Discussion

Our study describes a rapidly changing CRE epidemiology from 2011 to 2019 at a major academic medical center in the United States, and is one of the first to report both short- and long-term outcomes following the first year after isolation of CRE among hospitalized patients. ^{Reference Babiker, Clarke and Saul12} We found that survival improved over time, but the total burden of disease, as measured by recurrent infections and hospital re-admissions, remained significant. In fact, only a minority of patients who survived hospital discharge were discharged home, underscoring the added burden placed on long-term care facilities and nursing homes. ^{Reference Guh, Bulens and Mu23} We also identified a number of factors associated with improved clinical outcomes, including the year of treatment, pathogen type, and receipt of SOT prior to CRE infection. Although some of these factors may be expected, others were not, and attest to the need for continued clinical and molecular surveillance of CRE infections over time. Such data can be used to strategize treatment approaches in the current era of widely available β-lactam/β-lactamase inhibitor (BL/BLI) agents with in vitro activity against KPC and use of rapid molecular diagnostic tests.

Not surprisingly during this period, Most cases were due to ST258 KPC-producing K. pneumoniae; ^{Reference van Duin, Arias and Komarow2,Reference Sabour, Huang and Bhatnagar7,Reference Guh, Bulens and Mu23} however, the increasing frequency of non-CP CRE is particularly notable because these isolates are genetically diverse. Data collected across 49 hospitals showed that clinical outcomes of patients infected with non-CP CRE are similar to those infected with KPC-producing CRE. ^{Reference van Duin, Arias and Komarow2} These data, including those in the current report, contrast a prior single-center report of cases from 2013–2016 that identified a 4-fold increase in 14-day mortality rates among patients infected with CP- versus non-CP CRE. ^{Reference Tamma, Goodman and Harris24} Another key findings is a high proportion of CRE cases (28.6%; 300/1,049) that were unlikely to be reliably carbapenem-resistant based on the reported phenotype, corroborating a prior multicenter report. ^{Reference van Duin, Arias and Komarow2} Our approach in adjudicating CRE index cases was done through manual inspection of each phenotypic profile wherein standardized criteria were applied. Cases were excluded if carbapenem resistance was identified, but no other resistance to β-lactams, or if ampicillin and cefazolin were reported as susceptible. Although a more judicious approach would include repeat susceptibility testing on these isolates, this could not be done retrospectively. Notably, in a study of 1,040 CRE cases, 22% (228/1,040) of isolates were categorized as unconfirmed CRE after repeat testing in a central laboratory failed to identify carbapenem resistance. ^{Reference van Duin, Arias and Komarow2} Taken together, these data indicate that routine testing may misclassify a significant percentage of cases as CRE, particularly among non-CP CRE. Further studies are needed to confirm these findings prospectively with repeat testing in real time, exclusion of mixed cultures, and molecular investigation of the underlying mechanisms of resistance.

The increasing proportion of non-CP CRE at our center coincided with a decline in the number of KPC-producing CRE cases over time. This trend is consistent with a large epidemiological study across the Veterans Affairs healthcare system during the same period. ^{Reference Wilson, El Chakhtoura and Patel25} Interestingly, multiple distinct ST258 sublineages were identified and associated with device- and ward-level outbreaks, ^{Reference Marsh, Mustapha and Griffith26} but a discernable shift occurred between 2014 and 2015 from clade 1, KPC-2-producing K. pneumoniae to clade 2, KPC-3-producing K. pneumoniae. Although this is aligned with the availability of ceftazidime-avibactam in the United States in 2015, the clade 2, KPC-3 producing sublineage emerged prior to clinical use at our center. ^{Reference Marsh, Mustapha and Griffith26} It is more likely that adaptive mutations, including those in pfeA and rcsA, arose overtime allowing clade II to persist at our center. This persistence, however, was short-lived as rates of KPC-producing K. pneumoniae declined in 2017, a trend that is mirrored across the United States and globally. ^{Reference Shi, Ru and Wang27,Reference Duffy, Li and Czaja28} The reduction in ST258 KPC-producing K. pneumoniae was in fact a driving factor associated with a significantly lower rate CRE index cases by the end of the study period. To date, it is unclear why rates of KPC-producing CRE declined so precipitously across the United States, but notably similar trends have been identified among ST11 in China and other STs globally. ^{Reference Shi, Ru and Wang27} It is possible that interventions such as aggressive infection control, increased laboratory recognition of KPC, improved stewardship practices, and introduction of several novel BL/BLI agents with in vitro activity against KPC collectively affected these rates. ²⁹ Interestingly, rates of CP CRE have continued to decline in surveillance studies conducted since 2019, but as the proportion of KPC has declined, the rates of OXA-48 and NDM have slowly risen. ^{Reference Sader, Mendes, Carvalhaes, Kimbrough and Castanheira9} These data attest to the continued need for active surveillance studies to understand and adapt to continuing shifts in the CRE epidemiology, including the expanded use of whole-genome sequencing to identify clonal relatedness and expansion of new clones over time.

Next, our data confirm that overall outcomes for patients with CRE infections have improved over time. ^{Reference Babiker, Clarke and Saul12} Such changes can be clearly ascribed to improved treatment with BL/BLI agents rather than older agents like aminoglycosides and colistin. ^{Reference van Duin, Lok and Earley30,Reference Shields, Nguyen and Chen31} Molecular diagnostics have also improved detection of KPC, prompting earlier active therapy. ^{Reference Satlin, Chen and Gomez-Simmonds32} Although current guidelines recommend use of meropenem-vaborbactam, ceftazidime-avibactam, or imipenem-relebactam for the treatment of KPC-producing CRE, there is limited evidence to prioritize one agent over another. Comparative-effectiveness data for meropenem-vaborbactam and ceftazidime-avibactam is limited to a single retrospective cohort study of 131 patients with CRE infections where no significant mortality differences were identified. ^{Reference Ackley, Roshdy and Meredith33} It may be postulated that meropenem-vaborbactam should be preferred over ceftazidime-avibactam for cases of pneumonia considering pneumonia has been shown to be an independent predictor of failure for patients receiving ceftazidime-avibactam for CRE infections and the bronchial epithelial lining fluid exposure of ceftazidime-avibactam is only ∼30% of serum concentrations. ^{Reference Shields, Nguyen, Chen, Press, Kreiswirth and Clancy34,Reference Nicolau, Siew and Armstrong35} Nevertheless, future studies are needed, including those that evaluate robust long-term outcomes using the data described in this study as a key benchmark.

Mortality rates were lower among SOT patients with CRE in our study compared to non-SOT patients (Figure 2 ). These data extend our prior investigations and are consistent with case-control studies. ^{Reference Babiker, Clarke and Saul12,Reference Kalil, Syed and Rupp36} In a propensity score-adjusted analysis of patients with bacteremia, 28-day and 90-day mortality rates were significantly lower for SOT recipients compared with non-SOT patients. ^{Reference Kalil, Syed and Rupp36} In a separate analysis of patients with CRE infection specifically, SOT were shown to have a 7% lower risk 30-day mortality after inverse probability weighted adjustment. ^{Reference Boutzoukas, Dai and Cober37} Collectively the data suggest that immunosuppression, including use of corticosteroids, mitigates an inflammatory response in the setting of acute bacterial infections. It is also possible that SOT may have enhanced provider vigilance and improved access to health care. Furthermore, it is likely that clinicians exhibit lower thresholds to treat SOT patients when there are not clear signs of infection, which may portend improved longer-term outcomes.

When outcomes were stratified by CRE organism, we found the highest mortality rates with K. aerogenes compared to other species (Figure 2). Perhaps more importantly, our overall findings show that infection types, severity of illness, and clinical outcomes vary by CRE species, indicating that further investigations are needed to assess pathogen virulence, propensity for recurrent infections, and ultimately evolution of resistance to frontline therapies. To this end, it is clear that CRE encompass a wide range of pathogens, and further discrimination of the appropriate terminology is needed in medical education and treatment pathways. ^{Reference Livermore, Nicolau, Hopkins and Meunier38}

While the strengths of this study include its large sample size across a decade, standardized definitions of carbapenem resistance, and our ability to track patients within the same healthcare system over time, several limitations should be recognized. First, this study was performed retrospectively, and thus, the availability of clinical isolates was based on specimens that had been previously archived, and our determination of infection types was subject to documentation in the electronic health record. We were also unable to evaluate improved infection prevention practices in real time; however, outbreak detection and associated surveillance changes at our center have been previously reported. ^{Reference Marsh, Mustapha and Griffith26} Secondly, this study is an epidemiology study and specific treatment regimens were not assessed, but is a focus on future investigations. Third, our study period ended prior to the COVID-19 pandemic; however, contemporary data suggests that trends for lower rates of KPC isolation continued postpandemic. ^{Reference Sader, Mendes, Carvalhaes, Kimbrough and Castanheira9,Reference Shi, Ru and Wang27} Finally, this study was single center within the United States, and therefore external validity may be limited when extrapolating these results outside of the United States where epidemiologic and treatment patterns vary. Nevertheless, our data have unveiled several unique features of CRE infections, including a rapidly changing epidemiology and hypothesis-generating findings for patients with varying underlying diseases where long-term outcomes and healthcare utilization can be optimized.