major error limit defined by the CLSI standards for AST system validation [10], and even the 10% threshold suggested by Sarink et al. [1], their shift in resistance rates remained inconsistent, suggesting that these fluctuations may partly reflect inherent analytical variability rather than stable evolutionary shifts. In contrast, P. aeruginosa exhibited a rapid and sustained progression toward resistance: P. aeruginosa isolates frequently surpassed the 3% analytical noise level as early as day 1 and exceeded the 6% warning threshold by day 3, with resistance levels remaining consistently high thereafter. These findings reinforce the clinical recommendation of more aggressive AST monitoring for P. aeruginosa while suggesting that for species such as E. coli or S. aureus, follow-up AST may be safely deferred within specific intervals unless clinical failure is evident.

Drawing on a large Korean dataset, this study demonstrated that repeat AST frequency should be tailored to specific drug–bacterium combinations and time elapsed rather than following a “one-size-fitsall” approach. Laboratories can optimize resources by extending testing intervals for stable organisms that include S. aureus and E. faecalis, effectively implementing a “less is more” strategy [1]. Conversely, the rapid phenotypic adaptation of P. aeruginosa necessitates high-frequency monitoring and possible same-day testing from multiple critical sites. These findings may assist laboratories in refining repeat AST protocols by providing microbiological data that could support stratification based on microorganisms and, where feasible, initial susceptibility profiles. Although these findings support microorganism-based stratification for AST protocols, the long-term clinical impact of such optimized strategies warrants further validation.

This study has several limitations. Firstly, this study involved a single-center, retrospective analysis. Although the dataset was large, the findings may not be generalizable to other healthcare settings with different patient populations, antibiotic-prescribing practices, or local antimicrobial-resistance epidemiology. Secondly, while the clinical context is widely acknowledged as crucial for determining the necessity of repeat AST, especially in cases of persistent infection or high bacterial burden [3], our study primarily focused on patterns in microbiological data. Consequently, detailed clinical data, such as patient comorbidities, ongoing antimicrobial therapy, specific infection sites, clinical response to treatment, and reasons for repeat culture collection, were not included in our dataset. Thirdly, this study did not involve molecular typing; therefore, discerning whether the observed changes in susceptibility in “successive” isolates were due to the in-vivo evolution of resistance within the same bacterial strain or the acquisition of a new phenotypically similar but genotypically distinct strain was not possible. Furthermore, the reliance on an aggregate dataset lacking patient-level longitudinal tracking serves as a primary limitation, as it hindered our capacity to verify the clinical significance of observed transient spikes and implies that such fluctuations may reflect populationlevel variance rather than the true emergence of resistance within individual hosts.