Carbapenem-Resistant Klebsiella spp. in Healthcare-Associated Infections: a Narrative Review of Global Epidemiology, Resistance Mechanisms, and Clinical Outcomes (2020–2025)

1. Introduction

Healthcare-associated infections (HAIs) — defined as infections occurring in patients during the process of care in a hospital or other healthcare facility that were not present or incubating at the time of admission — represent one of the most significant and costly complications of modern medical practice. According to the third point prevalence survey coordinated by the European Centre for Disease Prevention and Control (ECDC), an estimated 4.3 million patients in EU/EEA acute care hospitals acquire at least one HAI annually, corresponding to a prevalence of 7.1% of hospitalised patients [1]. This burden translates into an estimated 2.6 million new cases and approximately 2.5 million disability-adjusted life years per year — a cumulative burden of 501 DALYs per 100,000 general population, exceeding that of influenza, tuberculosis and all other communicable diseases under ECDC surveillance combined; healthcare-associated pneumonia and primary bloodstream infection together account for more than 60% of this burden [2]. Critically, HAIs account for 71% of all infections caused by antibiotic-resistant bacteria, and up to 50% are considered preventable [3].

Among the diverse aetiological agents of HAIs, Gram-negative bacteria — and Klebsiella pneumoniae in particular — have assumed increasing clinical prominence over the past two decades. A member of the ESKAPE group of pathogens [4], K. pneumoniae is an opportunistic encapsulated bacillus capable of causing a spectrum of nosocomial infections including pneumonia, bloodstream infections (BSI), urinary tract infections (UTI), surgical site infections and meningitis [5].

The global burden attributable to this pathogen is substantial and disproportionately driven by resistance. In 2019, K. pneumoniae infections accounted for approximately 800,000 deaths worldwide, of which around 80% were associated with antimicrobial resistance; lower respiratory tract, bloodstream and intra-abdominal infections together accounted for nearly 90% of these deaths, with mortality driven predominantly by resistance to carbapenems and third-generation cephalosporins [6]. Successive Global Burden of Disease analyses of antimicrobial resistance place K. pneumoniae among the small group of pathogens responsible for the greatest share of resistance-associated mortality worldwide, and project a continued rise in that burden to 2050 [7]. This is reflected in current prioritisation frameworks: in the WHO Bacterial Priority Pathogens List 2024, carbapenem-resistant K. pneumoniae was the top-ranked pathogen overall, and carbapenem-resistant and third-generation cephalosporin-resistant Enterobacterales were assigned to the critical priority tier [8].

At European level, K. pneumoniae is the pathogen driving the deterioration in the antimicrobial resistance landscape. The estimated EU incidence of carbapenem-resistant K. pneumoniae bloodstream infections reached 3.51 per 100,000 population in 2024 (country range 0.02–20.31), 61.0% higher than in the 2019 baseline year and well above the 2030 target of 2.07 per 100,000, with a statistically significant increasing trend [9]. Resistance levels remain highest in southern, central and eastern Europe, where more than 35,000 deaths per year are attributed to infections with resistant bacteria across the EU/EEA [9]. Consistent with this, 32% of microorganisms recovered from microbiologically documented HAIs in the 2022–2023 ECDC point prevalence survey were antimicrobial-resistant [1].

The clinical threat posed by Klebsiella spp. has been dramatically amplified by the acquisition and horizontal transmission of mobile genetic elements encoding beta-lactamase enzymes. Extended-spectrum beta-lactamases (ESBLs) — primarily CTX-M enzymes — confer resistance to third-generation cephalosporins, while carbapenemases (notably KPC, NDM, OXA-48, VIM and IMP) render organisms resistant to last-resort carbapenem antibiotics [10]. These resistance traits, combined with co-transfer of genes encoding resistance to aminoglycosides, fluoroquinolones and colistin, have produced extensively drug-resistant and pandrug-resistant phenotypes for which therapeutic options are critically limited [11,12,13].

The epidemiological landscape has been further complicated by the emergence and global dissemination of hypervirulent K. pneumoniae (hvKP) strains into hospital environments [14]. More alarming still is the convergence of hypervirulence with carbapenem resistance into so-called CR-hvKP strains, which combine the immune evasion properties of hvKP with the near-untreatable resistance profile of CRKP. This dual-threat phenotype, predominantly associated with the ST11-K64 and ST11-KL47 clonal lineages, has been documented across multiple continents and is increasingly reported in intensive care settings [15,16].

The risk profile for CRKP acquisition is well characterised and largely modifiable. In a systematic review that compared CRKP cases against two different control groups, eight factors were associated with infection irrespective of the comparator chosen: admission to an intensive care unit, central venous catheter use, mechanical ventilation, tracheostomy, urinary catheter use, prior antibiotic use, and exposure to carbapenems and aminoglycosides. Prolonged length of hospital stay carried the single largest effect (OR 15.28), while prior hospitalisation, renal dysfunction, neurological disorders, nasogastric tube use, dialysis and exposure to quinolones, fluoroquinolones, glycopeptides and vancomycin were identified against one comparator only [19]. Earlier meta-analyses reported a broadly concordant profile [17,18]. The dependence of several estimates on the choice of control group is itself informative and explains part of the inconsistency between published risk-factor analyses [19].

These considerations are particularly relevant for Romania. According to the 2024 ECDC Surveillance Atlas of Infectious Diseases, Romania ranked third in Europe for carbapenem resistance among invasive K. pneumoniae isolates, with a resistance rate of 50.30%, surpassed only by Greece (60.20%) and Bulgaria (67.60%) [12,20]. Single-centre longitudinal data illustrate the trajectory and its therapeutic consequences: in a Romanian tertiary infectious diseases hospital, CRKP prevalence rose from 4.9% in 2010 to 41.6% in 2024, accompanied by an overall increase in antibiotic use — including a 186% rise in colistin consumption between 2019 and 2024 — while susceptibility of NDM plus OXA-48-like co-producing isolates was only 16.5% to colistin and 58.5% to cefiderocol [21]. The predominance of double carbapenemase producers distinguishes Romanian and other Eastern European isolates from Western European ones and further narrows therapeutic options [12,13]. This occurs against a background of high national antibiotic consumption [22], at a time when the EU population-weighted mean total consumption of antibacterials for systemic use was 20.3 DDD per 1,000 inhabitants per day in 2024 — 2% above the 2019 baseline and 4.4 DDD above the 2030 target of 15.9 [23].

Against this background, the present narrative review synthesises the evidence published between 2020 and 2025 on the global and regional epidemiology, molecular mechanisms, clinical outcomes and therapeutic management of Klebsiella spp. HAIs, with a focus on carbapenem-resistant strains, in order to inform infection control strategies, empirical treatment protocols and research priorities in high-prevalence settings such as Romania.

2. Literature Search Strategy

This article is a narrative review. It was not conducted as a systematic review and is therefore not reported according to the PRISMA 2020 statement; no protocol was registered and no new quantitative synthesis was performed. All pooled estimates presented below are cited from the published meta-analyses and surveillance reports in which they were originally calculated, and are attributed to those sources at the point of use.

PubMed/MEDLINE, Embase, Web of Science Core Collection and the Cochrane Central Register of Controlled Trials were searched for publications appearing between January 2020 and June 2025. The search was subsequently updated in July 2026 to incorporate key surveillance reports and meta-analyses published during manuscript preparation. The search combined Medical Subject Headings and free-text terms: (“Klebsiella pneumoniae” OR “Klebsiella oxytoca” OR “Klebsiella spp.”) AND (“healthcare-associated infection” OR “nosocomial infection” OR “hospital-acquired”) AND (“antimicrobial resistance” OR “carbapenem resistance” OR “ESBL” OR “KPC” OR “NDM” OR “OXA-48” OR “multidrug resistant”). Reference lists of retrieved articles and of relevant systematic reviews were screened for additional sources. Surveillance reports from the World Health Organization (including the GLASS database) and from the ECDC were consulted as grey literature.

Priority was given to systematic reviews and meta-analyses, multicentre cohort studies, national and supranational surveillance reports, and clinical guidelines. Publications were selected narratively by the authors on the basis of relevance to the themes addressed, methodological quality and recency; studies reporting exclusively community-acquired infections were not considered. Selection was not performed in duplicate, no formal risk-of-bias instrument was applied, and no assessment of publication bias or certainty of evidence was undertaken. These features are inherent to the narrative format and are addressed in Section 7.2.

3. Global and Regional Epidemiology of CRKP in Healthcare-Associated Infections

The most comprehensive quantitative estimate currently available is the systematic review and meta-analysis by Lin et al. [24], which pooled 61 studies comprising 513,307 patients with hospital-acquired K. pneumoniae infection from 14 countries and territories, published between 2008 and 2023. In that analysis, the global pooled prevalence of CRKP among nosocomial K. pneumoniae infections was 28.69% (95% CI: 26.53–30.86%), with very high statistical heterogeneity (I² = 99.8%). Regional estimates from the same source are summarised in Table 1.

At country level, the same analysis reported the highest prevalence in Greece (70.61%; 95% CI: 56.77–84.45%), followed by India (67.62%; 95% CI: 53.74–81.79%) and Taiwan (67.54%; 95% CI: 58.65–76.14%), while Germany, the United States, Tunisia and mainland China each reported rates below 20% [24]. Prevalence was substantially higher in countries with a low or middle socio-demographic index (29.77%) than in high-index countries (18.51%), and reached 62.31% in the intensive care subgroup, although the latter estimate derived from only two study populations and carries a correspondingly wide confidence interval [24].

European surveillance data complement this picture. The ECDC reports that the estimated EU incidence of carbapenem-resistant K. pneumoniae bloodstream infections rose by 61.0% between 2019 and 2024 [9], with a steep east–west gradient. Romania is among the most affected countries, at 50.30% carbapenem resistance among invasive isolates, behind Bulgaria (67.60%) and Greece (60.20%) [12,20]. Institutional Romanian series document the same trajectory locally, with CRKP prevalence rising from 4.9% in 2010 to 41.6% in 2024 in one tertiary infectious diseases hospital [21]; national figures reproduced in the Romanian literature indicate a rise in CRKP from 0% in 2007 to 47.8% in 2022 [13].

Two temporal observations frame these figures. Within the review period, single-centre Chinese data document measurable shifts in K. pneumoniae resistance patterns before, during and after the COVID-19 pandemic [28]. Over a longer horizon, the aetiology of healthcare-associated infection has moved from Gram-positive predominance towards multidrug-resistant Gram-negative organisms and emerging fungal pathogens, with carbapenem resistance in K. pneumoniae and Acinetobacter baumannii rising even as MRSA and Clostridioides difficile rates decline, and with low- and middle-income countries carrying a disproportionate Gram-negative burden [25].

4. Resistance Mechanisms and Molecular Epidemiology

Carbapenem resistance in Klebsiella spp. is driven predominantly by the acquisition of carbapenemase genes carried on mobile genetic elements, with KPC-type serine carbapenemases, NDM-type metallo-beta-lactamases and OXA-48-like oxacillinases accounting for the large majority of isolates worldwide; VIM and IMP enzymes represent minority genotypes [10,12]. The relative distribution of these enzymes is strongly geographically patterned rather than uniform, which is the single most consequential observation for empirical therapy. KPC predominates in the Americas, Italy, Greece and Israel; NDM is characteristic of the Indian subcontinent and increasingly of Eastern Europe; and OXA-48-like enzymes dominate in Turkey, North Africa and the Middle East [10,27].

The most detailed European evidence comes from the EURECA genomic survey, which characterised 687 carbapenem-resistant isolates from 41 hospitals in nine Southern European countries between 2016 and 2018 and compared them with the earlier EuSCAPE collection. blaKPC-like was the most prevalent carbapenemase gene, present in 46% of isolates, followed by blaOXA-48 in 39%, and eleven major clonal lineages were identified, most isolates belonging to the high-risk clones ST258/512, ST101, ST11 and ST307. Crucially, the pairing of gene and lineage was region-specific: blaKPC-like ST258/512 in Greece, Italy and Spain; blaOXA-48 ST101 in Serbia and Romania; blaNDM ST11 in Greece; and blaOXA-48-like ST14 in Türkiye [27].

Eastern European isolates are distinguished by the frequent co-production of two carbapenemases, most often NDM together with an OXA-48-like enzyme. This pattern has been documented repeatedly in Romanian tertiary hospitals [12,13,21] and in Ukraine, and contrasts with Western European isolates, in which single-enzyme production predominates [12,27]. Dual carbapenemase production is therapeutically decisive, because it simultaneously removes both ceftazidime-avibactam (inactive against metallo-beta-lactamases) and aztreonam monotherapy (hydrolysed by the co-produced serine enzymes) from the empirical armamentarium. In Greece, by contrast, KPC remains the most commonly encountered carbapenemase [12,26].

ESBL production, predominantly involving CTX-M-15 and CTX-M-27 enzymes, is highly prevalent among hospital-acquired Klebsiella isolates and frequently coexists with carbapenemase carriage and with plasmid-mediated AmpC enzymes [10,24]. Colistin resistance, including plasmid-mediated mcr determinants, has been reported in CRKP populations and is of particular concern where colistin remains a mainstay of therapy [10,11,21].

Molecular typing consistently identifies a limited number of high-risk clonal lineages responsible for the majority of CRKP isolates globally. ST258 and its derivative ST512, both predominantly KPC-carrying, have driven nosocomial outbreaks in Southern Europe, Israel and the Americas [12,13,26]. ST11 is the dominant clone in China, frequently carrying blaKPC-2 and increasingly associated with the acquisition of virulence plasmids that generate the CR-hvKP phenotype [29,31]; in one Chinese tertiary centre all blaKPC-2-positive isolates belonged to ST11, with predominant intra-hospital transmission of the ST11-KL64 subclone, whereas blaNDM-5-positive isolates were distributed across five distinct sequence types and showed evidence of both clonal and horizontal transmission [29]. ST101, predominantly associated with blaOXA-48-like, is prevalent in the Balkans, including Romania and Serbia, and has been linked to colistin resistance [12,27]. ST307 and ST147 have expanded across multiple continents, the latter carrying blaNDM and blaOXA-181 and being implicated in emerging ceftazidime-avibactam resistance [10,24,27]. Table 2 summarises the principal lineages and their carbapenemase associations.

5. Clinical Outcomes

5.1. Mortality

Carbapenem resistance is consistently associated with substantially worse survival. In the meta-analysis by Xu et al. [30], pooled mortality was 42.14% among 2,462 patients infected with CRKP, compared with 21.16% among patients infected with carbapenem-susceptible K. pneumoniae. Mortality varied markedly by infection site and by patient group, as summarised in Table 3. A subsequent meta-analysis of K. pneumoniae bacteraemia reported a pooled odds ratio for 30-day mortality of approximately 3.9 for CRKP compared with non-CRKP bacteraemia, with the excess mortality attributable to CRKP bacteraemia increasing over time [39].

Contemporary series are broadly consistent with these pooled estimates. A comparative analysis from a tertiary hospital in northern China confirmed significantly worse outcomes in carbapenem-resistant than in carbapenem-susceptible K. pneumoniae infection [32], while a nationwide Taiwanese multicentre study of CRKP bacteriuria illustrates the substantially more favourable prognosis of urinary tract involvement and the limited impact of antimicrobial therapy in that setting [33].

The additional mortality burden attributable to the convergence of carbapenem resistance with hypervirulence remains uncertain. In the meta-analysis by Chen et al. [15], which included 10 studies and 770 patients (224 with CR-hvKP and 546 with classical CRKP), the pooled odds ratio for mortality associated with CR-hvKP was 2.05 (95% CI: 0.89–4.75) — a trend towards higher mortality that did not reach statistical significance. Notably, the effect differed substantially according to how hypervirulence was defined, with a significant association in studies using phenotypic string tests (OR 4.16) but not in those using genotypic definitions (OR 1.05). Since most of the available evidence originates from East Asia, the global generalisability of these findings is limited, and the clinical impact of CR-hvKP should be regarded as biologically plausible but not yet quantitatively established.

5.2. Risk Factors for CRKP Acquisition

The most recent and largest synthesis is the meta-analysis by Jin et al. [11], which pooled 51 case-control and cohort studies comprising 13,860 patients (4,711 with CRKP infection and 9,149 with carbapenem-susceptible infection) published between 1991 and 2024, and examined 43 candidate risk factors, of which 31 were significantly associated with hospital-acquired CRKP infection. The strongest associations were found for antibiotic exposures — tigecycline (OR 5.97; 95% CI: 3.80–9.38), carbapenems (OR 4.79; 95% CI: 3.72–6.18), polymyxins (OR 4.25; 95% CI: 2.02–8.95) and glycopeptides (OR 3.40; 95% CI: 2.65–4.35) — and for the medical environment, notably intensive care admission (OR 4.27; 95% CI: 3.22–5.66) and a longer pre-infection hospital stay (mean difference 14.98 days). Among invasive procedures, bronchoscopy (OR 4.08; 95% CI: 1.40–11.92), tracheal cannulation (OR 3.72; 95% CI: 2.10–6.60), mechanical ventilation (OR 3.61; 95% CI: 2.72–4.78), central venous catheterisation (OR 3.39; 95% CI: 2.40–4.79) and dialysis (OR 3.00; 95% CI: 2.35–3.83) carried the highest risk; endoscopy and surgical drainage were newly characterised as distinct procedural risk factors in that analysis [11].

Two aspects of this synthesis merit emphasis. First, the reported associations should be read as markers of overall illness burden and healthcare exposure rather than as direct causal effects; the authors pooled unadjusted odds ratios, because heterogeneous confounder adjustment across the primary studies precluded meaningful pooling of adjusted estimates [11]. Second, and of direct relevance to the present review, subgroup analysis revealed systematic differences between Eastern and Western populations: cephalosporin (OR 2.68 vs 1.55) and fluoroquinolone exposure (OR 3.58 vs 1.89) were more strongly associated with CRKP infection in Western populations, whereas invasive procedures and polymyxin exposure carried substantially higher risk in Eastern populations (dialysis OR 4.47 vs 2.03; mechanical ventilation OR 4.22 vs 2.34; polymyxin OR 7.11 vs 3.02) [11]. This suggests that infection prevention priorities are themselves region-specific, favouring procedural control in Eastern settings and antimicrobial stewardship in Western ones.

Earlier meta-analyses reported a broadly concordant profile, implicating prolonged hospitalisation, tracheostomy, parenteral nutrition, immunosuppression and prior carbapenem exposure [17,18,19]. One notable discrepancy concerns underlying conditions: whereas Lin et al. [24] identified haematological malignancy as the strongest single predictor (OR 4.69; 95% CI: 2.80–7.88), Jin et al. [11] found no significant association for haematological malignancy, diabetes or hypertension, and instead identified respiratory disease (OR 2.69), kidney disease (OR 1.47), cardiovascular disease (OR 1.34) and male sex (OR 1.31) as significant. This divergence probably reflects differences in control group definition [19] and in the use of adjusted versus unadjusted estimates, and indicates that comorbidity-based risk stratification for CRKP remains unsettled. The practical implication is nonetheless consistent across syntheses: the dominant determinants — antibiotic exposure, intensive care admission and invasive devices — are modifiable.

5.3. Length of Stay and Resource Utilisation

Beyond mortality, CRKP infection imposes a substantial economic burden through prolonged hospitalisation, increased intensive care utilisation and the cost of salvage antimicrobial regimens. Health-economic modelling is informative in this respect: in a hybrid decision-tree and Markov model of CRKP bloodstream infection conducted from the Chinese healthcare perspective, ceftazidime-avibactam was found to be cost-effective as definitive therapy relative to polymyxin B monotherapy or polymyxin B-based combinations, against a willingness-to-pay threshold of USD 11,600 per quality-adjusted life-year [34]. The higher acquisition cost of newer agents may therefore be offset by reduced mortality and shorter hospitalisation, although such analyses are health-system specific and cannot be transferred directly to European settings.

6. Antimicrobial Treatment Landscape

6.1. Novel Beta-Lactam/Beta-Lactamase Inhibitor Combinations

Ceftazidime-avibactam is active against KPC- and OXA-48-type carbapenemases and is recommended in current IDSA and ESCMID guidance as the preferred agent for infections caused by strains producing these enzymes. Observational and cohort data indicate improved outcomes with ceftazidime-avibactam compared with polymyxin-based salvage regimens, including in high-risk neutropenic patients with KPC-producing Enterobacterales bacteraemia [35] and in single-centre retrospective series of CRKP infection [36].

Ceftazidime-avibactam has no direct activity against metallo-beta-lactamases. For NDM-producing strains, the combination of aztreonam with ceftazidime-avibactam exploits the stability of aztreonam to metallo-enzymes together with avibactam's inhibition of co-produced serine carbapenemases and ESBLs, and is the current combination of choice [10,35]. This distinction is directly relevant to Romanian practice, where NDM plus OXA-48-like co-production predominates [12,13,21].

Emergence of resistance to ceftazidime-avibactam, mediated predominantly by KPC mutations such as D179Y or by a shift towards metallo-beta-lactamase production, has been documented following exposure and represents a critical therapeutic vulnerability requiring active surveillance [37]. Cefiderocol offers an additional option, although susceptibility among Romanian NDM plus OXA-48-like co-producing isolates has been reported at only 58.5% [21].

6.2. Last-Resort and Adjunctive Options

Colistin and polymyxin B remain last-resort options for extensively drug-resistant CRKP infections, particularly where metallo-beta-lactamase producers predominate and newer agents are unavailable [38]. Their utility is constrained by nephrotoxicity, unfavourable pharmacokinetics and rising resistance: in a Romanian series, susceptibility of NDM plus OXA-48-like co-producing isolates to colistin was only 16.5%, against a background of a 186% increase in colistin consumption between 2019 and 2024 [21]. Tigecycline, fosfomycin and trimethoprim-sulfamethoxazole retain partial activity against some CRKP strains but are generally reserved for urinary tract infection or de-escalation because of pharmacodynamic limitations [16].

7. Discussion

7.1. Interpretation of the Evidence

Three themes emerge consistently from the evidence reviewed. First, the burden of carbapenem resistance in healthcare-associated Klebsiella infection is high globally — approximately 29% of nosocomial K. pneumoniae infections in the most recent pooled estimate [24] — but the global figure conceals a nearly five-fold gradient between high-income North America and South Asia. Second, carbapenem resistance approximately doubles crude mortality relative to susceptible infection, with bloodstream infection carrying the worst prognosis [30,39]. Third, and most consequential for practice, the molecular architecture of resistance differs fundamentally between regions, so that the correct empirical regimen in Greece is not the correct regimen in Romania or in India.

The Romanian situation exemplifies this last point. National surveillance places carbapenem resistance among invasive isolates at 50.30% [12,20], and institutional data document a predominance of NDM plus OXA-48-like co-producers [12,13,21] — precisely the combination against which ceftazidime-avibactam monotherapy fails and for which aztreonam-containing regimens or cefiderocol are required. Reported cefiderocol susceptibility of 58.5% and colistin susceptibility of 16.5% in this population [21] indicate that the therapeutic margin is already narrow. This argues strongly for routine carbapenemase typing rather than phenotype-only reporting in Romanian hospitals.

Two areas warrant particular caution in interpretation. The clinical impact of CR-hvKP, although biologically plausible and widely discussed, is not yet established quantitatively: the only dedicated meta-analysis found a non-significant pooled association with mortality and a strong dependence of the effect on the definition of hypervirulence used [15]. Similarly, the evidence supporting newer beta-lactam/beta-lactamase inhibitor combinations derives largely from observational cohorts rather than randomised trials, and is therefore susceptible to confounding by indication.

7.2. Limitations

This review has important limitations. It is a narrative rather than a systematic review: study selection was not performed in duplicate, no formal risk-of-bias instrument was applied, and no assessment of publication bias or of certainty of evidence was undertaken. Selection of sources therefore reflects the authors' judgement of relevance and may be subject to selection bias, and the synthesis should not be read as an exhaustive account of the literature.

The evidence synthesised also carries its own limitations. Statistical heterogeneity in the underlying meta-analyses was extreme (I² = 99.8% in Lin et al. [24]), reflecting genuine differences in case mix, surveillance intensity, susceptibility testing methodology and carbapenem breakpoint adoption; pooled prevalence figures should therefore be read as an indication of overall burden rather than as generalisable point estimates. Publication bias was formally detected in that analysis (Egger test, p = 0.001) [24]. Geographic coverage of the primary literature is uneven, with sparse representation of Africa, Latin America and much of Eastern Europe, so regional estimates for these areas rest on few studies; notably, the widely cited Western European estimate of 42.05% derives from 13 studies dominated by Italian data and is not representative of Northern Europe. The principal mortality synthesis [30] searched the literature only to December 2015 and predates both the widespread introduction of ceftazidime-avibactam and the post-2020 epidemiological shifts, so current case fatality may differ. Almost all underlying studies were observational, so residual confounding by comorbidity, illness severity and appropriateness of empirical therapy cannot be excluded and reported mortality differences should not be interpreted as fully attributable mortality; the principal risk-factor synthesis pooled unadjusted odds ratios for precisely this reason [11]. Finally, definitions of healthcare-associated infection and the extent of molecular characterisation varied considerably between sources.

7.3. Implications for Practice, Policy and Research

For clinical practice, the marked regional variation in carbapenemase distribution means that empirical therapy for suspected CRKP infection cannot be standardised internationally and must be guided by local molecular epidemiology, supported by rapid carbapenemase detection. For policy, the concentration of risk in modifiable exposures — carbapenem, quinolone and cephalosporin use, intensive care admission, invasive devices — supports investment in antimicrobial stewardship and device-associated infection prevention as the interventions with the greatest expected yield, alongside expansion of molecular surveillance capacity in high-prevalence Eastern European settings. For research, priorities include prospective multicentre cohorts from under-represented regions, standardised outcome and hypervirulence definitions to reduce heterogeneity, systematic surveillance of emerging resistance to ceftazidime-avibactam and cefiderocol, and randomised evidence in patients infected with dual-carbapenemase-producing strains, for whom current recommendations rest on the weakest evidence base.

8. Conclusions

Carbapenem-resistant Klebsiella spp. account for approximately one quarter to one third of healthcare-associated Klebsiella infections worldwide and are associated with roughly a doubling of crude mortality relative to carbapenem-susceptible infection. The burden is concentrated in South Asia, Southern and Eastern Europe and the Middle East, and is driven by regionally distinct carbapenemase profiles, with an accelerating trend towards dual carbapenemase production in Eastern Europe and the convergence of resistance with hypervirulence in East Asia. Romania, with carbapenem resistance exceeding 50% among invasive isolates and a predominance of NDM plus OXA-48-like co-producers, exemplifies the setting in which generic empirical recommendations fail. Locally adapted surveillance incorporating routine carbapenemase typing, rigorous antimicrobial stewardship and the rational deployment of novel beta-lactam/beta-lactamase inhibitor combinations are urgently required to contain this threat.