Study Design, Setting, and Patients

We conducted a multicenter cohort study during June 2016–April 2018, which included 15 hospitals in Lombardy (Figure 1). We asked each enrolled hospital to include data on all consecutively hospitalized adult patients who had >1 positive KPC-Kp isolate during their hospital stay. For patients hospitalized multiple times during the study period, we only considered the first hospitalization. For centers including patients during 2017, the year for which we had a full 12 months of data, we retrieved the administrative datasets of all admitted patients (Figure 1). We merged these data with those available in the KPC-Kp patient cohort database and used the combined dataset to describe KPC-Kp epidemiology in the hospitalized population.

The study protocol was first approved by the Research Ethics Committee of the coordinating center, Ospedale San Gerardo (Monza, Italy). Informed consent requirement was waived due to the study’s observational, noninterventional design. The study protocol was subsequently approved by the ethics committees of the 14 other participating centers. In accordance with local ethics committee requirements, 3 centers did not waive informed consent. Because this was an observational study, treatment for KPC-Kp infections was at the discretion of the attending physicians and no change to the center-specific surveillance protocol was required.

In all centers, intensive care unit (ICU) patients were tested for CRE at admission and weekly through rectal swab specimens or other surveillance cultures. The same protocol was applied heterogeneously in hospital wards in which patients are considered to be at higher risk of acquiring CRE, such as hematology, solid organ transplant, and geriatric units (Appendix Table 1). For the other wards, most centers performed surveillance rectal swab specimens at admission on the basis of major risk factors for CRE, such as previous CRE colonization, previous hospitalization during the 12 months before inclusion, or both. Of note, only 3 of the 15 participating centers, B, C, and I (Appendix Table 1), combined the 2 surveillance strategies described for specific wards and patients at higher risk of acquiring CRE.

Patient Classification

Patients were classified according to the most clinically relevant KPC-Kp isolate collected from them between hospital admission and discharge. Thus, for patients whose first isolate was attributable to colonization and a subsequent isolate was attributed to an infection, only the second isolate was considered. We used US Centers for Disease Control and Prevention criteria (13) to define diagnosed infection and diagnosis was confirmed by an infectious disease specialist. Infections were classified as KPC-Kp bacteremia when a blood culture was positive for a KPC-Kp strain with or without KPC-Kp–positive cultures from >1 other site and the patient had clinical signs of systemic inflammatory response syndrome requiring antimicrobial drug treatment. We defined nonbacteremic KPC-Kp infections by documented recovery of a KPC-Kp isolate from nonblood cultures, such as intra-abdominal wounds, urine, or bronchoalveolar lavage fluid; absence of KPC-Kp–positive blood culture during the index hospitalization; and clinical signs of infection.

In line with other studies (14), we classified KPC-Kp cases according to infection severity. We classified cases of KPC-Kp bloodstream or lower respiratory tract infections, and clinical presentation of septic shock, regardless of infection site, as severe infections. We classified infections from the urinary tract, surgical wounds, or other sites without septic shock as mild infections. We classified all cases identified through active surveillance as colonized when >1 culture sample grew KPC-Kp but the patient did not develop KPC-Kp infection during hospitalization.

Data Collection

For patients included in the KPC-Kp cohort, data were entered into the web-based case form after pseudonymization of personal data. Data were collected on demographic characteristics, medical history, underlying diseases, previous hospitalization, previous KPC-Kp infection, surgery <30 days before KPC-Kp isolation, invasive procedures <72 hours before KPC-Kp isolation, antimicrobial drug therapy <30 days before KPC-Kp isolation, dates of admission to hospital, and ward of isolation. Date of hospital discharge and patient status at discharge also were collected. The date and ward where the patient was hospitalized when KPC-Kp was isolated, the source of isolation, and resistance spectrum also were collected and entered into the web-based case record form. Antimicrobial treatment, including empirical treatment and post-antibiogram treatment regimen, were recorded. Empirical treatment was defined as adequate when it included >1 antimicrobial drug with in vitro activity against the KPC-Kp isolate. Data were collected in a web-based case report form.

For enrolled centers submitting patient data during 2017, we retrieved the clinical record datasets of all admitted patients after pseudonymization of personal information. To verify centers included all eligible patients, we retrieved the total number of patients with >1 KPC-Kp–positive isolate registered in the microbiology laboratory of each center and compared that with the total number of patients included in the cohort (Appendix).

Microbiology and Genomic Analysis

The clinical microbiology laboratory of each of the 15 participating centers performed isolate identification and routine antimicrobial susceptibility testing (Appendix). CRE was defined by using Clinical and Laboratory Standard Institute guidelines (15). All bacterial strains were sent to a central microbiological laboratory at Ospedale San Raffaele for whole-genome sequencing (Appendix).

Statistical Analysis

We estimated the prevalence of KPC-Kp in hospitalized patients in the region of Lombardy during 2017, the cumulative incidence of acquired KPC-Kp infections among hospitalized patients, and the cumulative incidence of acquired KPC-Kp infections occurring >48 hours after hospital admission among hospitalized patients in the same region. We calculated and reported crude estimates for all centers and estimates standardized by age and ward of isolation (Appendix).

To study the role of KPC-Kp infection severity on 15-day mortality rates, we considered a multivariable Cox proportional hazard model and the related hazard ratio (HR) estimates and adjusted by center for a random effect and number of days from hospitalization to KPC-Kp isolation. Colonized patients frequently have shorter hospital stays than infected patients. Because a shorter discharge time could affect our results, we performed a sensitivity analysis in which we excluded early-discharge patients. We performed a subgroup analysis to quantify excess mortality hazard due to septic shock among patients with bloodstream infections (Appendix).

We used multivariable mixed logistic regression models and accounted for clustering at the center level to evaluate the association between patient characteristics and delayed or inadequate empirical therapy, which we considered as outcome variables. We adjusted the models for age and type of KPC-Kp infection.

Center Characteristics

Among all centers, the median number of annual admissions was 27,600 (interquartile range [IQR] 18,287–40,000). Among 15 enrolled centers, 9 (60%) maintained enrollment over 12 consecutive months; centers had a mean enrollment duration of 13.8 months (Appendix Figure 1).

Patient Baseline Characteristics

Among 1,203 consecutive KPC-Kp–positive hospitalized patients found during study, 89.0% (1,071) were considered in the analyses and 11% (132) were excluded for various reasons (Figure 1). The median age among patients was 72 (IQR 61–80) years, 65% were male, and 35% were female; KPC-Kp was isolated from 275 (25.7%) ICU patients (Table 1). Among patients in the study cohort, >90% had >1 underlying condition, 40% of whom had congestive heart failure, peripheral vascular disease, or chronic renal failure. Severe infections were diagnosed in 221 (20%) patients and mild infections in 109 (10%) patients. Colonized patients (n = 741, 69.2%) had a median of 6 days between hospitalization and KPC-Kp isolation, which was much lower than for patients with severe (median 12 days) or mild (median 11 days) infections. Bloodstream infections accounted for 54% of all infections, and rectal swab samples accounted for 67% of all colonizations (Appendix Figure 2).

Distribution, Phylogeny, and Resistance Mechanisms of KPC-Kp Clones

Among the 1,071 patient strains isolated, 82 were from colonized patients included at the end of April 2018; these samples did not arrive at the central laboratory in time for genotyping. Of the 989 strains analyzed, 32 different sequence types (STs) were identified. The most numerous clones were ST512 in 45% (441), ST307 in 33% (326), ST258 in 7% (71), and ST101 in 6% (57) of isolates (Appendix Figure 3). We identified 2 KPC variants, KPC-2 and KPC-3, in 68% of isolates. KPC-2 was absent in ST512 but predominant in ST307 and ST258. Core-genome, single-nucleotide polymorphism (SNP) analysis revealed that ST512 was scattered across all centers, but ST307 was represented in smaller, more localized clusters (Figure 2; Appendix Table 3).

Epidemiology of KPC-Kp

During 2017, the estimated prevalence of KPC-Kp among hospitalized patients in the Lombardy region was 3.26 (95% CI 2.99–3.54) per 1,000 admissions. In the same region, the overall cumulative incidence of KPC-Kp infections was 1.00‰ (95% CI 0.86‰–1.16‰) and the incidence of acquired infections occurring >48 hours after hospital admission was 0.68‰ (95% CI 0.56‰–0.82‰). The proportion of patients infected at admission, considered imported infections, was ≈30% in most centers. We observed marked differences across centers even after standardization by age and ward of isolation, with values ranging from 1.62‰ (95% CI 1.07‰–2.18‰) in center A to 0.21‰ (95% CI 0.02‰–0.40‰) in center B (Appendix Figure 3).

Patient Outcomes

In-hospital death from all causes was 34% (95% CI 29.2%–39.6%) among KPC-Kp–infected patients and 21% (95% CI 17.7%–27.6%) among colonized patients. No differences emerged when we stratified for carbapenem-resistance mechanisms and the most prevalent clones (Appendix Table 4).

Mortality hazards (considering the first 15 days after KPC-Kp isolation) were much higher for patients with severe infection than for colonized patients, even after controlling for center, time between hospitalization and isolation, age, ward of isolation, and Charlson index (adjusted HR [aHR] = 1.93, 95% CI 1.40–2.66) (Table 2). In contrast, no excess mortality hazard was noted for patients with mild infections (aHR = 0.75, 95% CI 0.42–1.34) compared with colonized patients. When we analyzed the subgroup of patients with bloodstream infections, we found clinical manifestation of septic shock more than doubled the risk for death (HR = 2.71, 95% CI 1.46–5.02). We found comparable results when we excluded from the analysis 343 patients discharged alive before day 15 (data not shown).

Antimicrobial Drug Treatment

On the basis of susceptibility test results, we found that 54% (159/297) of patients infected with KPC-Kp received adequate empirical therapy (Appendix Table 5). Empirical treatment was most frequently adequate in patients with KPC-Kp colonization during the previous 12 months and in patients with severe infection (Appendix Table 5).

Fewer treatment delays (<4 days, which is considered the maximum acceptable waiting time to receive appropriate antimicrobial treatment) were reported for patients with severe KPC-Kp infection than patients with mild infections (Table 3). Patients reporting KPC-Kp colonization during the previous 12 months more frequently received prompt adequate therapy (p<0.001).

Among the 282 KPC-Kp–infected patients treated for their infections, 62 (22%) received an in vitro active drug plus carbapenem, but 29 (10%) patients received gentamicin, fosfomycin, or tigecycline monotherapy. The most common drug combination was colistin plus tigecycline plus carbapenem, which most frequently was administered to patients with severe infections. Ceftazidime/avibactam became available in Italy in February 2018, and 26/39 (66%) infected patients included after that date received it: 19/24 (79%) in the severe infection group and 7/15 (47%) in mild infection group (Appendix Table 6).