Incidence, clinical and genomic trends of hospital- and Non-hospital-onset KPC-producing Klebsiella pneumoniae infections before and during the COVID-19 era: a ten-year interrupted time series study

Abstract

Background

Infections caused by KPC-producing Klebsiella pneumoniae (KPC-KP) represent a persistent public health challenge. This prospective study examines ten-year trends, clinical features, and genomic epidemiology of hospital-onset (HOI) and non-hospital-onset (non-HOI, including healthcare-associated [HcAI] and community-acquired [CA]) KPC-KP infections following a 2012 outbreak. We evaluated the impacts of a 2014 antimicrobial stewardship program (ASP) and COVID-19-related infection prevention and control (IPC) measures, with emphasis on hospital-to-community dissemination.

Methods

We analysed a prospective, longitudinal cohort of patients (2012–2022) in a tertiary referral hospital. Interrupted time series and ARIMA models assessed ASP and IPC impacts on incidence density (ID). Cross-correlation analysis explored temporal associations between HOI and non-HOI trends. Whole-genome sequencing and PERMANOVA evaluated the genomic structure of ST512/KPC-3 isolates. Multivariable regression analysed the association between infection type and clinical outcomes.

Results

Among 467 patients, 33.2% had non-HOI (ID 0.53/1,000 admissions/month) and 66.8% HOI (ID 0.30, p = 0.39). Urinary tract infections predominated in non-HOI (52.9%), while bloodstream and respiratory infections were more common in HOI. Incidence density of HOI and non-HOI infections declined significantly following ASP implementation, with a 4-month lag suggesting sequential transmission dynamics. These reductions were maintained during the pandemic. Genomic data confirmed ST512/KPC-3 dominance and hospital-to-community spread, with temporal factors—rather than acquisition type—explaining genetic variation. Adjusted analyses showed similar 30-day mortality and treatment responses across HOI and non-HOI.

Conclusions

ASP and COVID-19 IPC measures contributed to maintaining low KPC-KP incidence. Genomic evidence underscores the role of temporal dynamics and clonal expansion in ST512/KPC-3 dissemination. Non-HOI infections are clinically significant and require targeted, system-wide surveillance and control strategies.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13756-025-01614-6.

Background

The dissemination of Klebsiella pneumoniae strains producing carbapenemases (KPC) has emerged as a critical issue in recent years, epidemiologically and clinically. KPC-producing Klebsiella pneumoniae (KPC-KP) strains are capable of hydrolysing carbapenems and often exhibit resistance to multiple antibiotics, posing significant challenges for the treatment of infections caused by these pathogens [1–3]. A substantial percentage of hospitalized patients are colonized by these microorganisms, which have been implicated in numerous outbreaks of severe nosocomial infections, such as bacteremia and ventilator-associated pneumonia, both associated with high mortality rates [1–4]. Colonization by KPC-KP increases the risk of subsequent infections and adverse outcomes, particularly in vulnerable patient populations [3, 5]. The persistence of KPC-KP carriage in hospital, healthcare and community settings represents a considerable epidemiological concern.

The COVID-19 pandemic has placed unprecedented strain on healthcare systems, leading to disruptions in routine practices that may influence the incidence of infections in both nosocomial and community environments. Notably, the pandemic has resulted in increased hospital admissions, particularly among patients with complex conditions requiring intensive care unit (ICU) admission, invasive procedures, and prolonged hospital stays. These factors, combined with shortages of healthcare personnel and protective equipment during the early stages of the pandemic, may have compromised infection prevention and control (IPC) measures. Additionally, concerns have been raised regarding elevated antimicrobial use during the pandemic, driven by empirical antibiotic prescribing even in the context of low rates of secondary bacterial coinfections [6–8]. Recent studies indicate that up to 75% of hospitalized COVID-19 patients received one or more antibiotics, with β-lactams, fluoroquinolones, and macrolides being the most prescribed classes [6, 7]. Furthermore, the increased use of broad-spectrum antibiotics, such as cefepime, piperacillin/tazobactam, carbapenems, and other last-resort agents like colistin and ceftazidime/avibactam, has been reported during pandemic peaks [9, 10]. These trends, coupled with the emergence of novel genomic lineages of carbapenemase-producing Enterobacterales [10–13], underscore the need for enhanced IPC practices and antimicrobial stewardship programs (ASPs) to mitigate the impact of antimicrobial resistance (AMR).

In 2012, the first significant outbreak of the KPC3-producing Klebsiella pneumoniae Sequence Type (ST) 512 strain occurred in a tertiary referral hospital in the South of Spain, following importation of a patient from Italy, where this clone was endemic [14]. In this study, we aimed to analyze a comprehensive, prospective, longitudinal cohort of patients with KPC-KP infections treated from 2012 to 2022 in this hospital. We assessed the incidence density and clinical and molecular epidemiology of hospital-onset (HOI) versus non-hospital onset (non-HOI) infections, including healthcare-associated (HcAI) and community-acquired (CA) cases. Additionally, we evaluated the impact of an ongoing antimicrobial stewardship program (ASP) and COVID-19 infection prevention and control (IPC) measures on infection trends using interrupted time series analysis (ITSA). These data are critical to inform national policies and clinical care.

Methods

Study design, clinical data and isolate collection

The KAPECOR cohort is a longitudinal, prospective, observational cohort of hospitalized adult patients at Reina Sofia University Hospital (HURS), a large teaching care center in southern Spain. The cohort includes patients with microbiologically confirmed infections by Klebsiella pneumoniae producing KPC carbapenemase (KPC-KP), and was established following the initial outbreak caused by the ST512/KPC-3 clone in 2012 [14].

For this study, clinical, demographic, epidemiological, and microbiological data were extracted from electronic health records (EHR), microbiology laboratory databases, and hospital pharmacy records. Additional information- such as prior hospitalizations, recent treatments (e.g., surgery, chemotherapy, antibiotics), and clinical outcomes- was obtained through HER review and, when necessary, confirmed via telephone interviews with patients or their relatives. Data collection followed a structured protocol using predefined case report forms maintained by the research team.

We included all consecutive cases of KPC-KP infections between June 2012 and April 2022, excluding colonization and reactivations within 30 days of a previous infection (Fig. 1). Infection episodes were defined according to the criteria of the European Centre for Disease Prevention and Control (ECDC) [15].

Fig. 1.

Study flow diagram. KPC-KP, KPC-producing Klebsiella pneumoniae; HOI, hospital-onset infections; non-HOI, non-hospital-onset infections; HcAI, healthcare-associated infections; CA, community-acquired infections. *Reactivation of a KPC-KP infection within 30 days of the first diagnosis

Infections were classified as non-HOI if the clinical sample was obtained before hospital admission (including the emergency department) or within the first 48 h of admission. Non-HOI infections were further categorized as healthcare-associated (HcAI) if the patient met one or more of the following criteria [16–18]: (i) hospitalization for ≥ 48 h, surgery, or dialysis within the previous 90 days; (ii) residence in a healthcare facility other than an acute care hospital; or (iii) antibiotic treatment, chemotherapy, or wound care within the previous 30 days. Cases that did not meet these criteria were classified as community-acquired (CA). All available KPC-KP isolates from these episodes were stored in an institutional microbial repository.

The study followed STROBE reporting guidelines (Supplementary Table S1).

Clinical variables

Patients were followed for 30-day clinical outcomes, including all-cause mortality, clinical response, and microbiological response, as well as for 90-day recurrence or death.

The primary outcome was 30-day all-cause mortality. Secondary outcomes included: (i) 30-day clinical response, defined as resolution or improvement of symptoms present at diagnosis, allowing for antibiotic discontinuation; ii) 30-day microbiological response, defined as negative follow-up cultures for KPC-KP from initially positive sites (excluding colonization samples); and (iii) 90-day recurrence, defined as a new episode of symptomatic infection by KPC-KP, confirmed by culture, occurring more than 30 days after the initial infection, either in the same or a different anatomical site.

All outcomes were assessed based on EHR documentation and microbiological results. Additional variables analyzed included demographic characteristics, pre-existing comorbidities, and risk factors. The Charlson Comorbidity Index [19] was used to assess baseline comorbidity burden. Other factors included prior colonization with KPC-KP, and clinical severity at the time of infection. Severity was assessed using the presence of septic shock [20], the McCabe risk score [21], and the INCREMENT-CPE mortality risk score [22].

Incidence-trend analysis: impact of ASP and IPC interventions

We conducted a descriptive analysis of the incidence density (ID), defined as the number of cases per 1,000 admissions per month, for both non-HOI infections and HOI infections. Temporal trends for these infection types were also compared.

Following the outbreak, a hospital-wide antimicrobial stewardship program (ASP) was implemented in September 2014 [23] (details summarized in Supplementary Tables S2-S3). During this period, infection prevention and control (IPC) measures—including contact precautions, environmental cleaning protocols, and hand hygiene promotion—remained stable and unchanged compared to the pre-ASP period.

In 2020, a comprehensive set of infection prevention and control (IPC) interventions was implemented at both hospital and community levels in response to the COVID-19 pandemic (summarized in Supplementary Table S4). The impact of these interventions was evaluated by comparing the monthly ID of KPC-KP HOI and non-HOI infections across three distinct periods: the pre-intervention period (March 2012-August 2014), the post-ASP period (September 2014-February 2020), and the post-COVID period (March 2020-April 2022).

Microbiological procedures

Initial biochemical identification and antimicrobial susceptibility testing of clinical isolates were performed using the Microscan WalkAway system (Beckman Coulter, panels NC53 or NC82). Species identification was confirmed using MALDI-TOF mass spectrometry (MALDI Biotyper, Bruker Daltonics GmbH & Co.). Additional susceptibility testing was carried out with Sensititre Gram Negative microdilution panels (Thermo Fisher Scientific, Madrid, Spain). Clinical categories for antimicrobial agents were interpreted according to EUCAST clinical breakpoints. KPC production was confirmed using the Xpert Carba-R assay (Cepheid, Sunnyvale, CA, USA) or the NG-Test CARBA 5 test (NGBiotech, Guipry, France). KPC alleles were identified by PCR followed by Sanger sequencing.

Whole genome sequencing analysis

Total DNA was extracted from all strains using the QIAamp DNA mini kit (Qiagen, Hilden, Germany). Paired-end genomic DNA libraries were generated using the Nextera XT DNA Sample Preparation Kit and sequenced on the Illumina HiSeq 500 platform, generating 2 × 150 bp paired-end reads (Illumina Inc., San Diego, CA, USA). Sequence data that support the findings of this study have been deposited in the European Nucleotide Archive with the primary accession code PRJEB66458.

The quality of the short reads was assessed using FastQC, followed by trimming with BayesHammer. Trimmed reads were assembled into contigs using Unicycler 0.5.0, and the quality of the resulting assemblies was evaluated with QUAST. Genome annotation was performed automatically using Prokka v1.14-beta.

Phylogenetic analysis was conducted using SNP-based methods implemented in PhaME, with a KPC-KP isolate from the 2012 outbreak index patient serving as the reference. The phylogenetic tree was generated with FastTree and visualized using the ggtree R package.

Statistical analysis

The incidence density (ID) of KPC-KP infections in the HOI and non-HOI groups was compared using linear mixed models with repeated measures. Interrupted time series analysis (ITSA) was conducted with a first-order autoregressive-moving-average model to assess changes in level and trend attributable to the interventions, as well as the absolute and relative effects of both interventions [23–25]. The relationship between the IDs of HOI and non-HOI infections was further analysed using the cross-correlation function of autoregressive integrated moving average (ARIMA) models.

For clinical variables, descriptive statistics were used, with proportions calculated for qualitative variables and median values (with interquartile ranges) for quantitative variables. Differences between nosocomial and non-nosocomial infections were assessed using the Chi-square test for categorical variables and either the Student´s t-test or Wilcoxon for continuous variables. The association between clinically relevant variables and clinical outcomes was analyzed using multivariable Cox regression or logistic regression models, as appropriate.

For genomic variables, a phylogenetic distance matrix was constructed. The genomic composition of the KPC-KP isolates was visualized using a principal coordinate analysis (PCoA) based on cophenetic distances derived from the phylogenetic tree. Permutational multivariate analysis of variance (PERMANOVA) was then performed using the adonis function from the vegan R package, with 2,000 permutations, to assess the marginal effects of acquisition type and sampling year on the variation in genomic composition among KPC-KP isolates.

Statistical analyses were performed using R software (version 3.0.1) with the cmprsk package for competitive risk analysis, SPSS version 26.0 (SPSS Inc.), and Salford Predictive Modeller software version 8.2, which includes CART for dichotomization of clinically relevant variables (Supplementary Figure S1). A two-tailed P value of less than 0.05 was considered statistically significant.

Results

Clinical epidemiology

During the 10-year study period, 467 cases of KPC-KP infections were recorded, of which 312 (66.8%) were hospital-onset infections (HOI), 121 (25.9%) healthcare-associated infections (HcAI) and 34 (7.3%) community-acquired infections (CA) (Fig. 1). The most frequent source was the urinary tract (34.7%), followed by bacteriemia (22.9%), respiratory tract (20.1%), and surgical site infections (13.3%) (Supplementary Table S5). Thirty-day all-cause mortality was 34.9% for HOI, 28.1% for HcAI, and 32.4% for CA cases (Supplementary Table S5, p = 0.396).

For the main analysis, HcAI and CA infections were grouped as non-hospital-onset infections (non-HOI) (Fig. 1, Table 1), based on their similar clinical profiles and outcomes in our cohort (Supplementary Tables S5, S6, S7). Table 1 compares HOI and non-HOI infections. Urinary tract infections (UTIs) were significantly more common in non-HOI (52.9%), while bloodstream infections (BSIs) and respiratory infections were more frequent in HOI (27.6% and 23.4%, respectively, p < 0.05). Patients with non-HOI were older (median age 79 vs. 68 years, p < 0.001), had higher Charlson Comorbidity Index scores (median 6 vs. 5, p = 0.012), and more often received active targeted therapy (92.3% vs. 81.7%, p = 0.003).

Table 1.

Comparison of non-hospital onset (non-HOI) versus hospital-onset (HOI) infections caused by KPC-KP

Characteristic	Overall N = 4671	Non-HOI N = 1551	HOI N = 3121	p-value2
Age, years	73 (59.0–82)	79 (70.0–85)	68 (57.0–79)	< 0.001
Female gender	208.0 (44.5%)	79.0 (51.0%)	129.0 (41.3%)	0.049
Hospitalization (days)	26 (13.0–45)	16 (10.0–33)	29 (16.0–52)	< 0.001
Risk factors
Charlson score	5 (3.0–7)	6 (4.0–7)	5 (3.0–7)	0.012
McCabe risk score				< 0.001
Nonfatal	142.0 (30.4%)	31.0 (20.0%)	111.0 (35.6%)
Rapidly fatal	202.0 (43.3%)	70.0 (45.2%)	132.0 (42.3%)
Ultimately fatal	123.0 (26.3%)	54.0 (34.8%)	69.0 (22.1%)
Steroid use (> 20 mg/day)	234.0 (50.1%)	62.0 (40.0%)	172.0 (55.1%)	0.002
Surgery, previous 3 months	166.0 (35.5%)	35.0 (22.6%)	131.0 (42.0%)	< 0.001
Prior KPC-KP colonization	188.0 (40.3%)	67.0 (43.2%)	121.0 (38.8%)	0.356
Hospitalization, previous 6 months	259.0 (55.5%)	93.0 (60.0%)	166.0 (53.2%)	0.164
Carbapenems, previous month	105.0 (22.5%)	20.0 (12.9%)	85.0 (27.2%)	< 0.001
Intensive care unit admission	64.0 (13.7%)	3.0 (1.9%)	61.0 (19.6%)	< 0.001
Infection
Type
Urinary tract	162.0 (34.7%)	82.0 (52.9%)	80.0 (25.6%)	< 0.001
Bacteriemia	107.0 (22.9%)	21.0 (13.5%)	86.0 (27.6%)	< 0.001
Catheter-related	11.0 (2.4%)	1.0 (0.6%)	10.0 (3.2%)	0.110
Unknown source	96.0 (20.6%)	20.0 (12.9%)	76.0 (24.4%)	0.004
Respiratory	94.0 (20.1%)	21.0 (13.5%)	73.0 (23.4%)	0.012
Surgical site	62.0 (13.3%)	16.0 (10.3%)	46.0 (14.7%)	0.185
Skin and soft tissue	26.0 (5.6%)	13.0 (8.4%)	13.0 (4.2%)	0.061
Intraabdominal	13.0 (2.8%)	1.0 (0.6%)	12.0 (3.8%)	0.069
Others	3.0 (0.6%)	1.0 (0.6%)	2.0 (0.6%)	> 0.999
Septic shock	64.0 (13.7%)	16.0 (10.3%)	48.0 (15.4%)	0.134
INCREMENT-CPE mortality risk score	6 (5.0–8)	6 (5.0–8)	6 (5.0–8)	0.018
Antibiotic therapy
Adequate empiric therapy	121.0 (25.9%)	32.0 (20.6%)	89.0 (28.5%)	0.067
Duration of empiric therapy (days)	6 (2.0–11)	5 (2.0–9)	6 (2.0–12)	0.030
Active targeted therapy	398.0 (85.2%)	143.0 (92.3%)	255.0 (81.7%)	0.003
Time from culture to targeted therapy (days)	3 (1.0–6)	3 (1.0–7)	3 (1.0–5)	0.077
Duration of targeted therapy (days)	8 (6.0–13)	8 (6.0–12)	9 (5.0–14)	0.765
Use of ceftazidime-avibactam	88.0 (18.8%)	34.0 (21.9%)	54.0 (17.3%)	0.228
Combined targeted therapy	127.0 (27.2%)	40.0 (25.8%)	87.0 (27.9%)	0.635
Duration of adequate therapy (days)	9 (6.0–14)	8 (6.0–12)	10 (6.0–15)	0.115
30-day clinical response	245.0 (52.5%)	82.0 (52.9%)	163.0 (52.2%)	0.893
30-day microbiological response	244.0 (52.2%)	81.0 (52.3%)	163.0 (52.2%)	0.998
30-day all-cause mortality	154.0 (33.0%)	45.0 (29.0%)	109.0 (34.9%)	0.201
90-day recurrence	46.0 (9.9%)	26.0 (16.8%)	20.0 (6.4%)	< 0.001

¹n (%); Median (Q1-Q3)

²Wilcoxon rank sum test; Pearson's Chi-squared test; Fisher's exact test

Within the non-HOI group (Supplementary Table S6), HcAI patients were younger than those with CA infections (median 77 vs. 81 years, p = 0.014), more likely to have received steroids (45.5% vs. 20.6%, p = 0.009), and to have undergone surgery in the previous 3 months (27.3% vs. 5.9%, p = 0.008). Notably, 43.8% of HcAI cases involved patients from health and social care centers.

Clinical outcomes showed no significant differences between HOI and non-HOI in terms of 30-day mortality (29.0% for non-HOI), clinical response (52.9%), or microbiological response (52.3%) (Table 1). However, 90-day recurrence was higher in non-HOI (16.8% vs. 6.4%, p < 0.001). No significant differences in mortality or recurrence were observed when comparing HcAI and CA within the non-HOI group (Supplementary Table S6). Kaplan–Meier curves and multivariable analyses confirmed no significant differences in adjusted mortality or clinical and microbiological response between HOI and non-HOI infections (Table 2, Supplementary Figure S2, Supplementary Tables S8–S10).

Table 2.

Adjusted association of non-hospital-onset (non-HOI) infections with clinical outcomes

Outcome variables at day 30	Adjusted risk (95% CI)	p value
Crude mortality*	0.89 (0.63–1.27)	0.522
Clinical response †	0.97 (0.66–1.44)	0.889
Microbiological response †	0.92 (0.62–1.37)	0.684

^*Crude mortality, hazard ratio (Cox regression)

^†Clinical and microbiological response, relative risk (logistic regression)

For complete analysis, see Supplementary Tables S8, S9 and S10

Incidence-trend analysis and impact of ASP and IPC interventions

In a previous study, we demonstrated that the antimicrobial stewardship program (ASP) implemented in 2014 significantly impacted on the reduction of nosocomial KPC-KP infections up to February 2020, prior to the onset of COVID-19 pandemic [23, 26]. In this period, infection prevention and control (IPC) measures remained unchanged compared to the pre-ASP phase, providing a stable baseline against which to evaluate the ASP’s effect. Building on this, the current analysis expanded the interrupted time-series approach to assess whether the additional IPC interventions introduced during the COVID-19 pandemic had any further impact on nosocomial infection trends. Furthermore, we extended the scope to evaluate the combined effects of ASP and COVID-19 IPC measures on the incidence and spread of non-nosocomial infections, encompassing HcAI and CA cases.

The incidence density (ID) of non-HOI infections was 0.30 cases per 1,000 admissions/month (range 0.0–0.60), compared to 0.53 for HOI infections (range 0.06–0.90, p = 0.39) (Fig. 2). Interrupted time series analysis (ITSA) revealed a significant reduction in the ID of KPC-KP infections following the introduction of ASP in both groups, with these decreases sustained throughout the study period (Table 3, Fig. 3). Hospital-onset infections showed a relative reduction of −99.7% (95% CI −102.8% to −96.5%) eight years after the ASP implementation, equivalent to an absolute reduction of −11.92 cases per 1,000 admissions/month (95% CI −14.19 to −9.64). Similarly, non-HOI infections experienced a relative reduction of −97.3 (95% CI −101.1% to −93.5%), with an absolute decrease of −4.59 cases per 1,000 admissions/month (95% CI −5.67 to −3.51) by the end of the study (Table 3). In contrast, COVID-19 IPC measures did not significantly alter the KPC-KP trends in either group, suggesting that the reductions observed were primarily driven by the earlier ASP intervention (Table 3, Fig. 3). Comparable patterns emerged when HOI and HcAI infections were compared to CA infections (Supplementary Table S11, Supplementary Figure S3).

Fig. 2.

Incidence density (ID) of hospital-onset (HOI) and non-hospital onset (non-HOI) infections before and after implementation of 2014 antimicrobial stewardship program (ASP) and 2020 COVID-19 infection prevention and control measures (IPC, dotted lines)

Table 3.

Interrupted time-series analysis (ITSA) of changes in incidence density trends of hospital onset and non-hospital onset KPC-KP infections

A) Hospital-onset infections (HOI)
	Comparison of changes in the indicated period relative to the previous period a		Changes at the end of the study relative to the indicated period d
Period	Level change b (95% CI)	Trend change c (95% CI)	Absolute change (95% CI)	Relative change (95% CI)
1. Pre-ASP (Jun 2012-Aug 2014)	0.374 (−0.001; 0.749)	0.095 (0.074; 0.116)	−11.918 (−14.192; −9.644)	−99.7 (−102.8; −96.5)
2. ASP (Sep 2014-Feb 2020)	−2.644 (−3.067; −2.221)	−0.099 (−0.121; −0.077)	−0.153 (−0.701;0.394)	−78.9 (−278.2; 120.4)
3. COVID-19 IPC (Mar 2020-Apr 2022)	0.369 (−0.086; 0.824)	−0.020 (−0.047; 0.006)	-
B) Non-hospital onset infections (non-HOI)
	Comparison of changes in the indicated period relative to the previous period a		Changes at the end of the study relative to the indicated period d
Period	Level change b (95% CI)	Trend change c (95% CI)	Absolute change (95% CI)	Relative change (95% CI)
1. Pre-ASP (Jun 2012-Aug 2014)	0.032 (−0.135; 0.199)	0.038 (0.028; 0.048)	−4.591 (−5.669; −3.514)	−97.3% (−101.1; −93.5)
2. ASP (Sep 2014-Feb 2020)	−0.744 (−0.939; −0.549)	−0.042 (−0.051; −0.032)	−0.023 (−0.228; 0.182)
3. COVID-19 IPC (Mar 2020-Apr 2022)	0.075 (−0.149; 0.299)	−0.004 (−0.017; 0.009)	-

Level change and absolute change are presented as monthly defined daily doses (DDDs) per 1,000 occupied bed days (OBDs). Trend change is measured in per-unit terms. Relative change is presented as percentage

ASP Antimicrobial Stewardship Program, CI confidence interval, IPC Infection Prevention and Control intervention

a Level and trend changes in the first period correspond to the regression value at t = 0 (intercept) and the initial trend, respectively

b Increase or decrease in the first month of the period relative to the expected value based on the prior trend

c Change in slope for the period relative to the previous trend

d Absolute or percentage difference between the expected value according to the period trend and the regression value at the end of the study

Fig. 3.

Interrupted time series analysis (ITSA) evaluating the impact of the antimicrobial stewardship program and COVID-19 infection prevention and control measures on the incidence density of KPC-producing Klebsiella pneumoniae infections: A) Non-hospital-onset (non-HOI) and B) hospital-onset (HOI) infections. Data are presented as monthly isolates per 1000 occupied bed days (OBDs). Solid grey line: KPC-KP incidence density time series. Solid black lines: pre-intervention and intervention trends. ASP, Antimicrobial Stewardship Program; IPC, Infection Prevention and Control intervention

An evolutionary analysis of KPC-KP infection incidence across the pre-intervention, ASP, and COVID-19 IPC phases revealed significant differences between HOI and non-HOI infections only during the pre-intervention phase (Supplementary Table S12, p = 0.046). When comparing CA infections with the rest (HOI and HcAI), significant differences were observed throughout the entire study period (p < 0.001) and across all individual phases (Supplementary Table S12). These results suggest that HOI and HcAI infections exhibited more similar incidence patterns over time compared to CA infections.

Cross-correlation analysis revealed that HOl infections influenced non-HOI infections, with a 4-month lag (r = 0.35, 95% CI 0.17 to 0.53). Moreover, HOI infections and HcAI infections affected CA infections with respective lags of 2 months (r = 0.24, 95% CI 0.06 to 0.42) and 4 months (r = 0.28, 95% CI 0.1 to 0.46) (Supplementary Table S13).

Genomic epidemiology of KPC-KP HOI and non-HOI infections

Whole genome sequencing was performed on 316 KPC-KP isolates, representing 67.7% of all infection episodes (Fig. 1), including 212 isolates from hospital-onset infections (HOI, 67.1%) and 104 from non-hospital-onset infections (non-HOI, 32.9%). Among the latter, 81 (25.6%) were isolated from HCAI infections and 23 (7.3%) from CA infections. Nearly all sequenced isolates (315/316; 99.7%) belonged to sequence type ST512 and carried the blaKPC-3 gene, confirming the clonal dissemination of the high-risk lineage ST512/KPC-3. Only one isolate was identified as ST280.

To explore the population structure and evolutionary dynamics of this clonal group, we performed SNP-based phylogenetic analysis on the 315 ST512/KPC-3 isolates. The resulting tree (Fig. 4A) illustrated a tight phylogenetic relationship, consistent with clonal spread.

Fig. 4.

Temporal signal outweighs acquisition type in genomic clustering of ST512/KPC-3 isolates. A Phylogeny of 315 ST512/KPC-3 K. pneumoniae genomes obtained from community-acquired (CA), healthcare-associated (HcAI) and hospital-onset (HOI) infections. A maximum likelihood phylogenetic tree was generated based on SNPs analysis. Colors indicate the time period of isolation. Numbers refer to individual KPC-KP isolates. B Results of PERMANOVA evaluating the association between cophenetic distances and acquisition type (HOI, HcAI, CA) and year of isolation

To identify factors associated with genomic variability, we calculated cophenetic distances and assessed the marginal and joint effects of acquisition type (HOI, HcAI, CA) and year of isolation using PERMANOVA (Fig. 4B). Year of isolation was significantly associated with genetic divergence (R² = 0.17, p < 0.001), while acquisition type showed no independent effect when both variables were included. These results are detailed in Supplementary Figure S4.

This finding was visually supported by Principal Coordinates Analysis (PCoA), shown in Supplementary Figure S5: when isolates were coloured by acquisition type, no clear clustering patterns emerged. In contrast, when coloured by year (grouped in two-year intervals), isolates showed partial temporal grouping, with clusters forming among those collected within the same time period.

Together, these findings support the conclusion that ST512/KPC-3 KPC-KP isolates from both HOI and non-HOI infections belong to a single, evolving clonal population. The genomic epidemiology suggests that ST512/KPC-KP strains diversified primarily through temporal genetic drift, rather than being structured by the type of acquisition. This highlights the endemic persistence and gradual evolution of a dominant clone that has continuously circulated across hospital and community settings throughout the ten-year study period.

Discussion

In this study, we analyzed the ten-year incidence trends, clinical outcomes and genomic epidemiology of infections caused by KPC-producing Klebsiella pneumoniae (KPC-KP) in a high-endemicity tertiary care setting over a decade since the initial outbreak in 2012. Our findings highlight a critical epidemiological shift: from a hospital-confined outbreak to sustained transmission into healthcare-associated and community environments. Cross-correlation analysis revealed a 4-month lag between hospital-onset and non-hospital-onset infections, suggesting a temporal sequence in which hospital cases preceded those in the community. Genomic analysis confirmed the spread of a single clonal strain, ST512/KPC-3, across all settings, with time of isolation—not acquisition type—emerging as the strongest correlate of genomic clustering. This points to a clonal expansion across institutional and community interfaces, emphasizing the permeability of these boundaries in the context of endemic multidrug resistance.

Impact of ASP and COVID-19 IPC measures

Our results confirm that the implementation of a comprehensive ASP in 2014 had a sustained effect on reducing the incidence of KPC-KP infections, and this trend persisted throughout the COVID-19 pandemic. While we initially presented ASP and IPC efforts as temporally distinct, it is important to clarify that IPC measures—including contact precautions, hand hygiene protocols, environmental cleaning, and staff education—were already in place from the beginning of the study period (i.e., during the pre-ASP phase) and remained stable during the ASP period. These measures were not newly introduced in 2014 but formed part of the baseline IPC practices maintained throughout the early and middle phases of the study. It was only during the COVID-19 phase that IPC strategies were intensified and adapted in response to the specific challenges of the pandemic (Supplementary Tables S2–S4). Therefore, the observed impact of the 2014 intervention should be interpreted as the result of a coordinated ASP implemented on a foundation of ongoing, stable IPC practices.

During the pandemic, although global trends suggested increased antimicrobial use and reduced compliance with infection control measures targeting multidrug-resistant organisms (MDROs) [27–37], our center maintained low incidence rates. This"neutral impact"may be explained by the synergy between pre-existing ASP/IPC infrastructure, early implementation of COVID-19–specific IPC policies, and reduced hospital occupancy during lockdown periods. Comparable findings were reported in Taiwan [38] and Italy [39], where robust IPC and stewardship policies limited the impact of COVID-19 on MDR transmission.

The introduction of ceftazidime–avibactam during the ASP period likely contributed to improved outcomes and lower transmission by enabling targeted treatment of KPC-KP infections. However, its limited availability during the early years of implementation, and its restricted use by indication, prevented its independent effect from being statistically assessed in our cohort.

Clinical characteristics and outcomes

Although patients with non-HOI infections were older and had more comorbidities, clinical outcomes—including 30-day mortality and treatment response—were similar across groups. This reinforces the clinical significance of community-onset KPC-KP infections in endemic settings and highlights the need for early recognition and appropriate management.

However, the 90-day recurrence rate was significantly higher among non-HOI cases (16.8% vs. 6.4%, p < 0.001). This finding may reflect incomplete eradication, delays in follow-up, or challenges in outpatient antimicrobial stewardship—particularly among patients discharged to long-term care or community settings.

Clonal expansion of ST512/KPC-3 and genomic insights

Our genomic findings confirm the clonal nature of KPC-KP infections in our cohort, with nearly all isolates belonging to the high-risk ST512/KPC-3 lineage. This supports the scenario of a long-standing, endemic clone responsible for both hospital-onset (HOI) and non-hospital-onset (non-HOI) infections in our setting.

Importantly, although all isolates shared close genetic similarity, analysis of intra-clonal variability revealed that temporal factors—and not the epidemiological setting of acquisition—best explained genetic divergence. Specifically, while acquisition type (HOI, HcAI, CA) had minimal explanatory power, year of isolation consistently emerged as a significant driver of phylogenetic structure, as evidenced by both PERMANOVA and Principal Coordinates Analysis.

This suggests that evolution within the ST512/KPC-3 population is shaped by gradual, time-linked genomic drift, rather than compartmentalized outbreaks within specific care settings. The absence of clustering by acquisition type indicates ongoing transmission across healthcare boundaries, highlighting the lack of clear demarcation between hospital and community reservoirs in high-endemicity contexts.

These findings underscore the importance of longitudinal molecular surveillance, not only to track the persistence of dominant clones but also to detect subtle genetic changes that may impact resistance mechanisms or transmissibility. In stable clonal lineages such as ST512/KPC-3, temporal genomic shifts may still carry epidemiological and therapeutic relevance.

Implications for classification and prevention strategies

Our findings also contribute to the ongoing debate regarding how to classify healthcare-associated infections [16]. Following the original criteria proposed by Friedman et al. (2002) and subsequent adaptations such as the modified Duke-2002 definition [16], we categorized infections as community-acquired (CA), healthcare-associated (HcAI), or hospital-onset (HOI). While previous literature has grouped HcAI with nosocomial infections based on shared risk factors and outcomes [17, 18, 40], our results suggest a more nuanced scenario in the context of KPC-KP endemicity.

Clinically, HcAI cases in our cohort were more similar to CA than HOI in terms of age, comorbidity burden, and infection type. This justified grouping HcAI and CA as non-HOI for adjusted analyses—a strategy supported by prior work, including Salamanca-Rivera et al. [41].

From a genomic perspective, our findings suggest that intra-clonal variation within ST512/KPC-3 isolates is driven more by time of isolation than by epidemiological acquisition category. This reinforces the notion that KPC-KP transmission spans a continuum across hospital, long-term care, and community settings, rather than being confined to discrete silos.

These results support the need for integrated infection prevention and control (IPC) and antimicrobial stewardship (ASP) strategies that go beyond hospital walls. As the ST512/KPC-3 clone continues to evolve and disseminate across interconnected care networks, robust surveillance—including genomic tools—will be essential for early detection, containment, and targeted response.

Limitations

This study has several limitations. As a single-center analysis, the generalizability of our findings may be limited. However, given the continued detection of KPC-KP outbreaks in Spain [42, 43], our results likely reflect broader patterns relevant to other high-endemicity settings. We also did not assess asymptomatic colonization, which could help further characterize transmission dynamics. Lastly, although prior work at our center demonstrated reductions in carbapenem and quinolone consumption following ASP implementation [26], antimicrobial use during the COVID-19 period was not directly measured in this analysis.

Conclusion

Our study demonstrates that in a high-endemicity setting, sustained antimicrobial stewardship and infection control interventions are effective in reducing and maintaining low incidence rates of KPC-KP infections, not only within hospitals but across community interfaces. Genomic evidence suggests a clonal expansion of ST512/KPC-3, with acquisition timing—not acquisition source—best explaining isolate genetic relatedness. Non-hospital-onset infections are clinically significant and represent a key target for preventive strategies. Future policies must integrate genomic surveillance and long-term, multifaceted interventions that span both hospital and community healthcare environments to curb the spread of multidrug-resistant organisms.

Authors’ contributions

J.T.C., E.P.N., A.C., M.R.R., I.G.A. and L.M.M. conceived and designed the study. MRR., A.C., E.P.N., I.G.A., V.G.S., J.A.M.S, D.S., J.J.P.C., M.M., C.R., E.R.A., J.D.T.P., J.J.C., I.M., I.S. and all members of the KLEBMAN study group acquired the study data. J.T.C., E.P.N., A.C., M.R.R., T.L.V., B.G.G., I.G.A., M.M. and G.P. analysed and interpreted the data. E.P.N, V.G.S, J.A.M.S and M.M.S. performed and analysed the genomic studies. J.T.C., E.P.N., T.L.V., M.R.R., A.C., B.G.G., and L.M.M. drafted the manuscript. All main authors and members of the Study Group made critical revisions to the manuscript. J.T.C., E.P.N. and L.M.M. obtained the funds for the study. J.T.C., E.P.N., A.C., M.R.R., L.M.M. and I.G.A. verified all data. All authors had full access to all the data in the study and had final responsibility for the decision to submit for publication.

Funding

This work was supported by “Instituto de Salud Carlos III (ISCIII)” and co-funded by the European Union [grant number PI21/01199, KLEBMAN Project, to JTC; grant number PI23/00546, KLEBGEN Project, to E.P.N.]; “Center of Biomedical Investigation Network for Infectious Diseases (CIBERINFEC), ISCIII” [grant number CB21/13/00049]; and “Universidad de Cordoba, Consejería de Economía, Conocimiento, Empresas y Universidad de la Junta de Andalucía” [grant number 1381158-R, Programa Operativo FEDER Andalucía 2014–2020, to LMM and EPN].

Data availability

Sequence data that support the findings of this study have been deposited in the European Nucleotide Archive with the primary accession code PRJEB66458. The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

This study was approved by Reina Sofía University Hospital Institutional Review Board (code HURS-4785), which waived the requirement for written informed consent.

Consent for publication

Not applicable.

Competing interests

J.T.C. received honoraria for participating in an advisory board and educational activities for Pfizer, MSD and Shionogy, and research grants from Pfizer and MSD. AC received honoraria for participating in educational activities for Pfizer and Angelini. L.M.M. has been a consultant for MSD and Shionogi, has served as speaker for Merck, Astra-Zeneca, Astellas, Fastinov, Menarini and Shionogi and has received research support from Shionogi, Pfizer and MSD. All other authors declare no competing interest.