![[Feedback]](/icons/banner/feedbackACT.gif){kind=icon}
![[For Subscribers]](/icons/banner/subscriberACT.gif){kind=icon}
![[Archive]](/icons/banner/archiveACT.gif){kind=icon}
![[Search]](/icons/banner/searchACT.gif){kind=icon}
![[Contents]](/icons/banner/tocACT.gif){kind=icon}
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Previous Article | Next Article 
Antimicrobial Agents and Chemotherapy, March 2007, p. 839-844, Vol. 51, No. 3
0066-4804/07/$08.00+0 doi:10.1128/AAC.00901-06
Copyright © 2007, American Society for Microbiology. All Rights Reserved.
Department of Medicine, University of Maryland, Baltimore, Maryland,1 Department of Epidemiology and Preventive Medicine, University of Maryland, Baltimore, Maryland,2 VA Maryland Healthcare System, Baltimore, Maryland3
Received 21 July 2006/ Returned for modification 21 November 2006/ Accepted 17 December 2006
 |
ABSTRACT |
|---|
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
INTRODUCTION |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
In general, it is reasonable to assume that the administration of antibiotics directed at the causative pathogen as early as possible in the course of infection would improve patient outcomes. This principle often triggers physicians to use broad-spectrum antipseudomonal agents in cases of presumed sepsis, which may lead to overuse of these drugs, and a subsequent increase in adverse events, increased costs, and antimicrobial resistance. It is therefore important to ensure that this principle is based on a true association. However, this hypothesis has not been clearly verified in P. aeruginosa bacteremia. Although several studies have found a distinct association between adequate empirical therapy and outcomes in cases of P. aeruginosa bacteremia (1, 3, 25), others have suggested that such associations do not exist (2, 13, 22, 23).
These studies are, by necessity, observational in nature, and therefore adequately controlling for confounding variables is more difficult. In addition, methodological differences in the way these studies were conducted may explain the conflicting results. First, the term "empirical therapy" is not defined consistently across the literature, and the time at which empirical therapy was assessed varied. Second, aggregate scores used to control for severity of illness were often measured at a time when they did not accurately reflect a patient's baseline risk of mortality. Third, many studies on this subject were designed to evaluate the outcomes of bacteremias caused by gram-negative organisms in general and thus had smaller numbers of patients with P. aeruginosa bacteremia, thereby decreasing their statistical power for this subgroup.
Our study sought to evaluate the effect of inappropriate empirical therapy at three distinct time points on mortality and length of stay among patients with P. aeruginosa bacteremia, while controlling for severity of illness before the onset of the bacteremia.
|
MATERIALS AND METHODS |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Variables electronically collected included age, gender, blood culture source, date and time the blood culture of interest was obtained, patient location at the time of blood culture, length of stay (index blood culture collection date to hospital discharge date), time prior to blood culture collection (hospital admission date to index blood culture collection date), time to susceptibility results (index blood culture collection date to date of receipt of antibiotic susceptibility testing results), and susceptibility results for the organism. We identified all coinfecting species and their resistance profile for patients who had more than one organism cultured out of the index blood culture (polymicrobial bacteremias).
Medication administration records for each patient were examined to determine whether and when the patient received appropriate empirical antibiotics. Additional information such as vital signs, the presence of a ventilator at admission or at culture collection, and data about central lines were also collected from patient records.
All microbiological data were collected retrospectively through the UMMC central data repository, and all blood cultures and susceptibility tests were performed by the UMMC clinical microbiology laboratory as part of the patients' standard care, according to the recommendations of the Clinical and Laboratory Standards Institute (formerly the National Committee for Clinical Laboratory Standards) (26).
Variable definitions. Empirical antibiotic therapy was defined as the antibiotic therapy received by the patient between 8 h before the index blood culture was drawn and the time antibiotic susceptibility testing results were available. Empirical antibiotic therapy was considered appropriate if it included antipseudomonal antibiotics to which the specific Pseudomonas isolate displayed in vitro susceptibility. Although aztreonam may be effective for P. aeruginosa bloodstream infections, neither aminoglycoside nor aztreonam monotherapy were considered adequate therapy for P. aeruginosa bacteremia in the present study and were classified in the inappropriate group (12, 27). To more completely evaluate the impact of empirical therapy, we assessed adequate therapy in three distinct windows: between 8 h before the time the culture was obtained and 24 h afterward, between 24 and 48 h after the culture was obtained, and from 48 h after the culture was obtained to 4 h after the time the antibiotic susceptibility testing results were available from the UMMC microbiology laboratory. Susceptibility results are updated automatically in the electronic patient record available to the clinician as soon as they are uploaded onto the laboratory database by the technician. The prior positive blood culture result then gets flagged as a new result, prompting the clinician to access the newly updated information. Patients who received appropriate therapy during one time period were considered to have received appropriate therapy for all subsequent time periods. For example, if a patient started receiving appropriate therapy between 24 h and 48 h after the index blood culture was obtained, this patient was considered to be on appropriate therapy at the subsequent time-period (from 48 h to receipt of susceptibility results) as well.
We collected data on the presence and removal of central lines. Central lines were defined as all central venous catheters, peripherally inserted central catheters (PICC lines), dialysis catheters, and medication ports. The nursing flow sheets were reviewed to identify the presence of a central line at the time of culture collection and at 24 and 48 h after that time point. Multiple lines at the time of culture were not considered separately and were coded as removed only if all of the lines present at culture had been removed.
Severity of illness prior to the bacteremia of interest was assessed by calculating a modified acute physiology score (APS) based on the APACHE III score (19) at admission and again at 24 h before the time the culture was obtained. We chose this latter time point in order to make sure that the aggregate score accurately reflects the baseline severity of illness for each patient and did not include values that occurred as a consequence of the bacteremia. The APACHE III score was designed for use among intensive care unit (ICU) patients. Since the present study included participants who may not have been in an ICU at the time of the bacteremia, we modified the score by excluding variables that were irrelevant to our study population such as pulmonary arterial gradient, urine output, neurological status, and ventilator data.
The presence of preexisting comorbid conditions was determined using the chronic disease score (CDS), which utilizes patient medications as indicators of the existence of comorbid conditions (32). In the present study, the CDS was calculated based upon the patient medications ordered within the first 24 h of hospital admission, as has been done in other studies (17, 24).
Statistical analysis. All data were collected and entered into a Microsoft Access database (Microsoft Corp., Redmond, WA) and analyzed by using SAS (SAS Institute, Cary, NC) software version 9.1. The Fisher exact test and chi-square test were used to compare categorical variables, and the Student t test was used for continuous variables. The Wilcoxon rank-sum test was used to test the difference between the median values for non-normally distributed continuous variables.
Multivariable regression was used to control for confounding variables. Since the APS at the time of culture collection and the APS at the time of admission are inherently correlated variables, they were entered into the regression models separately during analysis. All biologically plausible variables with a P value of <0.20 in the bivariable analysis were considered for inclusion in the final multivariable regression models. As our primary variable of interest, appropriate empirical antibiotic therapy was forced into all multivariable models. Multivariable analysis was performed using backward selection logistic regression for the outcome of death. Linear regression was used for the outcome of length of stay (LOS). Because of the non-normal distribution of LOS, this variable was natural log transformed in the linear regression analyses. In this form, the exponentiated model coefficients represent the ratio of the natural log of LOS between levels of the independent variables. Patients who died in the hospital were excluded from the LOS analyses because their hospital stay was shortened by death. All tests of significance were two-tailed, and P values of 
0.05 were considered significant.
|
RESULTS |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
Of the isolates recovered during the study period, 82.6% were susceptible to piperacillin-tazobactam, 83.2% were susceptible to imipenem, and 84.5% were susceptible to cefepime. Twenty-seven isolates were tested for susceptibility to polymixin B, and none of these were resistant (Table 2).
|
Hospital mortality. Sixty-one (36.5%) patients died during their hospital stay. In bivariable analysis, there was a trend toward a protective effect of receiving appropriate antibiotics between 8 h before and 24 h after the index blood culture collection (odds ratio [OR] = 0.88; 95% confidence interval [CI] = 0.38 to 1.4), between 24 and 48 h (OR = 0.68; 95% CI = 0.33 to 1.4), and at the time susceptibility results were available (OR = 0.71; 95% CI = 0.26 to 1.93). Compared to survivors, hospital nonsurvivors were significantly more likely to be older (OR = 1.02; 95% CI = 1.00 to 1.05), ventilated at the time of culture (OR = 2.14; 95% CI = 1.09 to 4.20), and have a higher modified APS at admission (OR = 1.04; 95% CI = 1.01 to 1.07) and at the time of culture (OR = 1.06; 95% CI = 1.03 to 1.09). In addition, time prior to blood culture collection was significantly longer in patients who died compared to those who survived (11.05 days compared to 4.3 days; P = 0.01).
In the multivariable logistic regression analysis (Table 3), after we adjusted for the APS at time of culture and age, there was a trend toward a protective effect of appropriate antibiotics at 24 h (OR = 0.93; 95% CI = 0.45 to 1.92), at 48 h (OR = 0.66; 95% CI = 0.29 to 1.49), and at the time susceptibility results were available (OR = 0.96; 95% CI = 0.31 to 2.93). The variables age and APS were modeled continuously, such that the odds ratios are representative of a 1-year or one-point increase, respectively.
|
In multivariable linear regression analysis, there was no association between receiving appropriate therapy between 8 h before and 24 h after the time of culture and an increase in LOS (ratio of LOS = 1.08; P = 0.66). Patients who received appropriate antibiotics at subsequent windows, i.e., between 24 and 48 h after the culture was obtained (ratio of LOS = 0.92; P = 0.62) and from 48 h until the time the susceptibility results were available (ratio of LOS = 0.93; P = 0.74), also did not have a significant change in their length of stay (Table 3).
Significant predictors of longer LOS in the multivariable analysis were time before blood culture collection (P < 0.01), the presence of a ventilator at admission (P = 0.03), and the presence of a ventilator at culture (P < 0.01).
Forty-eight (28.7%) patients had a positive blood culture within 12 h of their hospital stay. These patients had a lower mortality than patients whose index blood culture was collected later in their hospital stay. For these patients, in a multivariable model, APS at admission was a predictor of mortality (OR = 1.06. 95% CI = 1.03 to 1.09), and receiving appropriate therapy in the first 24 h was not statistically significantly associated with mortality (OR = 0.79, 95% CI = 0.43 to 1.44).
|
DISCUSSION |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
As expected, the independent predictors of mortality in our cohort were age and APS at culture. Increased time prior to blood culture collection (time from hospital admission to when the index blood culture was obtained) was associated with an increased length of stay. The presence of a central line at the time of culture was also an independent predictor of an increased length of stay.
The belief that early appropriate therapy should improve survival is intuitive but has not been clearly established in the literature. In fact, several studies of the influence of appropriate therapy in bloodstream infections have failed to find an association between adequate therapy and mortality. Zaragoza et al. found no difference in mortality in 166 ICU patients with bacteremia (34). Likewise, Bryan et al. (2) reported that appropriate therapy on the first calendar day after culture did not improve survival in an analysis of 1,186 episodes of gram-negative bacteremia. In contrast, these authors found that appropriate therapy, when started on the second calendar day after culture, had a protective effect compared to patients who received appropriate therapy from the onset of bacteremia. These researchers attributed these findings to patient factors, hypothesizing that patients who had survived the initial period of bacteremia without treatment were also more likely to survive their hospital stay. Several other large studies of gram-negative bacteremia have shown a statistically significant association between appropriate antibiotics and mortality (5, 13, 15, 20-23, 30). However, these studies included all gram-negative organisms and thus may have been underpowered to show a difference in survival for those patients who specifically had P. aeruginosa bacteremia.
To our knowledge, despite several publications on patient outcomes in P aeruginosa bacteremia (1, 3, 4, 16, 31), only two other studies were specifically designed to evaluate the impact of appropriate empirical antibiotics on the outcomes of P. aeruginosa bacteremia. Micek et al. reported on 305 patients with P. aeruginosa bacteremia and concluded that the association between appropriate initial therapy and hospital mortality was statistically significant but only after the variable for comorbidity (all patient refined diagnosis related group, APR-DRG) was removed from the multivariable model, which is not epidemiologically a suggested practice (25). On the other hand, Kang et al. established that the 30-day mortality in patients who experienced a delay in receiving appropriate antibiotics (after excluding those who did not get appropriate definitive therapy) tended to be higher, but this was not statistically significant, a finding similar to our results (14).
These inconsistent findings may be attributable to variations in the methodology used, differences in the study population, or differences in the distribution of other confounding variables among the studies. In reviewing the methodology, the studies differed in their definition of empirical therapy, the time at which severity of illness was measured, and whether septic shock was controlled for in the final regression models.
Antibiotics are generally defined as empirical if they are chosen before susceptibility results are known. In practice however, microbiology laboratories often report preliminary results of positive blood cultures (e.g., non-lactose-fermenting gram-negative rod), prompting treating physicians to change antibiotics based on an increased index of suspicion for one or the other organisms. Thus, the time point at which therapy is truly empirical is difficult to define. A cross-sectional look at the adequacy of therapy can misclassify patients who either were already receiving early definitive therapy or those who had received appropriate therapy earlier in their course. Micek et al. defined inappropriate antimicrobial treatment as "microbiological documentation of infection that was not adequately treated at the time the causative microorganism and its antibiotic susceptibility were known." This definition therefore incorporates both empirical and definitive therapy. Kang et al. defined empirical therapy as effective if administered within 24 h after the blood culture was obtained or delayed if more than 24 h had elapsed. However, this definition did not take into consideration any antibiotics administered before the blood culture was collected, nor did it extend the assessment period to the time at which susceptibility results were available and thus did not analyze the full empirical therapy window.
Severity of illness may be an important confounder in patients who have bacteremia. A high severity of illness at the onset of the bacteremia may lead to a higher mortality but may also lead to more aggressive treatment early on. Ideally, when controlling for severity of illness as a predictor for mortality, it should be assessed before the onset of the bacteremia and would therefore reflect the patient's baseline risk of mortality. Measuring severity of illness exactly at the time the blood culture is drawn or any time after the onset of bacteremia would include variables that are a consequence of the current illness and therefore be in the causal pathway to mortality. These should not be controlled for since they would lead to an underestimate of the effect of inappropriate therapy on outcome (28). Likewise, respiratory failure and circulatory shock are usually the result of bacteremia and should not be adjusted for in the analysis (7, 29). Both of the previously published studies measured the severity of the illness at the time of bacteremia and adjusted for septic shock in their analyses.
In the present study, we calculated the severity of illness at a time point 24 h before the index blood culture was obtained whenever possible, therefore avoiding the inclusion of intermediate variables in our analysis. We also assessed appropriate therapy at three distinct time points, taking into account antibiotics administered before the blood culture was obtained and at various time points until the susceptibility results were available.
Our study was limited by a relatively low statistical power due to the fact that few patients in our cohort received inappropriate empirical therapy. This may be explained by the fact that most of the isolates of P. aeruginosa included in our study were susceptible to piperacillin-tazobactam, a beta-lactam antibiotic that is recommended for use in our institution in cases where a nosocomial or gram-negative infection is suspected. Although P. aeruginosa is rarely a contaminant of blood cultures (33), it is possible that some of the patients in this study were incorrectly classified as having a true bacteremia. We did not have data on the number of blood cultures positive with P. aeruginosa and thus were unable to try to sort out true P. aeruginosa pathogens from contaminants. Also, for patients who had polymicrobial bacteremia, we were unable to completely assess whether all coinfecting species were being appropriately treated along with the P. aeruginosa bacteremia. However, statistical analyses where patients with polymicrobial bacteremia were excluded did not yield any different results. Hospital mortality may also be influenced by inappropriate definitive therapy or early termination of appropriate therapy, neither of which we assessed in our study.
In conclusion, P. aeruginosa bacteremia is an important cause of hospital mortality but the effect of appropriate empirical therapy on survival is still unclear. We found that appropriate empirical therapy over time had a trend toward decreasing mortality and length of stay but neither was statistically significant. Thus far, none of the studies designed to assess the impact of appropriate empirical therapy on outcomes have been able to show a protective effect of appropriate empirical therapy on mortality without controlling for intermediate variables such as septic shock and severity of illness or not, including potentially important confounding variables in their final analyses. Therefore, further studies designed to evaluate the exact time point at which administration of an appropriate antibiotic is critical to survival are necessary. Identifying the time at which appropriate therapy becomes critical to patient outcomes is particularly important because of the delicate balance that exists between the beneficial effect of appropriate therapy on patient survival and the development of antimicrobial resistance due to excessive use of broad-spectrum antimicrobial agents.
|
ACKNOWLEDGMENTS |
|---|
We thank Colleen Reilly and Jingkun Zhu for database maintenance and abstraction.
|
FOOTNOTES |
|---|
Published ahead of print on 28 December 2006.
|
REFERENCES |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Clin. Vaccine Immunol. | Clin. Microbiol. Rev. |
|---|---|
| J. Clin. Microbiol. | ALL ASM JOURNALS |
