| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1 Center for Anti-Infective Research and Development, Hartford Hospital, Hartford, CT 06102, USA
2 Division of Clinical Microbiology, Hartford Hospital, Hartford, CT 06102, USA
Correspondence
David P. Nicolau
dnicola{at}harthosp.org
Received 16 January 2008
Accepted 29 April 2008
0.001). The utilization of a carbapenem increased progressively over the course of the study. Moreover, the time to initiation of carbapenem therapy was reduced from 15.7±4.9 to 0.1±2.0 days (P 0.001) after implementation of this automated detection system. Overall, clinicians responded positively to the ESBL automated detection report, as gauged by the increased utilization of a carbapenem and the earlier initiation of appropriate therapy; however, reductions in length of stay and mortality were not observed in this infected population.
Abbreviations: ESBL, extended-spectrum β-lactamase.
 |
INTRODUCTION |
|---|
 TOP INTRODUCTION METHODSRESULTS AND DISCUSSIONREFERENCES |
|---|
A previous study at our institution conducted between October 2001 and 2002 observed that the prevalence of ESBLs was 6.1 % at Hartford Hospital (Dandekar et al., 2004). Prior to this study, clinician consensus was that the ESBL phenotype did not exist locally. It was later noted that our patients infected with ESBL-producing organisms had increased lengths of hospitalization and the likelihood of initial antibiotic course failure (Lee et al., 2006). In accordance with CLSI (2006) recommendations, an automated screening/confirmation (two-step) testing process was instituted in 2002 using the Microscan Walkaway (Dade Behring) automated susceptibility testing system. Currently, it is unknown whether the implementation of such an automated testing process will improve the appropriateness of therapy, patient outcomes or both; therefore, we undertook the following study to assess the response of clinicians to the ESBL report and its subsequent impact on patient care.
|
METHODS |
|---|
|
TOPINTRODUCTIONMETHODSRESULTS AND DISCUSSIONREFERENCES |
|---|
Patients were excluded if they were discharged, died or were placed on palliative care prior to or on the day of the ESBL report. Patients were also excluded if they were colonized with an ESBL. Colonization was considered when any patient was detected with an ESBL via a superficial culture and any combination of the following circumstances occurred: a negative deep-site culture (e.g. bronchoscopic alveolar lavage/bronchoscopy, deep-wound culture), no signs/symptoms of infection (e.g. leukocytosis, fever, purulent secretions), negative radiological findings or no antibiotic treatment directed at the positive culture. Only the first confirmed non-urine ESBL culture of each patient was considered for inclusion.
The hospital's institutional review committee approved this study. An informed consent waiver was granted as all data were currently in existence and no patient-specific interventions were conducted for the retrospective study. The collection of data was in compliance with the Health Insurance Portability and Accountability Act of 1996.
Microbiological testing. Susceptibility results for all relevant antibiotics (aztreonam, cefoxitin, ceftazidime, cefotaxime, ceftriaxone, cefepime, piperacillin/tazobactam, meropenem, imipenem, ciprofloxacin, levofloxacin, gentamicin, tobramycin, and trimethoprim/sulfamethoxazole) were tested and interpreted using the Microscan Walkaway system following CLSI (2006) guidelines. This also includes the methods explained below that were used to confirm ESBL phenotypes.
Prior to November 2002, the automated susceptibility testing device only screened Klebsiella species and E. coli isolates for ESBLs if they were resistant to ceftazidime (MIC >8 µg ml–1). Along with this, a message report of the final microbiology laboratory results from the culture would be given on the hospital's computerized provider order entry system, stating that the organism may be an ESBL producer and clinically resistant to all cephalosporins and aztreonam. After November 2002, a 2-day process was performed, where an initial screening occurred for all Klebsiella species and E. coli isolates with an MIC 
2 µg ml–1 for ceftazidime or cefotaxime. The next day, the presence of ESBL phenotypes was confirmed by determining a threefold change in MIC following the addition of clavulanic acid to ceftazidime or cefotaxime. The reporting and notifications of the ESBL were similar except that the message now confirmed the organism as an ESBL producer.
Measurements and definitions. Two definitions of appropriate treatment for ESBLs were explored during the analysis. The primary definition of appropriate therapy was defined as the initiation of a carbapenem antibiotic. Secondly, we analysed outcomes defining appropriate treatment as the prescribing of any antibiotic (e.g. fluoroquinolones, trimethoprim/sulfamethoxazole and aminoglycosides) that the organism was reported as being susceptible to. The clinician's acceptance of the ESBL report (i.e. the date the final report was released) was defined as a change in antibiotic therapy to an appropriate therapy as defined above. A patient treated prior to the report with a carbapenem or antibiotic that the organism was susceptible to, and remaining on that therapy after the report, was considered to have met the definition of acceptance. The time to appropriate antibiotic therapy was measured from the availability of the ESBL report to the initiation of appropriate treatment. Infection-related length of stay was measured from the date of culture to the last antibiotic received for the ESBL infection. All-cause mortality was defined as death due to any reason at the end of hospitalization, whilst infection-related mortality was defined as death while receiving antibiotics for the initial ESBL infection, without any other obvious cause of death.
Data collection. Once patients were identified, a single investigator conducted the data collection by reviewing the medical records for all patients. Data were collected using a standardized tool that included patient demographics (age, gender, race), hospital admission date, hospital discharge date, status at discharge (alive or dead), date of ESBL culture, source of culture, type of infection (e.g. pneumonia, bloodstream infection, wound infection), all microbiology reports, antibiotic courses (drug, dosage, start date/time, stop date) within 30 days prior to and after detection of the ESBL, location within the hospital and the request of an Infectious Diseases consult. The acute physiological and chronic health evaluation (APACHE) II score and the Charlson co-morbidity index were calculated within 24 h of the date of the ESBL culture (Knaus et al., 1985; Charlson et al., 1987).
Analyses.
Dichotomous variables (e.g. gender, race, infection site) were compared between groups using a 
2 test. Continuous variables (e.g. age, APACHE II score, length of hospital stay) were compared using one-way analysis of variance or a Mann–Whitney rank sum test, where appropriate. The proportion of patients who met the definition of clinician acceptance of the ESBL report within each annual group was compared using a 2 test. The numbers of patients who received an appropriate antibiotic and had an Infectious Diseases consult as a result of the ESBL report were calculated and compared over each year. All statistical analyses were conducted using SigmaStat 2.03 (SPSS). A P value of less than 0.05 was defined as statistically significant.
|
RESULTS AND DISCUSSION |
|---|
|
TOPINTRODUCTIONMETHODSRESULTS AND DISCUSSIONREFERENCES |
|---|
Table 1
demonstrates the demographics and characteristics for the included patients separated by cohort. The Charlson co-morbidity score was significantly greater among patients in the post-detection cohort, which was consistent with a trend towards a greater APACHE II score in this group. Moreover, all patients in the pre-detection cohort were infected with Klebsiella species, whilst 30 % of patients in the post-detection cohort were infected with E. coli. Apart from these differences, the patient groups were quite similar. Of note, the most common site of infection was the lung.
|
, report acceptance based on the use of a carbapenem was significantly greater in the post-automated compared with the pre-automated detection years (69.2 vs 20 %, P 0.001). When observed more closely, a general trend towards higher acceptance rates based on the prescribing of a carbapenem was noticed after 2002 (Fig. 1). However, when defining appropriate therapy based on the use of any agent that the organism was susceptible to (Table 2), there was no difference (80.8 % vs 66.7 %, P=0.301) in acceptance of the ESBL report. The time in which patients received appropriate antibiotics was <1 day for both definitions (agents that the organism was susceptible to and carbapenem only) in the post-detection cohort. This was in contrast to the pre-detection period, which showed a 5-day delay for agents that the organism was susceptible to and a 15-day delay until the initiation of a carbapenem (P 0.001 for both).
|
|
0.001 and P 0.002, respectively). Whilst most patients were already under the care of an Infectious Diseases clinician at the time that the ESBL was reported, there were a greater number of new requests for Infectious Diseases consults after the report in the post-detection cohort (16.7 vs 0 %, P=0.118). Despite these findings, there were no differences noted in patient outcome. Currently, no studies are available to document the utility of ESBL automated screening and confirmation. In this study, we were able to observe that clinician acceptance of the ESBL report with automated detection did occur after implementation of the automated testing system, as noted by the increased utilization of a carbapenem. However, clinicians were not influenced by automated detection when measured solely on the use of an agent that the infecting organism was deemed susceptible to. This is probably because clinicians often choose an agent that the organism is susceptible to according to the microbiology report, regardless of the presence of an underlying resistance mechanism such as an ESBL. As carbapenems are generally regarded as the antibiotics of choice for the treatment of non-urine ESBL infections, it is not surprising that their use remained low during the time period when ESBLs were not automatically confirmed.
One study observed the influence of non-automated microbiological testing (disc diffusion) of ESBLs on antimicrobial therapy (Gavin et al., 2006). The authors reported that 94 % of the patients' antibiotics were changed to appropriate therapy (based on the agent's susceptibility) after the ESBL was detected and reported. Although a similar rate of appropriate therapy (88.4 %) was observed in our study, significance was only noted when therapy was changed to a carbapenem. Moreover, the carbapenems were the antibiotics prescribed most often in our study and their utilization has increased year after year since the implementation of the automated ESBL system.
The other important finding in our study was the time in which many of these changes in therapy occurred. In general, the pre-detection cohort was shown to have a >5-day delay in the initiation of appropriate therapy from the time of report. These staggering numbers are even more profound when it is noted that Infectious Diseases clinicians were caring for >50 % of these patients prior to the automated report and only about one-third of these patients received a carbapenem. Because Infectious Diseases consultants had the capability of prescribing a carbapenem without antibiotic restrictions, the delay in prescribing would not have reflected a delay due solely to authorization. Once automated detection was implemented, this delay in therapy diminished and the prescribing of a carbapenem increased progressively.
When data for the request of Infectious Diseases consultations were analysed, it was noted that significantly more consultations were undertaken in the post-automated detection cohort. These results initially indicate that automated detection of ESBLs allowed more requests for Infectious Diseases consultations; however, a closer observation suggests otherwise. A substantial amount (83.8 %) of the Infectious Diseases consultations occurred several days to weeks prior to the ESBL report; this is possibly the result of the severely ill population, the frequency of prior infections, nosocomial acquisition of the ESBL or some combination of these factors. Additionally, a greater number of patients in the post-detection phase were admitted from a specialized nursing facility, which is associated with increasing prevalence of infections with antimicrobial-resistant organisms and multidrug-resistant outbreaks (Wiener et al., 1999; Bonomo, 2000; Crnich et al., 2007). With Infectious Diseases clinicians involved in >90 % of the cases in the post-detection cohort, we were mainly measuring the response of these specialists to the ESBL report. Additionally, patients with documented ESBL infections received appropriate therapy with a carbapenem more frequently, and the initiation of appropriate therapy was earlier after the implementation of the automated system. However, this did not lead to significant reductions in length of stay or mortality. The overall mortality rates (infection related and all-cause) were similar in both cohorts. Moreover, the length of stay (overall hospital, intensive care unit and infection-related stay) in the two cohorts was also similar (Table 2
).
The major limitation of this work is that it is a single-centre study and these results may not be reflective of outcomes and practices at other institutions. Another factor is the small patient population (15 patients) in the pre-detection cohort, which may have prevented us from obtaining more pronounced differences in clinical outcomes. Additionally, the retrospective design made it difficult to capture the responses (i.e. antibiotic choices) of non-Infectious Diseases clinicians to the ESBL report. Whilst a full economic evaluation of this automated testing system has not been undertaken, the fact that reductions in the length of stay (the largest component of hospital costs) were not detected suggests that it is unlikely that this system would be deemed cost-effective at our institution.
The use of an automated ESBL screening and confirmation system showed significant improvements in the prescribing of a carbapenem by clinicians, most of whom were Infectious Diseases consultants. Specifically, patients not only received these perceived drugs of choice more frequently, but therapy was also initiated more quickly after system implementation. Contrary to the lack of differences in clinical outcomes (i.e. length of stay and mortality), the MicroScan Walkaway automated susceptibility testing improved awareness at our institution for patients at risk of infection from ESBL-producing members of the Enterobacteriaceae, as shown by the increased carbapenem utilization and early appropriate therapy. Overall, additional investigations should be undertaken to confirm these observations at other institutions.
|
ACKNOWLEDGEMENTS |
|---|
|
REFERENCES |
|---|
|
TOPINTRODUCTIONMETHODSRESULTS AND DISCUSSIONREFERENCES |
|---|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| INT J SYST EVOL MICROBIOL | J MED MICROBIOL | MICROBIOLOGY | J GEN VIROL | ALL SGM JOURNALS |
