The incidence of 3GCREB infections among patients admitted to German university hospitals in 2014/15 was < 1%. Additional fluoroquinolone resistance was frequent in particular in CAI, while additional carbapenem resistance was rare, both in CAI and HAI (0.006 and 0.008 per 100 patients; HAI incidence density 0.014 per 1000 patient days). An interesting finding was that among CA-LRTI caused by 3GCR Klebsiella spp., the percentage of additional carbapenem resistance (28%) was substantial and significantly higher than in 3GCR Klebsiella spp. causing HA-LRTI (6%, p = 0.015).
In an admission prevalence study performed in parallel, we screened a minimum of 500 patients per hospital for rectal 3GCREB carriage on admission (day 1–3, admission day = day 1). This study showed a 3GCREB colonization prevalence of 9.5% . Therefore, we conclude that the colonization rate of patients in university hospitals is about 16 times higher (=9.5%/0.58% patients with 3GCREB infection) than suggested when only patients with 3GCREB infections are considered. The reason for this high colonization prevalence and for a CAI incidence comparable to HAI is most likely the case mix in tertiary care university hospitals due to their position at the end of the treatment chain. On admission, many patients most likely have received previous antimicrobial treatment and have higher co-morbidity scores than patients in other hospital types.
The incidence of CAI causing 3GCREB that were also resistant to fluoroquinolones exceeded that of HAI (Table 2). One reason for this may be the high antibiotic use reported by patients admitted to the participating hospitals (34% of the 3GCREB-negative and 53% of the 3GCREB-positive patients) . Another reason may be an enhanced use of fluoroquinolones in outpatient care . This excess of fluoroquinolone use might be caused (among other reasons) by over prescription and non-adherence to antibiotic prescription guidelines in outpatient care [11,12,13]. A reduction of fluoroquinolone prescriptions would be desirable.
In 2013/14, the EU mean of carbapenemase-producing (CP) E. coli and K. pneumoniae was found to be 0.025 per 1000 patient days and 0.006 per 1000 patient days were reported for Germany (EuSCAPE study) . We observed a CRE HAI incidence density for Klebsiella spp. of 0.006 (95% CI 0.000–0.008) and for E. coli of 0.003 per 1000 patient days (95% CI 0.001–0.005). The EuSCAPE study showed that among CRE, 70% of the K. pneumoniae and 30% of the E. coli produced carbapenemases . Combining these percentages of CPE among CRE with our CRE data yields an estimated CPE incidence density for Klebsiella spp. and E. coli of 0.005 per 1000 patient days (Klebsiella spp. 0.004 and E. coli 0.001 per 1000 patient days). Thus, our data is comparable to the EuSCAPE data for Germany .
In contrast to other Enterobacteriaceae species, 3GCR Klebsiella spp. showed a high percentage of carbapenem resistance among CA infections, especially in LRTI and BSI. In a UK study, the prevalence of carbapenem resistance in clinically relevant K. pneumoniae specimens was also due primarily to community-acquired isolates (70%) . The EU mean of carbapenem resistance in invasive K. pneumoniae isolates increased from 2012 to 2015 to 8%. In two European countries, carbapenem resistance was observed in over 25% of K. pneumoniae (Italy 34% and Greece 62%). In isolates from Germany, carbapenem resistance was rare (0.1%) . Surveillance in German ICUs showed that the carbapenem resistance rate of K. pneumoniae in clinically relevant specimens increased from 2001 to 2015 to 1.5% . A large admission prevalence study found a low CRE admission prevalence in Germany (comparable to UK, both 0.1%) [6, 16]. In light of this data, we conclude that even with increasing trends Germany currently still is a low CRE prevalence region.
The ATHOS project was a prospective observational study. One major limitation is the lack of patient-based information on previous healthcare contacts. Therefore, a classification into the important “healthcare-associated” category was not possible. Instead, the somewhat arbitrary day 3 limit for classification into CA and HA was applied, as it is commonly used for surveillance of HAI . Nonetheless, we have some insight into the healthcare contacts of our patient mix from our patients in the admission prevalence study sample, 9% of whom had stayed in a rehabilitation centre, 5% in a long-term care facility and 26% in another hospital in the 6 months prior to admission . Some CAI may, in fact, be healthcare-associated and thus the incidence of CAI may be overestimated in our analysis. This CA/HA classificiation was also used for SSI. We cannot exclude that some CA-SSI cases may be readmitted cases. An advantage of the study is the inclusion of general wards. We can describe the incidence of 3GCREB infections in German university hospitals comprehensively and are not limited to ICU data. However, due to the hospital-wide surveillance, we lack ward-specific denominator data and are not able to calculate department-specific incidences.
One strength of the study is the inclusion of most Enterobacteriaceae species, since other studies are often restricted to E. coli or K. pneumoniae. Such studies are likely to underestimate the real incidence of CRE. We found a major part of CRE-HAI caused by Enterobacter spp. (0.006 per 1000 patient days, 95% CI 0.000–0.008). Imipenem resistance of E. cloacae complex has increased over the last years in German ICUs . In addition, long-term surveillance by the US Veterans Health Administration of CRE in patients also showed a steady increase of carbapenem resistance in E. cloacae . However, we have noticed that Vitek®2 frequently overreported imipenem resistance. For future CRE studies, we suggest including Enterobacter spp. and using additional, more reliable diagnostic methods (e.g. disk diffusion test or agar gradient diffusion) to determine if the high frequency of Vitek®2 imipenem resistance represents the true epidemiology. Another drawback of using routine diagnostic methods is that we were unable to compare sequence types of 3GCREB causing CAI and HAI.
As a result of the 3GCREB admission prevalence study performed in parallel  and the surveillance data for infections, we can for the first time estimate the 3GCREB prevalence to be about 16-times higher than indicated by the 3GCREB infection incidence (0.58%). This will enable other university hospitals with a similar patient mix to estimate roughly the dimension of colonization prevalence present in their patients. Furthermore, we conclude that very few 3GCREB carriers are identified using clinically indicated diagnostic procedures. Therefore, we believe the majority of hospitals underestimate the extent of 3GCREB prevalence. This study gives a comprehensive description of the incidence of CRE found in German university hospitals.