Abstract
BACKGROUND: Carbapenem-resistant Klebsiella spp. is a major global health threat. Overuse of carbapenem in healthcare settings, particularly ICUs, has driven the rise of these resistant bacteria. This study aimed to apply dynamic regression forecasting to predict future trends of carbapenem-resistant Klebsiella spp., with analysis of the temporal relationship between carbapenem consumption and resistance serving as the foundation for building accurate forecasts to inform antimicrobial stewardship (AMS) strategies. METHODS: This retrospective study analyzed hospital- and ICU-level carbapenem consumption and resistance data for 1,727 Klebsiella spp. isolates (1,043 ICU) from a tertiary care hospital in Alexandria, Egypt (2019–2023). Cross-correlation analyses identified lag structures between consumption and resistance, and these lagged consumption measures — hospital carbapenem use expressed as defined daily doses per 1,000 patient-days (DDD/1000PD) per quarter and ICU use expressed as days of therapy per 1,000 patient-days (DOT/1000PD) per semester —were incorporated as exogenous regressors in dynamic regression time-series models to forecast future resistance trends. RESULTS: Carbapenem resistance exhibited by Klebsiella spp. in the hospital significantly increased from 30.2% (95% CI: 20.6% to 39.8%) in quarter 1 of 2020 to 57.5% (95% CI: 49.3% to 65.7%) in quarter 4 of 2023, accompanied by a parallel increase in the burden of resistant isolates and a substantial rise in carbapenem consumption from 30 DDD/1000PD in quarter 1 of 2019 to 131 DDD/1000PD in quarter 4 of 2023. Similarly, in the ICU, carbapenem resistance jumped from 25.8% (95% CI: 18.5% to 33.1%) in semester 1 of 2019 to 67.0% (95% CI: 58.6% to 75.3%) in semester 2 of 2023, accompanied by an increase in the burden of resistant isolates and by a surge in carbapenem consumption from 196 DOT/1000PD in semester 1 of 2019 to 473 DOT/1000PD in semester 4 of 2023. Strong correlations were found between one lagged period of carbapenem consumption and resistance rates (r = 0.80 for hospital, r = 0.86 for ICU). A marked increase in hospital carbapenem consumption was observed following the onset of the COVID-19 pandemic, with a post-pandemic level change of 61.82 DDD/1000 patient-days. Dynamic regression forecasting predicted stabilization of carbapenem resistance at approximately 50% at the hospital level (95% prediction interval [PI]: 32.6–67.1), while ICU-level models projected a continued upward trajectory in resistance over time (95% PI: 31.3–97.6). CONCLUSIONS: Dynamic regression forecasting revealed persistently high carbapenem resistance in Klebsiella spp. at the hospital level and a continued rising trend in the ICU, occurring in temporal association with carbapenem use. These projections support antimicrobial stewardship by informing trend awareness and highlighting the need for targeted interventions and sustained surveillance to mitigate further escalation. SUPPLEMENTARY INFORMATION: The online version contains supplementary material available at 10.1186/s12879-026-13168-y.