Infections, antibiotic treatment and mortality in patients admitted to ICUs in countries considered to have high levels of antibiotic resistance compared to those with low levels

Background Antimicrobial resistance is an increasing concern in ICUs worldwide. Infection with an antibiotic resistant (ABR) strain of an organism is associated with greater mortality than infection with the non-resistant strain, but there are few data assessing whether being admitted to an intensive care unit (ICU) with high levels of antimicrobial resistance is associated with a worse outcome than being admitted to an ICU with low rates of resistance. The aim of this study was, therefore, to compare the characteristics of infections and antibiotic treatments and patient outcomes in patients admitted to ICUs in countries considered as having high levels of antibiotic resistance and those admitted to ICUs in countries considered as having low levels of antibiotic resistance. Methods Data from the large, international EPIC II one-day point prevalence study on infections in patients hospitalized in ICUs were used. For the current study, we compared the data obtained from patients from two groups of countries: countries with reported MRSA rates of ≥ 25% (highABR: Greece, Israel, Italy, Malta, Portugal, Spain, and Turkey) and countries with MRSA rates of < 5% (lowABR: Denmark, Finland, Netherlands, Norway, and Sweden). Results On the study day, 1187/2204 (53.9%) patients in the HighABR ICUs were infected and 255/558 (45.7%) in the LowABR ICUs (P < 0.01). Patients in the HighABR ICUs were more severely ill than those in the LowABR ICUs, as reflected by a higher SAPS II score (35.6 vs 32.7, P < 0.05) and had longer median ICU (12 days vs 5 days) and hospital (24 days vs 16 days) lengths of stay. They also had higher crude ICU (20.0% vs 15.4%) and hospital (27.0% vs 21.5%) mortality rates (both P < 0.05). However, after multivariable adjustment and matched pair analysis there were no differences in ICU or hospital mortality rates between High or LowABR ICU patients overall or among those with infections. Conclusions Being hospitalized in an ICU in a region with high levels of antimicrobial resistance is not associated per se with a worse outcome. Electronic supplementary material The online version of this article (doi:10.1186/1471-2334-14-513) contains supplementary material, which is available to authorized users.


Background
The prevalence of infection is high among patients admitted to ICUs and is one of the main causes of ICU mortality [1,2]. In 1992, a one-day European point-prevalence study on infections in intensive care (EPIC), performed in 1417 ICUs in 17 Western European countries [3], showed that approximately 45% of all patients present on the ICU on the study day were infected. The mean prevalence of ICUacquired infections was 21%, but varied widely from 10% in Switzerland to 32% in Italy [3]. EPIC II was conducted 15 years later, on May 8, 2007 [2]. Data were collected from 13796 adult (>18 years) patients from 1265 ICUs in 75 countries and showed that 51% of patients had at least one infection.
Antimicrobial resistance is an increasing concern in ICUs worldwide. Rates of resistance vary considerably across different countries and regions; for example, ICUs in southern Europe generally have higher rates of resistance than countries in northern Europe and Scandinavia [4]. Several studies have reported that infection with an antibiotic resistant strain of an organism is associated with greater mortality and costs and longer ICU lengths of stay than infection with the non-resistant strain [5][6][7][8][9]. An important question is whether being admitted to an ICU with high levels of antimicrobial resistance is associated with a worse outcome than being admitted to an ICU with low rates of resistance, but there are few data available on this topic. In the present study, therefore, we used data from the large EPIC II study [2] to compare the characteristics of infections, antibiotic treatments and patient outcomes in patients admitted to ICUs in countries considered as having high levels of antibiotic resistance, as recorded in the Antimicrobial Resistance Surveillance System (EARSS) 2007 Annual Report, and those admitted to ICUs in countries considered as having low levels of antibiotic resistance. We hypothesized that there would be a difference in outcomes in patients admitted to ICUs in countries with high levels of antimicrobial resistance compared to those admitted to ICUs in countries with low levels of resistance.

Methods
The worldwide EPIC II 1-day point-prevalence study of infection and demographics of critically ill patients was performed on 8 May 2007 [2]. Data were collected for 13,796 adult patients in 1265 participating ICUs from 75 countries on the study day. Infection was defined according to the criteria of the International Sepsis Forum [10]. Microbiological analyses were performed locally. Participating ICUs were asked to provide patient followup until hospital discharge or for 60 days.
For the purposes of this analysis, we selected two groups of countries based on the country rate of methicillinresistant Staphylococcus aureus (MRSA) reported in the European Antimicrobial Resistance Surveillance System (EARSS) 2007 Annual Report [11]: countries with MRSA rates of ≥ 25% in the 2007 report (Greece, Israel, Italy, Malta, Portugal, Spain, and Turkey) and countries with MRSA rates of < 5% (Denmark, Finland, Netherlands, Norway, and Sweden) (see Appendix for list of participating centres in these countries). Severity scores, source of infection, pathogens, antibiotic resistance, antibiotic therapy (prophylactic and therapeutic, as defined by the attending physician at each centre), co-morbidities, lengths of stay and outcomes were compared in patients in the two groups of countries.
The EPIC II study was approved by the ethics committee of Erasme Hospital, Belgium, the coordinating center. Local ethical committee approval at each participating center (see Appendix for list of participating centers) was expedited or waived because of the observational nature of the study.

Statistical analyses
Statistical analyses were performed using IBM SPSS Statistics 20 for Windows (IBM, Armonk, NY, USA). The Kolmogorov-Smirnov test was used, and histograms and Q-Q plots were examined to verify if there were significant deviations from the normality assumption of continuous variables. Difference testing between groups was performed using Mann-Whitney test, χ 2 test, or Fisher's exact test as appropriate. Logistic regression analysis was used to determine the mortality risk associated with admission to an ICU in a country with high levels of antimicrobial resistance or one with low levels of antimicrobial resistance. To remove any bias of confounding variables for the association between the areas of ICU admission and mortality, a propensity score was estimated for each of the ICU areas, with the following variables considered as factors: type of admission, source of admission, comorbidities, age, sex, mechanical ventilation, haemofiltration or haemodialysis, infection, SAPS II score and type of microorganism. After checking that balance on all covariates that were used in the propensity model had been achieved, odds ratios with mortality as the dependent variable were estimated by logistic regression using two strategies: by matching and by introducing the propensity score in the model [12,13]. Reported odds ratios were adjusted for propensity score, hospital and organizational related factors, including type of ICU (closed vs. open, community vs. university, surgical vs. medical), number of ICU beds, number of nurses, number of physiotherapists, presence of 24-h ICU physician coverage, length of ICU stay prior to study day, percentage of gross domestic product spent on health care generated using the World Health Organization Statistical Information System (based on data from 2006 [14]) and geographical region. Data are presented as median [interquartile range (IQR)] or number (%), as appropriate. All tests were two-tailed, and a P < 0.05 was considered statistically significant.

Results
In the EPIC II database, there were 2204 patients included from ICUs in countries with high levels of antimicrobial resistance (HighABR) according to the EARSS 2007 Annual Report [11] and 558 from ICUs in countries with low levels of antimicrobial resistance (LowABR) Table 1). The median SAPS II score was significantly higher in HighABR than in  vs 31 [25-39] P < 0.001) and chronic renal failure and chronic obstructive pulmonary disease (COPD) were significantly more common (Table 1). Patients from HighABR ICUs were more often admitted from the emergency room and less often from the operating/recovery room than were patients from LowABR ICUs. Patients from HighABR ICUs had longer ICU and hospital lengths of stay than patients from LowABR ICUs; the number of days in the ICU prior to the study day was also longer in patients admitted to HighABR ICUs (Table 1).

Infections and Microbiology
More patients in the HighABR ICUs than in the Low-ABR ICUs were infected on the study day (53.9% vs 45.7%, P < 0.01). The respiratory system was the most common site of infection in all patients (Additional file 1). The sites of infection were similar in the two groups except for abdominal infections, which were more common in patients admitted to LowABR ICUs (24.3% vs 18.4%, P = 0.03), and bloodstream infections, which were more common in patients admitted to HighABR ICUs ( Table 2), whereas methicillin-sensitive S. aureus (MSSA), enterococci and anaerobes were more common in Low-ABR ICU patients.

Antibiotic use Treatment
Use of therapeutic antimicrobials was significantly more common overall in HighABR ICU patients than in  LowABR ICU patients (54.6% vs 44.4%, P < 0.001). When considering groups of antimicrobials, carbapenems, aminoglycosides, quinolones and glycopeptides were more commonly used in patients in HighABR ICUs than in LowABR ICUs, whereas cephalosporins were used more frequently as treatment in LowABR ICUs (Table 3). For individual antimicrobials, piperacillin-tazobactam, imipenem/meropenem, amikacin, oxazolidinone and tigecycline were used more frequently in patients in HighABR ICUs; cefazolin, cefuroxime, benzylpenicillin, oxa-/cloxa-/flucloxacillin, erythromycin and metronidazole were used more frequently in patients in LowABR ICUs (Table 3).

Prophylaxis
Use of prophylactic antimicrobials was significantly less common in HighABR ICU patients than in LowABR ICU patients (22.2% vs 32.8%, P < 0.001). Among the agents most commonly used were aminoglycosides (tobramycin), antifungals (amphotericin B) and cephalosporins (Additional file 2).

Mortality
Crude ICU (20.0% and 15.4%) and hospital (27.0% vs 21.5%) mortality rates were higher in patients admitted to HighABR than in those admitted to LowABR ICUs (both P < 0.05); however, these differences were not present after adjustment in multivariable analysis or the matched pair analysis (Table 4). For subgroups of infected or non-infected patients there were no significant differences in crude or adjusted mortality rates between HighABR and LowABR ICUs. Patients with infections due to Pseudomonas aeruginosa had crude ICU and hospital mortality rates of 32.4% and 42.7%, respectively. Corresponding percentages for patients with Acinetobacter infections were 31.7 and 37.3, respectively, and for MRSA 32.3 and 43.1, respectively. Because of the relatively small sample sizes, we were unable to compare outcomes for the different pathogens in the two groups.

Discussion
The main message of this study is that hospitalization in an ICU in a country with high levels of antimicrobial resistance is not an independent risk factor for worse outcome. We chose to select countries with likely high and low antimicrobial resistance using reported MRSA rates, although acknowledge that general resistance rates may have been different; nevertheless, most countries with high MRSA rates do seem to have high resistance levels among other organisms [11]. Several studies have shown increased mortality rates in patients with MRSA infections compared to those with MSSA infections [15,16].
In an earlier analysis of the EPIC II database, MRSA infection was associated with an increased risk of death compared to MSSA infection (OR 1.46 [95% CI 1.03-2.06], P = 0.03) [17]. Using data from the European Antimicrobial Resistance Surveillance System and the Burden of Resistance and Disease in European Nations project, De Kraker et al. [15] reported that 27,711 episodes of MRSA blood stream infections (BSIs) were associated with 5,503 excess deaths. The same authors compared two prospective cohorts of patients from 13 ICUs in 13 European countries, one with MRSA and one with MSSA BSIs, each matched with control patients without the respective BSI [16]. MRSA and MSSA patients had higher 30-day and hospital mortality than the control patients, and MRSA patients had greater 30-day mortality (OR 1.8, P = 0.04) than MSSA patients. Other studies have reported similar mortality rates in MRSA and MSSA patients, but longer lengths of stays and higher associated costs [6,7,18]. Infections due to difficult to treat pathogens, i.e., P. aeruginosa, Acinetobacter spp and MRSA, were in general associated with higher crude ICU and hospital mortality rates in our study than other infections and were significantly more prevalent in HighABR ICUs; however, because of small sample sizes, comparisons between groups were not possible. We know from other ICU studies that these pathogens are associated with increased attributable mortality if empirical treatment is not appropriate [19][20][21]. Unfortunately, we are unable to determine whether or not treatment was appropriate in our study, although we did show that broadspectrum antibiotics, such as carbapenems, piperacillin-tazobactam, amikacin, tigecycline, oxazolidinone and vancomycin, were more commonly used for treatment in the HighABR ICUs. Similarly, we did not record antibiograms for the different species and have no data on rates of resistance, except for S. aureus and ESBL-producing Gram-negative bacteria. Infections caused by enterococci and anaerobes were more common in LowABR ICU patients which may explain the more frequent abdominal infections and use of metronidazole in these ICUs.

Conclusions
Infections were more prevalent among patients admitted to ICUs in HighABR countries than in LowABR countries. Antibiotic therapy differed markedly, with broader spectrum antimicrobials being used more frequently in HighABR countries. ICU patients in HighABR countries were sicker and had longer ICU and hospital stays and higher crude ICU and hospital mortality rates, which could have a marked economical impact; however, after multivariate adjustment there were no differences in ICU or hospital mortality rates between the two groups of patients.
Appendix: List of participating centres for this EPIC II substudy by country, alphabetically