Background
There is an alarming increase in antibiotic-resistant Gram-negative infections [
1‐
3]. Among Gram-negative infections,
Pseudomonas aeruginosais is one of the most common gram-negative bacteria causing nosocomial and healthcare-associated infections (HAIs) in hospitalized patients [
4]. The World Health Organization places carbapenem-resistant
P. aeruginosa as a critical priority pathogen that desperately requires new treatment options [
5]. Increasing rates of multidrug-resistant (MDR)
P. aeruginosa in HAIs and among hospitalized patients is a major public health problem [
6]. MDR
P. aeruginosa infections in the hospital setting are associated with poor outcomes including increased resource utilization and costs, morbidity, and mortality [
7].
MDR
P. aeruginosa account for 13–19% of HAIs each year in the US. The increasing level of resistance in MDR
P. aeruginosa is often attributed to patient-to-patient transmission of resistant strains as well as newly acquired resistance owing to previous antibiotic exposure. As per standardized international terminology, MDR is defined as non-susceptibility to at least one agent in three or more antimicrobial categories and XDR is defined as non-susceptibility to at least one agent in all but two or fewer antimicrobial categories (i.e. bacterial isolates remain susceptible to only one or two categories) [
8]. In severe systemic infections, consideration of MDR/XDR
P. aeruginosa when selecting treatment is warranted to ensure timely and appropriate initial therapy. Lack of effective antibacterial therapies against
P. aeruginosa infections severely limit effective therapeutic options and consequently, lead to inappropriate initial therapy that adversely impacts health outcomes [
9]. Typically it takes 48 h to definitively identify MDR
P. aeruginosa, which can be a detrimental delay for these patients.
There are considerable gaps and inconsistencies in knowledge regarding risk factors associated with the occurrence of MDR
P. aeruginosa in nosocomial infections and HAIs. Identifying risk factors predicting acquisition of MDR
P. aeruginosa or identifying subgroups of patients who are at an increased risk for acquisition of MDR
P. aeruginosa in the hospital setting will assist in providing timely and appropriate treatment. While a recent review examined the risk factors independently associated with extensively drug-resistant (XDR)
P. aeruginosa, there has not been a comprehensive analysis of contemporary literature reporting all levels of resistance (MDR or XDR or resistant) of
P. aeruginosa infections [
6]. This systematic review of recent published literature evaluates risk factors that are independently associated with acquisition of microbiologically identified MDR or XDR, or resistant (single drug class)
P. aeruginosa from various sites among inpatients, hospitalized in any of the following setting including wards, intensive care units, and other types of inpatient settings.
Methods
We performed a systematic search in the MEDLINE, Cochrane Library, and EMBASE databases for citations indexed from January 01, 2000 through December 31, 2016. The initial search strategy (Additional file
1: Table S1) includes terms related to the pathogen (
P. aeruginosa), mode of infection (nosocomial, hospital-acquired, healthcare-acquired, hospital-associated, healthcare-associated), and risk factor assessment (risk factors, predict, risk score, risk assessment, and multivariate analysis). Additionally, search terms for Gram-negative infections, beta-lactamases, and metallo-beta-lactamases attributed to carbapenem resistance were included to increase the yield of studies with resistant (single drug class)
P. aeruginosa. These three databases were searched because they index most of the published citations. To supplement this search, we reviewed reference lists of eligible studies using an iterative process to maximize inclusion of relevant data. We did not perform grey literature searches or systematically search for unpublished data. We reviewed abstracts from conference proceedings if they were indexed in the three aforementioned databases.
Study definitions
Our primary objective was to examine studies evaluating risk factors associated with acquisition of MDR or XDR, or resistant P. aeruginosa among inpatient adult patients evaluated in the hospital setting. In this review, the hospital setting comprises all types of units, including intensive care units, emergency room/casualty, or other wards where patients were located at the time of collection of the resistant P. aeruginosa microbiological specimen and underwent further treatment in the hospital setting. The acquisition of resistant P. aeruginosa could have been either community- or hospital-acquired, but all patients must have been evaluated and treated in the hospital setting. We accepted MDR and XDR P. aeruginosa as defined by individual studies, although we noted considerable heterogeneity in the definitions. Non-MDR P. aeruginosa was also defined by the individual studies and varied in definition to include P. aeruginosa infections other than MDR such as susceptible P. aeruginosa (susceptible to all antipseudomonal agents) or resistant to any one class of drug. Non- P. aeruginosa was also defined by the individual studies and varied in definition to include pathogens other than P. aeruginosa.
Eligibility criteria
Studies evaluating patients hospitalized in a ward, intensive or critical care unit, or any other inpatient healthcare setting (including chronic care facilities), were eligible for inclusion. Studies examining patients with either hospital-acquired or community-acquired were included, provided that they were evaluated in a hospital /inpatient setting. Risk factors that predict acquisition of MDR or XDR, or resistant P. aeruginosa were included. Risk factors were categorized into patient characteristics, hospital characteristics, and treatment characteristics.
The exposure and comparators of interest included: MDR or XDR P. aeruginosa versus resistant P. aeruginosa (to any one class of antibiotic); MDR or XDR P. aeruginosa versus susceptible P. aeruginosa; MDR or XDR P. aeruginosa versus control (any other Gram-negative pathogen or any infection with unspecified pathogen); resistant P. aeruginosa versus susceptible P. aeruginosa. The outcome of interest was microbiologically confirmed acquisition of MDR or XDR, or resistant P. aeruginosa. For studies that reported both the outcome of acquisition of MDR or XDR P. aeruginosa and single-drug resistant P. aeruginosa, we included the study only once (for the outcome of acquisition of MDR or XDR P. aeruginosa) in the analysis. We accepted all study designs except non-comparative studies, case reports, and case series. Studies that reported multivariable results adjusting for any potential confounders were eligible. No minimum study duration or follow-up time was required for inclusion.
We excluded studies of mixed infections that had more than 20% Gram-positive infections (to ensure most (80%) of the evaluated infections in eligible studies were Gram-negative infections) and studies solely examining P. aeruginosa resistant due to metallo-beta-lactamases. Studies published in languages other than English were excluded.
All citations identified by literature searches were independently screened by two researchers. Upon the start of citation screening, we implemented a training session where all researchers screened the same articles and conflicts were discussed. We iteratively continued training until we reached consensus regarding the nuances of the eligibility criteria for screening. During double-screening, we resolved conflicts through group discussions.
We assessed the methodological quality of each study based on predefined criteria using the Agency for Healthcare Research and Quality (AHRQ) and the New-Castle Ottawa risk of bias tool, which probes risk of selection, performance, detection, attrition, reporting, and other potential biases. Each study was extracted by one experienced methodologist (GR, EA, and JC). The extraction was reviewed and confirmed by at least one other methodologist. Any disagreements were resolved by discussion amongst the study team. Data were extracted into customized Microsoft Excel™ forms. We tested the data extraction forms on several studies and revised as necessary before full data extraction. Extracted data included variables addressing population characteristics, including severity of illness, description of exposure and comparator groups, sites of infection, outcome definitions, enrolled and analyzed sample sizes, study design features, and multivariate results. Any missing or unavailable data were deemed as not reported information. We performed random effects model meta-analyses of eligible studies if at least three or more studies were sufficiently similar in population, exposure, and outcomes [
10]. If appropriate, we also conducted meta-regression analyses to evaluate study features explaining potential heterogeneity. However, due to the presence of substantial clinical heterogeneity, we also qualitatively compared results across studies (for example, source of infection).
All analyses were performed in Stata version 14 (StataCorp, College Station, Texas) with the metan, metareg, and metabias functions. We tested between study heterogeneity with the Q statistic (significant when
p < 0.10) and quantified the extent of heterogeneity with the I
2 statistic. It is common practice to interpret I
2 > 50% as representing substantial inconsistency or significant statistical heterogeneity [
11].
When applicable, results were also presented in table format. All meta-analyses of adjusted or multivariate results were presented in forest plots. The summary results examining the relationship between risk factors and acquisition of MDR or XDR, or resistant P. aeruginosa were presented as adjusted odds ratio or relative risks with 95% confidence intervals.
Discussion
This systematic review of the literature summarizes risk factors that predict acquisition of MDR P. aeruginosa, XDR P. aeruginosa, and other resistant P. aeruginosa. Our meta-analysis identified that the risk factors for acquisition of MDR or XDR P. aeruginosa varied with regard to the type of control used for comparison. ICU stays (either as current or prior to current episode of infection) and use of quinolones was significantly associated with acquisition of MDR or XDR P. aeruginosa, compared with resistant or susceptible P. aeruginosa. Use of prior antibiotics including, cephalosporins, quinolones, or carbapenems, and prior hospital admissions predicted an increased risk of MDR or XDR P. aeruginosa versus non-P. aeruginosa.
The increasing rates of MDR or XDR, or resistant
P. aeruginosa are a worldwide public health problem. Resistant strains of
P. aeruginosa are associated with high mortality and increased resource utilization [
66]. The emergence and spread of MDR
P. aeruginosa may be associated with misuse or overuse of antimicrobials. Our meta-analysis found a consistent association between the use of quinolones and acquisition of MDR or XDR
P. aeruginosa P. aeruginosa. However, association between carbapenem use and acquisition of MDR or XDR
P. aeruginosa was not consistent across comparisons. While some primary studies did report a relationship between MDR or XDR
P. aeruginosa and prior use of antibiotics [
14,
30], our meta-analysis identified statistically significant associations between carbapenem use and acquisition of MDR or XDR
P. aeruginosa and the significance of their association can vary by the type of comparator examined. Nonetheless, our results confirm that prior use of specific antimicrobials is an important risk factor for acquisition of MDR or XDR
P. aeruginosa. Antimicrobial resistance occurs over time, often mediated by gene mutations and misuse or overuse of antimicrobials may accelerate this process. Understanding the epidemiology of MDR or XDR
P. aeruginosa will be necessary to overcome infections with these resistant pathogens.
This review finding can be supported in the context of a recent review on MDR or XDR
P. aeruginosa, which examined multivariate risk factors reported in eight included articles [
67]. The authors of this review did not perform a meta-analysis but identified prior antimicrobial therapy, medical devices, patient-related characteristics, and environmental sources as risk factors for MDR or XDR
P. aeruginosa. Our review, in which a meta-analysis was performed, demonstrated that admission location, prior admission, or use of quinolones was significantly associated with acquisition of MDR or XDR
P. aeruginosa. In addition to reviewing MDR or XDR
P. aeruginosa, we reviewed single drug-resistant
P. aeruginosa. In meta-analysis, prior use of piperacillin-tazobactam, vancomycin, or carbapenems were significantly associated with acquisition of carbapenem-resistant versus susceptible
P. aeruginosa. In meta-analysis, prior fluoroquinolone use was a statistically significant predictor of subsequent quinolone-resistant versus susceptible
P. aeruginosa.
There are limitations associated with this systematic literature review. To begin with, the study inclusion criteria were limited to published studies indexed in English. Secondly, studies eligible for inclusion were heterogeneous with respect to the definition of exposure, site of infection with P. aeruginosa, and risk factors included in the study. As a result, a combined meta-analysis may misrepresent the true picture. We mitigated these issues by using a random effects model in our analyses and by limiting analyses to similar comparators. However, in many instances there were risk factors that were statistically significant in two studies, but unfortunately a third study was not available to permit the exploration of this factor through meta-analysis. Additionally, our study results may not be generalizable to all regions as studies from certain regions (for example, Western European nations for MDR or XDR P. aeruginosa and Brazil for carbapenem-resistant P. aeruginosa) were overrepresented in our sample published literature.
Confounding in observational studies reporting unadjusted data is a well-known source of bias. To mitigate this bias, we included risk factors that were analyzed in multivariate analyses. The choice of the control group (e.g. patients with infection with susceptible strains or patients without any infection at all) may also have influenced the results. We attempted to address this issue using similar comparison groups in the meta-analyses. As with any evidence synthesis, the limitations of the data available in primary studies will transfer into limitations of the systematic review. For example, the control group was not well defined in studies, specifically in studies that reported non-
P. aeruginosa as controls. The description of these controls was unclear if the control group did or did not include MDR infection of a non-Pseudomonas bacteria. The results section describes variability across studies with regard to study characteristics, outcome assessment, and the relationship between risk factors and the outcome of acquisition of resistant
P. aeruginosa. We could not conduct additional subgroup analyses by site of infection, or explain differences across studies using stratified analyses due to the small number of available studies for any particular comparison. An insufficient number of studies (< 10 studies) precluded evaluatingthe potential for publication bias with funnel plots and Egger’s tests for small study effects [
68].