INTRODUCTION
The burden of healthcare-associated infections (HAIs) is especially high in developing countries, most of which are in the tropics1. Although work processes are a main cause of HAIs2, ecological determinants are increasingly recognized to be relevant epidemiological factors. As an example, recent studies report the seasonality of infections caused by gram-negative bacilli in healthcare settings3,4.
Acinetobacter baumannii infections are an interesting example of HAIs, with a high incidence in tropical climates5, seasonality6 and an increasing incidence of multidrug resistance7. Acinetobacter baumannii is one of the most common agents found in HAIs in Brazilian hospitals and poses a particular threat for patients in high-risk units8,9.
Several risk factors for nosocomial acquisition of A. baumannii (overall and multi-drug-resistant isolates) have been reported. These factors include the severity of illness and the use of invasive devices and antimicrobials10–12. However, many of these risk factors are similar to those reported for other pathogens13. Therefore, competition among microorganisms affecting closed populations subjected to the same ecological pressures may be a possibility in this context.
Our study was designed to assess the effects of such competition. We analyzed, from an ecological perspective, the influence of several pathogens on the monthly incidence rates of healthcare-associated bloodstream infections (HA-BSIs) caused by A. baumannii.
METHODS
Study setting and period
The study was conducted in a teaching hospital in an interior part of the State of São Paulo. The hospital has 450 active beds and is a tertiary referral hospital for an area with 500,000 inhabitants. Intensive care units comprise approximately 10% of hospital beds. The hospital has a microbiology laboratory and an infection control committee. Our study analyzed data over a time span of 72 months, from 2005 through 2010.
Study design
The study had an ecological design. We examined the monthly incidence rate data for HA-BSIs (overall or caused by specific pathogens or groups of pathogens). The total number number of months for which data were analyzed was 72. The outcome of interest was the incidence rate of HA-BSIs caused by A. baumannii.
Data and definitions
A search was performed using the hospital microbiology laboratory database for the period 2005 to 2010. To generate monthly data, we counted the number of patients with positive blood cultures. Patients with positive cultures (from any site) in the month previous to admission or within the first three days after admission were excluded, in accordance with the Society for Healthcare Epidemiology of America (SHEA) definitions for healthcare-associated pathogens14. We also excluded duplications, defined as the isolation of the same agent from different blood cultures obtained from the same patient within a 30-day period.
To generate a database of monthly incidence rates, we defined HA-BSIs on the basis of a single blood culture that tested positive for one or more typical pathogens. For atypical agents (e.g., coagulase-negative staphylococci [CoNS]), two positive blood cultures collected within 24h were required15. For the purpose of our study, we did not differentiate primary (or catheter-related) bloodstream infections from those secondary to other primary sites.
Data on the durations of hospital stays (patient-days) were recovered from hospital administrative files. Monthly incidence rates were expressed as cases per 10,000 patient-days.
Statistical analysis
Analysis was performed using the Statistical Package for the Social Sciences (SPSS) version 19.0 (IBM, Armonk, NY, USA). Briefly, we applied univariate and multivariate models of Poisson regression, selecting the monthly incidence of HA-BSIs caused by A. baumannii as the outcome. Infections caused by other agents or groups were included as predictive factors. A value of p < 0.05 was considered statistically significant. To present the data graphically, we repeated the analysis using linear regression, with similar results.
RESULTS
In the study period, the overall incidence rate of HA-BSI was 15.8 per 10,000 patient-days. Acinetobacter baumannii ranked third in incidence among the etiologic agents, with an incidence of 2.5 per 10,000 patient-days. In the multivariate analysis (Table 1), the incidence rate of A. baumannii was negatively associated with the incidence rates of Staphylococcus aureus (rate ratio [RR]=0.88; 95% confidence interval [CI]=0.80-0.97), Enterobacter spp. (RR=0.84; 95%CI=0.74-0.94), and a pool of gram-negative pathogens that individually were of low incidence (e.g., Serratia spp., Citrobacter spp., and several others). We also found a positive association between the incidence of CoNS and A. baumannii.
TABLE 1 – Results of Poisson regression models for the associations between the incidences of several pathogens and the incidence of Acinetobacter baumannii.
| Pathogens | Univariate analysis | Multivariate analysis | ||||
|---|---|---|---|---|---|---|
| Crude RR | 95%CI | P | adjusted RR | 95%CI | P | |
| Fungi | ||||||
| Candida spp. | 0.90 | 0.81-0.99 | 0.04 | 0.94 | 0.84-1.04 | 0.23 |
| Gram-negative bacilli | ||||||
| Pseudomonas aeruginosa | 1.06 | 0.97-1.17 | 0.21 | 1.05 | 0.94-1.15 | 0.38 |
| Enterobacter spp. | 0.84 | 0.74-0.94 | 0.003 | 0.84 | 0.74-0.94 | 0.004 |
| Klebsiella spp. | 1.07 | 0.99-1.16 | 0.08 | 1.09 | 0.99-1.18 | 0.053 |
| Escherichia coli | 1.09 | 0.99-1.19 | 0.06 | 1.10 | 0.99-1.22 | 0.08 |
| Other gram-negative bacilli | 0.91 | 0.87-0.95 | 0.001 | 0.92 | 0.87-0.97 | 0.002 |
| Gram-positive cocci | ||||||
| Staphylococcus aureus | 0.89 | 0.82-0.97 | 0.008 | 0.88 | 0.80-0.97 | 0.01 |
| Coagulase-negative staphylococci | 1.10 | 0.98-1.24 | 0.11 | 1.19 | 1.03-1.37 | 0.02 |
| Enterococcus spp. | 0.92 | 0.81-1.05 | 0.32 | 1.01 | 0.87-1.17 | 0.88 |
| Other gram-positive cocci | 0.87 | 0.72-1.06 | 0.23 | 0.95 | 0.78-1.15 | 0.61 |
RR: rate ratio, corresponding to the proportional increase (or decrease) in the incidence rate of Acinetobacter baumannii for each unit increase in the incidence of other pathogens. 95%CI: 95% confidence intervals. Note:Statistically significant results are presented in bold.
Although Poisson regression did not presuppose a linear correlation between the incidence rates, linear regression analysis showed a negative impact of Staphylococcus aureus (Figure 1), Enterobacter spp. (Figure 2) and several minor enterobacteria on the incidence A. baumannii. To illustrate these relationships, Figure 3 presents the time-series of the monthly incidence rates for S. aureus, Enterobacterspp. and A. baumannii.
FIGURE 1 Univariate correlation between the incidences of healthcare-associated bloodstream infections caused byStaphylococcus aureus (x-axis) and Acinetobacter baumannii (y-axis; Pearson’s correlation coefficient = -0,23). Note: Incidences are shown as infections per 10,000 patient-days.
DISCUSSION
Infection control practitioners are faced with several challenges in the control of multi-drug-resistant organisms (MDROs). Guidelines emphasize active surveillance and prompt institution of isolation precautions to contain the spread of MDROs throughout healthcare settings16,17. Nevertheless, most current evidence comes from studies focusing on gram-positive pathogens, such as methicillin-resistant Staphylococcus aureus (MRSA) and vancomycin-resistant enterococci16,18. However, a systematic review found insufficient evidence that these recommendations are effective19.
The debate becomes more complex when A. baumannii20 is considered. There are still gaps in our knowledge of the role of antimicrobial use, inanimate reservoirs, asymptomatic colonization and even continuous introduction from environmental sources outside healthcare settings21,22. In addition, the risk factors for the acquisition of drug-resistant A. baumannii isolates may vary with the extent of colonization pressure (i.e., the intensity of exposure to other patients harboring this agent) experienced by patients12. It is difficult to choose separate, appropriate strategies for fighting A. baumannii for each specific setting and period of time.
An ecological approach may help in delineating optimal strategies. We hypothesize that healthcare-associated pathogens compete for reservoirs, both animate and inanimate. Competition between microorganisms has been widely documented in nature23 and in the human microbiota24. It is plausible that the same phenomenon occurs within hospitals. If so, we can estimate trends in the incidence of A. baumannii (or any pathogen of interest) in relation to the epidemiology of other microorganisms. An understanding of these relationships may guide the choice of which microorganisms are priority targets for infection control efforts during specific periods. Such an understanding of these relationships may also further our understanding of the behavior of endemic pathogens (e.g., A. baumannii) when challenged with the introduction of emerging agents (e.g., carbapenem-resistant enterobacteria)25,26.
Our study found a negative association between the incidence of A. baumannii and the incidences of both S. aureus and Enterobacter spp. These findings are noteworthy because they are consistent with our impression, based on experience, that waves of MRSA and multi-drug-resistant A. baumannii infections alternate over time. The finding that competition exists between Acinetobacter baumannii and Enterobacter was unexpected, however, because a previous study in our hospital found that the incidences of both agents (A. baumannii and Enterobacter spp.) are seasonal, with peaks during the warmer months of the year7. The negative association between A. baumannii and other, low-incidence gram-negative pathogens was only perceptible when the latter were grouped together, due to a lack of statistical power when they were considered individually. Nevertheless, this finding suggest the possibility that other associations could have been identified if more cases occurred or if the sampling period had been longer. Our study also found a positive association between the incidence rates of CoNS and A. baumannii. While the reasons for this finding are unclear, it suggests that future studies should focus on the competition between staphylococcal species (both coagulase-positive and negative) and A. baumannii.
Our study was limited by the retrospective design and relatively short time span considered. We used a retrospective ecological design, and aggregate data may not be indicative of risks related to individuals (an ecological fallacy). However, ecological analysis is warranted when the implications of interventions are essentially collective, as is the case for many infection control policies27. Another limitation of our analysis is that the direction of causality could not be determined in this study. While we found negative correlations between the incidences of pathogens, we cannot be sure whether, for instance, the incidence of S. aureus (or Enterobacter spp.) affects that of A. baumannii or whether the opposite occurs. Thus, we must interpret the results in terms of their predictive capability, not in terms of causality. Finally, our study focused only on blood cultures. While this strategy may have lowered the statistical power of our study, we believe that blood cultures are more robust markers of infection than are other microbiological samples.
In conclusion, we found that the incidence of HA-BSIs caused by A. baumannii was negatively associated with the incidence of S. aureus, Enterobacter spp., and other gram-negative pathogens. Our findings suggest that competition between pathogens occurs in hospital settings and that the effects of this competition should be considered when developing infection control policies.
