Characteristics and Clinical Implications of Carbapenemase-Producing Klebsiella pneumoniae Colonization and Infection, Italy

We found 15-day mortality rates were higher for patients with severe infections than for those with mild infections or colonization.


Patient Infection Classification
When a single patient had multiple KPC-Kp infections, the most clinically relevant infection was considered in the analysis. For example, if a patient experienced a urinary tract infection followed by a bloodstream infection, only the bloodstream infection and related isolate were considered.

Personal Data Pseudonymization Process
The patient's name and date of birth were pseudonymized automatically with the generation of a patient identification number saved in the database. Only center staff could visualize personal data saved in a separate file.

Definition of Acquired Infections Adopted in All Centers
Patients with acquired infections were those with an infection arising >48 hours of admission (1). For infectious events, clinical presentation of septic shock, defined as sepsis with organ dysfunction and persistent hypotension despite volume replacement (2); chronic renal failure; antimicrobial treatment including empirical treatment, which was considered adequate when it included >1 drug with in vitro activity against the KPC-Kp isolate; and post-antibiogram treatment regimen were recorded.

Microbiology and Genomic Analysis of Strains
During the study period, the laboratory of each of the 15 participating centers collected consecutive, nonreplicate, clinical isolates of KPC-Kp from any site of infection or colonization of patients enrolled in the study.
Isolates were identified with the Vitek 2 system (bioMérieux, https://www.biomerieux.com) or matrix-assisted laser desorption/ionization-time-of-flight mass spectrometry by MALDI Biotyper (Bruker Daltonics GmbH, https://www.bruker.com) or Vitek-MS (bioMérieux). Each hospital conducted antimicrobial susceptibility testing via automated systems and according to standard protocols. Eleven centers used the Vitek   (Pinf = 1000*(Nadm+Nstay)/Ntot), and the cumulative incidence of acquired infections occurring >48 hours of hospital admission among hospitalized patients in the region of Lombardy in 2017 (CI = 1,000* Nstay/Ntot). Prevalence/incidence were reported as crude rates and standardized by age (lower or higher than 66 years, the median of the 170,699 patients admitted) and ward of isolation (ICU, infectious diseases, surgery, oncology/hematology, and other medical wards), according to a direct method (stdP). The standard population was the overall adult population of patients admitted in 2017 in these 8 centers, excluding day admissions and pediatric admissions. We calculated 95% CI by using Poisson distribution (Appendix Figure 3).
To study the role of KPC-Kp infection severity on 15-day mortality rates in KPC-Kp patients, we considered the exposure variable of KPC-Kp infections as severe, mild, or colonized. We conducted a survival analysis in which the time of KPC-Kp isolation was taken as time of origin (i.e., t = 0), and the event was hospital death occurring <15 days of KPC-Kp isolation, thus we censored hospital stays at 15 days. Colonized patients were selected as the reference category since they represented the best available control group because they were hospitalized during the same time and at the same locations in which the KPC-Kp-infected cases arose, and had comparable clinical characteristics (Appendix Table 2). Multivariable Cox proportional hazard frailty models on 15-day hospital mortality rates were used to estimate both crude and adjusted hazard ratios (HRs). To select covariates in the Cox proportional hazard mixed models, we constructed several Cox proportional hazard mixed models (i.e., center was entered as random effect) to identify factors independently associated with 15-day hospital mortality rates. In particular, we included the following: previous colonization (yes/no), previous hospitalization (yes/no); isolation ward ICU (yes/no); antimicrobial therapy in the 30 days before hospitalization (yes/no); major surgery in the 30 days before isolation (yes/no); Charlson Index; underlying conditions, including congestive heart failure (yes/no), peripheral vascular disease (yes/no), cerebrovascular disease (yes/no), chronic lung disease (yes/no), chronic renal failure (yes/no), cancer (yes/no), or diabetes (yes/no); central venous catheter at isolation, urinary catheter at isolation, immunosuppressive therapy (yes/no); carbapenem-resistance mechanism (KPC3 vs KPC2); and major clones ST512 (yes/no) or ST307 (yes/no). Only the number of days between admission and KPC-Kp isolation were considered for KPC-Kp patients as a covariate in the first models. A center-specific random intercept was also included, to adjust for potential When the main targeted therapeutic regimens were compared between the severe and mild infection groups, we used χ 2 and Fisher exact tests to compare categorical variables and ttest or Mann-Whitney U test to compare continuous variables. We performed all analyses by using SAS 9.4 software (SAS Institute, Inc., https://www.sas.com) and considered p≤0.05 statistically significant.