This study demonstrated that BSI caused by different GNB had significantly different clinical characteristics by comparing the risk factors such as causes of ICU admission, underlying diseases, sources of infection, treatment and invasive procedures.
MDR-K. pneumoniae BSIs
A meta study [16] found that the main risk factors for carbapenem-resistant K. pneumoniae (CR K. pneumoniae) infection included: long hospital stay or ICU stay, previous use of carbapenems and steroids, CVC implantation, mechanical ventilation and tracheostomy, with the first three as the major risk factors. Our study found that all the above risk factors included a higher proportion of K. pneumoniae infections (mainly CR K. pneumoniae), but a less proportion of A. baumannii infection, compared with P. aeruginosa and E. coli infection. Other studies [8, 17, 18] have confirmed some of the above risk factors, and found that CR K. pneumoniae colonization is a strong risk factor for later infection. In our study, K. pneumoniae was found in 64.6% of patient specimens before the infection onset, and most of them were colonized bacteria. K. pneumoniae can asymptomatically colonize the skin, mouth, respiratory tract and intestines, and intestinal colonization is closely related to subsequent infections [19-21]. Thus, we suspected that K. pneumoniae in this study might mainly come from the direct or indirect gut microbiota. One study found that the intestinal decontamination can decrease the incidence rate of ICU-acquired GNB BSI by 45%, while the selective oropharyngeal decontamination only decreased by 33% [22]. However, in a randomized trial of 8665 patients [23], oropharyngeal decontamination or SDD did not reduce mortality in ICU-acquired MDR-GNB BSI compared with the standard care. The all-cause mortality in our study was 63.3%, thus higher than 44.5% [8] (mortality due to CR K. pneumoniae BSI in ICU patients). Another study [24] found that mortality rates varied with severity, since in high-risk patients the mortality rate of CR K. pneumoniae BSI is 77%, while it is 45% in middle-risk patients. Therefore, we hypothesized that the mortality was closely related to the general condition and treatment strategy.
MDR-E. coli BSIs
Although E. coli is present in the normal intestinal flora, it can enter and survive in a sterile parenteral environment, which often leads to urinary tract infections and BSI [25]. Indeed, E. coli is the most important pathogenic bacteria in urinary and abdominal infections [26] and BSI. One study reported that 76% of E. coli BSIs come from home, and the most common source of home-acquired E. coli BSIs was the urinary tract, while the gastrointestinal tract was the common source for nosocomial E. coli BSIs [27]. Our observations were similar with the above observations. Patients in the E. coli group were admitted to the ICU mainly for septic shock or renal failure secondary to urinary or abdominal infection, and BSI often occurred within 48 hours after the transfer from the general ward or community to the ICU, thus, the previous duration of hospital stay and ICU stay were short. In this study, most MDR-E. coli were ESBLs-producing strains and Ajao et al. [28] found that colonization and exposure to multiple antibiotics are independent risk factors for infection with E. coli-producing ESBLs. However, these factors were not significantly present in our study due to short hospital stay and unclear out-of-hospital medication for some patients. Overall, the E. coli group was significantly different from the other groups due to its special source of infection, and the mortality rate was low thanks to the ease of keeping its infections under control and easy access to sensitive antibiotics.
MDR-P. aeruginosa BSIs
Unlike other bacteria, P. aeruginosa not only has high virulence, but also has inherent resistance to some antibiotics and is prone to drug resistance. Our study found that P. aeruginosa BSIs were common in immunocompromised patients with a previous chemoradiotherapy and neutropenia, Mccarthy et al. found similar findings [29]. In our study, patients after surgery often developed P. aeruginosa BSI, which might be related to the previous intestinal colonization [30, 31]. Indeed, P. aeruginosa often causes pneumonia and can escape from the lungs by destroying the pulmonary vascular barrier to cause BSI [32]. This might explain the pulmonary origin of P. aeruginosa BSI in our study. Similar to other studies [3, 33], but different from the other groups, in the present study P. aeruginosa often caused burn wound infection, which subsequently led to BSI. The 30-day mortality rate was 41.5% in our study, similar to the 37% reported by other studies [30]. Thaden et al. found that P. aeruginosa BSI is related to increased mortality compared with other GNB BSIs, and this effect persists after adjustment for comorbidities, resistance, and treatment factors [34], thus, P. aeruginosa virulence might play an important role in lethality. However, by vertical comparison of the risk factors leading to death on day 2, 7, and 28 after P. aeruginosa BSI, Mccarthy [29] et al. found that from day-7 onwards patient comorbidities become increasingly important, suggesting that the 30-day mortality depends on the underlying disease to some extent.
MDR-A. baumannii BSIs
It is well known that ICU is a very common place to contract an infection, especially from A. baumannii, which mainly includes the following two reasons. As regard patient's factors: long hospital stay [5], experience of invasive procedures (CVC [35], surgery [1, 36], mechanical ventilation [37]), exposure to antibiotics [1, 36] such as carbapenems, weak immunity (previous hormone therapy [36], tumor patients with neutropenia [38]). As regard the contamination of wards [39] A. baumannii is a bacterium commonly present in ICU that is difficult to eliminate and it often spreads among the ward crowd, resulting in a high risk of A. baumannii colonization for patients [40, 41]. Some of these aspects were also found in our study. Compared with other groups, A. baumannii group had the longest hospitalization or ICU stay, the highest total proportion of invasive operations, and exposure to more than 3 classes of antibiotics (carbapenem and piperacillin tazobactam were the main antibiotics used). Our hypothesis is that it is difficult to reduce the occurrence of A. baumannii infection in ICU, except for the decontamination and improvement of the awareness of the cleaning operations of the medical workers.
Drug resistance
Carbapenems were often selected for empirical treatment in our study, which is an important factor causing the onset of carbapenem-resistant A. baumannii, K. pneumoniae and P. aeruginosa. Carbapenems and tigacycline were rarely used in E. coli group when admitted to hospital, thus, the resistance rate was significantly lower in this group. However, with drug-resistant genes resulting in drug-resistant bacteria, they can widely spread in hospitals and communities, consequently resulting in an increased incidence of carbapenem-resistant E. coli and CR K. pneumoniae infections in China [42]. In China, ST11 has been considered as the dominant CR K. pneumoniae strain, but new ST11 CR A. baumannii is emerging [43], which is hypervirulent, multidrug resistant, and transmissible, potentially resulting a real superbacteria that could pose a serious threat to public health. The polycolisin-resistant E. coli has been found in China [42], which undoubtedly adds to the current sad situation.
Prognostic analysis of MDR-A. baumannii BSIs
Overall, previous studies found that the prognosis of A. baumannii BSI in ICU patients is different depending on the underlying disease, illness severity, drug resistance, therapies used, post-infection appearance and infection source. The following characteristics often lead to poor prognosis: improper initial antibiotic treatment [9, 38], comorbidities (tumours [38, 44, 45], liver cirrhosis [46], chronic obstructive pulmonary disease and chronic renal failure [37]), high drug resistance, neutropenia [38], previous high-dose hormone therapy [47], previous surgery[9], infection originated from the respiratory tract [35, 46], high SAPS score [8], Pitt score[35], APACHE II score [37, 38, 46]. However, the development of septic shock [47] and the severity of the disease are the factors most closely related to prognosis, and our study also found these two key points. A study [9] focused on ICU patients found that patients developing septic shock after MDR-A. baumannii BSI have a 23.8% higher mortality rate than patients without septic shock (82.5% vs 58.7%). An early study [10] grouped by APACHE II score found that the mortality of 15 < APACHE II scores ≤ 25 group was 14.3%, and the mortality of 25 < APACHE II scores ≤ 35 group was 3 times of that of 15 < APACHE II scores ≤ 25 group. In addition, Yang et al. found that in case of A. baumannii BSI, the death rate of the carbapenem MICs ≥ 8 mg/l group was twice that of the MICs ≤ 48 mg/l group [48]. In our study, hypoalbuminemia and a previous hormone therapy were independently associated with death, suggesting that nutritional and immune statuses are very important. Because the access to colistin was limited in our hospital, common drug treatments mainly included tigecycline in combination with cefoperazone/sulbactam, and monotherapy with amikacin or sulfamethoxazole trimethoprim, but none of the monotherapies has been reported as associated with survival. In addition, in vitro treatments with sensitive antibiotics that improved the prognosis were not found.
Limitations and strengths
Some limitations are present in this study. Firstly, this was an observational research with its inherent defects. Secondly, it is a single-centre study with relatively few cases included, thus, differences between groups could not reflect a common situation worldwide. Thirdly, in the analysis of the prognosis, the adjustment for the severity of the disease and antibiotic regimen was not performed. However, this study was based on ICU patients, reducing the influence of ward difference and convenient to compare among patients of similar illness severity. The clinical characteristics of four common MDR-GNB BSIs in ICU were compared to better identify the characteristics of different BSIs. In addition, the response of vital organs or systems after infection was evaluated in order to explore the pathogenicity of MDR-A. baumannii BSI.