This retrospective case–control study was conducted in the medical intensive care unit (MICU) at Winthrop University Hospital, a 600-bed tertiary care university hospital, between January 1, 2010 and May 31, 2010. Only active and symptomatic infections, receiving antimicrobial treatment, were included. MDRO positive patients had organisms resistant to one or more classes of antimicrobial agents recommended as first line therapy and were identified from: blood, sputum, urine or tissue cultures, collected at the time of suspected infection. Included organisms were: Acinetobacter baumannii, Escherichia coli, Pseudomonas aeruginosa, Klebsiella pneumoniae, Proteus mirabilis, Morganella morganii and Serratia spp.

Other organisms identified, were vancomycin-resistant enterococcus spp. (VRE) and methicillin-resistant Staphylococcus aureus (MRSA) bacteria. Susceptibilities to all antimicrobial agents were determined and interpreted according to criteria of the Clinical and Laboratory Standards Institute (CLSI) by disk diffusion susceptibility.10 Intermediate susceptibility was considered as resistance, based on CLSI standards10 and because this approach has been used in other MDRO studies in the ICU.11–13 Only MDROs isolated more than 48h after MICU admission and within 15 days after MICU discharge were considered, and duplicates were excluded. All patients had cultures obtained at the time of suspected infection, but if cultures were negative, they were considered MDRO negative.

Controls were patients with infection requiring antimicrobial therapy, but with no MDROs in any type of cultures. We included culture negative patients in the control group because culture-negative sepsis is common in the United States, with 49% of hospitalized patients with sepsis being culture-negative.14 Patients could be included more than once, if there were different admissions for each infection. The study was approved by the Institutional Review Board of the Winthrop University Hospital. Patient consent was waived because data were collected and analyzed anonymously and retrospectively from patient records. All research data obtained were de-identified, handled, stored and shared with confidentiality.

Patient characteristics including, age, gender, race, body mass index, past medical and/or surgical history, comorbid illness and clinical course (total length of stay, and MICU length of stay) were recorded. Antimicrobial drug exposures were assessed during hospital admission and until discharge. Duration (days of use) of antimicrobial exposure, central venous catheter, and bladder catheter were also analyzed. Days of antibiotics, ventilation and hospitalization were further separated, into days prior and after the diagnosis of infection. The recorded outcomes were: mortality, discharge to home, or to a nursing home/long term facility.

The CDC definitions were used for infections at different anatomic sites.15 The definition of MDRO used was adopted from the CDC recommendations for the management of MDR organisms in healthcare settings that defines MDR organisms as bacteria that are resistant to one or more classes of antimicrobial agents recommended as first line therapy.1

Vancomycin-resistant Enterococcus spp. (VRE) was defined as the Enterococcus faecium/facaelis species that were resistant to vancomycin by standard susceptibility testing methods or by results from any FDA-approved test for VRE detection from specific specimen sources. Methicillin-resistant Staphylococcus aureus (MRSA) was defined as Staphylococcus aureus cultured from any specimen that tested oxacillin-resistant, cefoxitin-resistant, or methicillin-resistant by standard susceptibility testing methods.

Antimicrobial classes were: penicillins (oxacillin), third-generation cephalosporins (ceftazidime, cefotaxime, ceftriaxone), fourth-generation cephalosporins (cefepime), carbapenems (meropenem, etrapenem), anti-pseudomonal penicillins (piperacillin-tazobactam), aminoglycosides (amikacin, gentamicin), monobactams (aztreonam), glycopeptide (vancomycin) quinolones (levofloxacin, moxifloxacin), and glycylcyclines (tigecycline). Susceptibility to colistin was not considered.

The distinction between colonization and infection when the site of infection was the urinary tract or the lungs, was made by evaluating clinical criteria such as the presence of fever and the adequacy of organ perfusion. All included patients and controls met criteria for infection.

Immunosuppression was defined, as active solid or hematologic malignancy, leukopenia (absolute neutrophil count <1500cells/μl blood), chronic immunosuppressive treatment, prior radiation and use of corticosteroids at a dose of least 10mg/d for 15 days.

Prior hospitalization, prior antimicrobial use and prior surgical history were defined as a hospital stay, antimicrobial administration and undergoing major surgery respectively, within 3 months before the index hospitalization.

Liver dysfunction was considered present in any patient with of a bilirubin concentration over 2.0mg/dl or with liver cirrhosis, while renal insufficiency was defined in a patient with a creatinine level above 2.0mg/dl or requirement for dialysis. Cardiovascular failure was defined when inotropic drugs were required. Respiratory dysfunction was defined as inadequate gas exchange requiring endotracheal intubation and mechanical ventilation.

Previous history of hyperlipidemia was defined in a patient who was diagnosed with an abnormal level of lipids in the blood and was currently being treated. Previous history of hypothyrodism was defined in a patient who was diagnosed with hypothyroidism and was receiving therapy.

Initial empiric antibiotic treatment was defined as appropriate if the antibiotic prescribed within 24h of obtaining cultures matched the in vitro susceptibility of the presumed etiologic pathogen, in a patient with a positive culture at the time of infection. The doses used were in accordance with current standards. For example, in patients with normal renal function, we prescribed ertapenem 1g q24h, meropenem 1g q8h, ceftriaxone 2g q24h, cefotaxime 1g q8h, ceftazidime 1g q8h, cefepime 1g q6h, aztreonam 2g q8h, vancomycin loading dose 25–30mg/kg, followed by 15mg/kg q12h, linezolid 600mg q12h, levofloxacin 750mg q24h, piperacillin-tazobactam 4.5g q8h, gentamicin 7mg/kg, daptomycin 6–8mg/kg IV q24h, tigecycline 100mg IV loading dose, followed by 50mg q12h. By 2010, the CLSI approved new breakpoints for meropenem, imipenem, and doripenem that define susceptibility as a MIC of ≤1μg/ml and resistance as ≥4μg/ml.16

Continuous variables were tested for normality using histograms and Kolmogorov–Smirnov test. If the data were markedly non-normally distributed, they are presented as medians (interquartile range) and categorical data are presented as percentages. Fisher's exact test was used to compare MDRO (positive vs. negative) for binary variables and Wilcoxon rank-sum test for continuous variables. We used all variables that were significant in univariate logistic regression at p<0.05 in the multiple logistic regression model, with a stepwise selection method to find independent predictors of MDRO. Risk was assessed using odds ratios. All calculations were performed utilizing SAS 9.2 (SAS Institute, Cary, NC) for Windows and results were considered statistically significant when p<0.05.

Three hundred and thirteen (313) patients were included in this study. Demographic and clinical characteristics comparing cases and controls are shown in Table 1. Although Coagulase-negative Staphylococci (CoNS) can cause true bacteraemia, repeated blood cultures (from a total of 9 subjects) did not isolate CoNS with either the same antibiogram or isolated another pathogen. Therefore these CoNs isolates were considered as a contaminant and not a pathogen. MDRO were found in 127 patients (41.7%), while 177 (58.2%) did not have MDRO (control population) present in cultures. Of the 177 control patients, a non-MDRO pathogen was present in 65 and cultures were negative in 112. Among the MDRO positive and negative patients, infection may have been incubating at the time of hospital admission in 21 (16%) and in 9 patients (12%) respectively, although all infections were diagnosed after 48h in the ICU, and thus met the criteria for nosocomial infection. Interestingly all these patients had healthcare exposure risk factors [history of prior hospitalization and one or two comorbidities (COPD, diabetes)] and thus were similar to HAIs in terms of various comorbid conditions, source of infection, pathogens and mortality rate and therefore we considered them as HAIs.

The MDRO patients had a median age of 75 years vs. 69 years in controls (p=0.06). There were no statistically significant differences between the two groups with respect to gender and body mass index. The comorbidities were similar, except that more MDRO positive patients had immunosuppression (60.0% vs 41.2%, p=0.001), prior antimicrobial use (12.6% vs 6.2%, p=0.06) and a history of prior hospitalization (33.1% vs 19.2%, p=0.007).

Among the 127 patients with MDRO infection, 38 (29.9%) had septic shock,. Septic shock was present in 35 (19.7%) of the 177 patients with non-MDRO infection. All other patients had active infection with sepsis. Of the 127 MDRO positive patients, previous colonization with the same MDRO was not identified, although routine surveillance cultures were not collected.

In those with MDRO infection, the site of infection was: the urinary tract in 55%, the lung in 35% and other sites (abdomen, skin and soft tissues) in 10%. For those with non-MDRO infection, the site of infection was: the urinary tract in 32%, the lung in 38%, and other sites in 30%.

The identity and frequency of the pathogens are shown in Table 1, with methicillin-resistant Staphylococcus aureus, Enterococcus faecium, Acinetobacter baumannii, Klebsiella pneumoniae, Escherichia coli and Pseudomonas aeruginosa being the most common. After univariate analysis (Table 2), the clinical factors associated with MDRO were: infection and cardiac disease as a cause of admission, total days in hospital, days in MICU, days on the ventilator, days with a central venous catheter, tracheostomy, prior hospitalization, prior antibiotic use (before hospitalization), immunosuppression, past history of hyperlipidemia, use of vasopressors, past history of surgery and white blood cell count >10,000/mm3 on admission.

The multivariate analysis included all variables with a univariate p-value of <0.05 (Table 2), and the only clinical features significantly associated with MDRO were: infection as a cause of admission (p<0.0001, OR=3.3), total days in hospital (p=0.005, OR=1.04) total days of ventilation (p=0.013, OR=1.07), immunosuppression (p=0.01, OR=2.04), past history of hyperlipidemia (p=0.008, OR=2.2), past surgical history (p=0.033, OR=1.82), age (p=0.032, OR=1.02) and white race (p=0.025, OR=0.44). Days of ventilation and days in hospital were further separated, into days before and after the onset of infection.

The median time for acquiring an MDRO infection while in the hospital was 8 days (range of 1–48 days). Each additional hospital day increased the risk of having an MDRO by about 5%.

Days of hospitalization (the day of hospital admission was calendar day 1) were separated, into days prior and days after the diagnosis of infection. Both groups spent a similar number of days in the hospital prior to infection, but the MDRO positive patients spent a greater median number of days after infection than the MDRO negative patients. The median number of days in hospital after the diagnosis of infection was 13 and 10 days, respectively (p=0.006) for the MDRO positive and negative patients (Table 3). Culture negative control patients were not included in this analysis.

In the analysis of days on ventilation prior and after the diagnosis of infection we observed that more days on ventilation (prior to infection) were present for MDRO+ versus MDRO− patients (p=0.009). Additionally MDRO patients spent more median days on ventilation after infection than MDRO negative patients (p=0.054). Culture negative control patients were not included in this analysis.

MDRO positive patients had a greater median number of antibiotic days before infection and after infection, compared to MDRO negative patients. Among MDRO positive and negative patients the median number of days of antibiotics prescribed prior to the infection was 2 and 0 days respectively (p=0.004) (Table 3). After the diagnosis of the infection among the MDRO positive and negative patients the median number of days of antibiotics prescribed was 8 and 6 days respectively (p=0.009) (Table 3).

Multivariable analysis did not identify any specific antibiotic as independently associated with MDROs.

Appropriate empiric antibiotic therapy (matching in vitro susceptibility to the isolated pathogen) was administrated in 105 out of 127 (82.6%) of MDRO positive cases. 72/105 (68.5%) patients received appropriate therapy within 24h of the onset of infection. Among MDRO negative patients with an identified pathogen, 60/65 received antibiotic therapy, which was appropriate in 54/60. The use of appropriate therapy was not associated with reduced mortality in either the MDRO positive or MDRO negative patients.

No statistically significant difference in 28-day ICU mortality rate was observed between MDRO positive and MDRO negative patients (20% vs 12.4%, p=0.07). In addition, there was no difference in the percent of patients discharged to home, nursing home, hospice or other long term care facilities (Table 1).