In this study, we report on a molecular and infection surveillance of carbapenem-susceptible A. baumannii on a surgical intensive care unit in a tertiary care centre. During the one-year study period, an initial polyclonal increase of A. baumannii was observed, followed by the emergence of two dominant clones.
A pathogen-based surveillance is part of the vertical approach in infection control ; it is mostly conducted for MDR bacteria, e.g. carbapenem-resistant A. baumannii, in German hospitals . Hence, there is no detailed surveillance data about the occurrence and epidemiology of carbapenem-susceptible A. baumannii. Furthermore, we found only a few descriptions of outbreaks of carbapenem-susceptible A. baumannii in the literature, e.g. . A recent meta-analysis reported that three quarter of Acinetobacter outbreaks in the literature were attributed to MDR bacteria. However, the definition of MDR is used inconsistently, and is not always in line with international recommendations . The underreporting of outbreaks of non-MDR A. baumannii might be due to a surveillance bias, as IPC programmes put the main focus on MDR bacteria. Another reason might be a publication bias, as MDR infections are more difficult to treat and as MDR A. baumannii has high transmission rates [1, 7]. Moreover, MDR A. baumannii has become endemic in a lot of countries. The carbapenem-resistance rate exceeded 75% in invasive isolates in some south-eastern European countries in 2016 . On German intensive care units, 43% of clinical isolates were resistant to imipenem in 2015 . This study was conducted in a low endemic setting of carbapenem-resistant A. baumannii (only sporadic appearance and almost exclusively related to a hospital stay abroad).
The epidemiology of A. baumannii can be quite complex. Simultaneously occurring endemic and epidemic MDR A. baumannii clones are described in the literature, making detection and control difficult [1, 28]. The worldwide emergence of carbapenem-resistant A. baumannii is caused by a few successful clonal lineages . The molecular epidemiology of non-MDR A. baumannii is less clear, but the population of non-MDR isolates is described as less homogenous . Based on our results, the dynamics of transmission cannot simply be explained by an “endemic setting” of A. baumannii. Noteworthy, nearly all acquisitions were hospital-acquired. Moreover, the general reservoir of A. baumannii is unknown, as it is found almost exclusively in the hospital environment [1, 5]. Appearance and transmission of A. baumannii is quite common on intensive care units [7, 28] and occurs via contaminated hands after contact with the inanimate environment or colonized/infected patients. A recent review showed a positive association between infection/colonization with A. baumannii and exposure to rooms previously occupied by patients with A. baumannii . In our study, we found direct and indirect evidence for both modes of transmission (patient-to-patient or environment-to-patient). The variety of environmental A. baumannii isolates, clonally and non-clonally related to clinical isolates, found on frequently touched surfaces, demonstrates the widespread environmental contamination and supports the importance of hand hygiene. We were not able to find definitive epidemiological links for all patients. Of course, colonized health-care workers or colonized (undetected) patients cannot be completely ruled out. It can be quite cumbersome to establish spatiotemporal links between patients, who are often transferred to several wards and units during their hospital stay. Interestingly, we also found evidence for transmissions from the SICU to and within other wards. Some electronic surveillance systems can help to detect patient contacts, e.g. the Hybase software used in this study. However, they fail if several patients from different wards are affected or in cases of indirect transmissions (room contact). Easy to use software solutions are lacking to analyse and visualize complex patient transfers.
The transmissions of the clonal strains were stopped with a combination of standard hygiene measures, intensified environmental cleaning and disinfection and rectal screening for carbapenem-susceptible ACB-complex. Contact precautions were only applied to a minority of patients as stated above. The expanded screening was introduced in the middle of the study period and resulted in a significantly earlier detection of ACB-complex. Before the introduction, community-acquired isolates might have been misclassified as hospital-acquired due to the late detection. Neither the sensitivity of patient screening for (MDR) A. baumannii colonization nor the ideal screening loci are well known. Recent studies showed a low sensitivity of detection in colonized patients  or demonstrated that perirectal screening might be more appropriate than rectal screening for A. baumannii . Nevertheless, in our opinion the combination of three screening loci (rectal and nose/throat) and subsequent inoculation on a cefpodoxime-containing media, as performed in our study, reach a sufficient sensitivity.
We also applied UV light to disinfect complex surfaces  in combination with two times standard cleaning and disinfection over a period of 2 months. As rooms were blocked for hours during this procedure, these additional measures were stopped. Therefore, the impact of additional environmental cleaning and disinfection remains unclear in our study. However, intensified room cleaning and disinfection was described as an effective IPC measure during outbreak periods . We also observed a relevant environmental contamination, especially on difficult-to-clean regulators of the endotracheal suction system. As nearly all of them were contaminated with skin flora and other pathogens (e.g. Klebsiella pneumoniae, Enterococcus faecium), they might have been a reservoir of transmission. As there is no dedicated cleaning staff for terminal cleaning and disinfection of medical products in our hospital during the late and night shifts, this task is assigned to the nurse at times of patient-to-nurse ratios of sometimes 3:1 or worse. This ratio is not unusual on German intensive care units .
The A. baumannii infection rate of approx. 50% was described during outbreaks, with no difference between MDR- and non-MDR infected patients . In our cohort, we observed a hospital-acquired A. baumannii infection rate of 33.3% (12 out of 36 patients). The majority of our patients showed the relevant risk factors for colonization/infection with A. baumannii, already well described for MDR Acinetobacter, such as prolonged hospital stays at an ICU, mechanical ventilation or prior antimicrobial therapy [7, 28]. Selective pressure of broad-spectrum penicillins or third-generation cephalosporins, both favoured in our hospital, might have played a decisive role.
We encountered some pitfalls of such surveillance. A proper and reliable (molecular) identification of the isolates to the species level is needed, as the different ACB-complex species differ in their hospital epidemiology. Especially A. baumannii and to a certain extent A. nosocomialis are known to cause outbreaks in hospital settings [1, 5]. In our opinion, it is important to consider microbiological results after the patients’ transfer to other wards. We would have missed seven patients if we had only included specimens collected on the SICU. Hence, a hospital-wide surveillance appears more appropriate to identify possible transmission pathways, but requires extended IPC resources. Furthermore, a typing method with high discriminatory power is crucial. In our opinion, RAPD and PFGE are suitable methods for typing of A. baumannii. However, in times of next generation sequencing (NGS), a “random access” database may be more appropriate. For example, a gene-by-gene approach with an accepted allele scheme (cgMLST) can be used over longer periods . The application of a cgMLST scheme was recently demonstrated by Willems et al. in an outbreak of A. baumannii .
There are a few limitations to this study. First, we only analysed one isolate per resistance pattern per patient. Patients colonized/infected with more than one strain might have remained undetected. Secondly, a formal case–control study to determine risk factors for carriage of carbapenem-susceptible A. baumannii was not performed. However, our aim was to describe the surveillance results and the molecular epidemiology. Thirdly, we only applied NGS for isolates that were highly related using PFGE and RAPD for economic reasons. Related isolates showing different PFGE/RAPD patterns, though unlikely, might have gone unnoticed.