Strategies of initiation and streamlining of antibiotic therapy in 41 French intensive care units

Initiation of antibiotic therapy (AT) in intensive care unit (ICU) patients is a critical issue. The importance of empiric AT covering all pathogens responsible for infections has been highlighted on many occasions [1‐4]. The need for urgent AT was also emphasized in a study demonstrating a 7% increased mortality for each hour of delayed empiric AT in patients with severe sepsis and septic shock [5]. The time to the first dose of AT has been emphasized in the recommendations of the surviving sepsis campaign [6] and has become a measure of quality of care in ICU patients [7‐9]. The difficulty in differentiating infectious from noninfectious etiologies in critically ill patients is also a major driver of antibiotic prescribing in ICUs leading to the development of new diagnostic tests [10]. On the other hand, the parsimonious choice of AT drugs has also been stressed to curtail the emergence of resistance and contain the cost [11, 12].

Most studies addressing the issue of AT have focused on appropriateness, while few longitudinal surveillance studies have analyzed the decision-making process [1, 2, 4, 13‐15]. To more clearly understand AT current prescribing practices in ICU patients, a prospective multicenter observational study was performed to describe the modalities of initiation (frequency, timing) of AT, the reason for changes (streamline/de-escalate therapy) and identification of independent factors associated with mortality in patients receiving new AT during their ICU stay.

To our knowledge, this study represents the largest cohort addressing the AT decision-making process in ICU patients. More than 60% of patients received AT during their ICU stay and one third of them required new AT initiated out-of-hours in half of the cases.

Observational studies have their own limitations. A limited number of centers participated in the survey with heterogeneous activity and case-mix in teaching and non-teaching institutions. All microbiology laboratories followed the same guidelines published by the French Society of Microbiology [21], decreasing the heterogeneity of the management and decision-making process. The duration of the study was not sufficient to take into account seasonal changes in antibiotic prescriptions. In the study design, the decision-making process was deliberately addressed rather than considerations linked to the quality of antibiotic prescription in terms of pharmacokinetic/pharmacodynamic (pK/pD) parameters or adherence to local protocols. This issue might be relevant as the lack of correlation with microbiologically appropriate AT could be due to poor quality of the antibiotic prescription. In addition, the delay in starting new AT was not documented. This is a critical point in addressing the issue of relationship between mortality and AT and admittedly is a weakness of our study. No distinction was made between community-acquired and nosocomial infections. Finally, methodological issues could be considered as limitations. This is the case for appropriateness of antibiotic therapy assessed by local investigators, the duration of antibiotic therapy not determined and hospital mortality not assessed. Consequently, the results of this study should be interpreted cautiously, although this descriptive study can be assumed to reflect "real life" conditions.

In a single-center prospective study, Bergmans et al. reported that 36% of patients had at least one infection during their ICU stay and were treated for infection on 48% of all patient-days [14]. In a 15-month study in a surgical ICU using computerized patient data management systems, Hartmann et al. observed that 58% of the patients received AT [24]. In a single-center prospective audit, Warren et al. reported that 77% of admissions received at least one AT during their ICU stay [13]. In this paper, 17% of AT were initiated prior to ICU admission and 45% of patients received antibiotics for suspected or proven sepsis [13]. In a study performed in 23 Swedish ICUs over a two-week period, the median proportion of patients receiving antibiotics was 74% (range 24 to 93%); 64% of all prescriptions corresponded to empiric AT with only minor differences between units [15]. In a Turkish six-month single-center study, AT was prescribed in 61% of all admissions and empiric therapy accounted for 46% of cases [25]. In the EPIC II study, 9,084 (71%) of 13,796 adult patients in 1,265 ICUs from 75 countries were receiving antibiotics in this point prevalence study [26].

In more than 70% of our patients receiving AT, treatment was initiated before the results of Gram-stained direct examination and at the time of direct examination in more than 90% of these patients. In a prospective Spanish multicenter study in severe sepsis, the authors observed that 66% of patients received broad-spectrum antibiotics during the first six hours after presentation [7]. In a recent French multicenter study performed over six months in 2006, the authors reported that antibiotic therapy was administered within the first three and six hours following the diagnosis of severe sepsis or septic shock in 46% and 61% of patients, respectively [27]. Other studies addressing the delay of AT have reported similar observations of treatments administered within the first three to six hours in 60 to 86% of patients [8, 9].

Heterogeneity of practice with regard to microbiological sampling was not a surprise. In a previous observational study addressing the treatment of postoperative pneumonia, we reported that 14% of the patients received empiric AT without pulmonary samples having been taken [28]. While half of these of patients were hospitalized in ICU at the time of diagnosis only 6% of them developed ventilator associated pneumonia. These were mainly the less severe cases. The second major source for early treatment without sampling was the absence of round-the-clock microbiological laboratory facilities. This was not the case in our study where all ICUs had direct access to the laboratory.

To our knowledge, very few studies have evaluated whether initiation of AT during out-of-hours modifies the appropriateness of antimicrobial therapy. We hypothesized that out-of-hours could be associated with a lower proportion of appropriate AT, especially among the least experienced ICU physicians. Interestingly, no such differences were observed considering appropriateness of AT or outcome. This point was also not observed in centers without written guidelines. However, the proportion of fellow prescribers is too small to draw any conclusions. Inexperienced physicians may have a tendency to start a broad spectrum AT regime and this perhaps explains why no correlation between level of training and appropriateness was found. Local protocols and guidelines might play a protective role in that more antimicrobials are prescribed more securely by on-call doctors and more often at the beginning of infection probably ensuring earlier initiation of treatment.

Defining appropriateness of AT is a major challenge. This issue can be assessed in many ways. Gyssens et al. have developed an interesting algorithm to assess comprehensively the quality of antibiotic prescriptions [29]. Basing it only on a match between the antibiotic given and the results of susceptibility testing is the commonest approach used in the literature and makes sense with regard to patient outcome in severe infections. However, this mode of prescribing is perhaps short-sighted. Even if broad spectrum AT is much more likely to be "appropriate" than limited spectrum AT in the circumstances, the ecologic issues and risks of emergence of resistance with such a policy are major concerns.

De-escalation following AT appears to vary considerably, depending on the initial diagnosis from 23% of all antibiotic prescriptions [13] to 64% in patients with septic shock [4]. However, in many instances, no microbiologic confirmation is obtained or susceptibility testing is not available, which raises the issue of de-escalation. This has been frequently demonstrated where there is suspicion of pulmonary infection, as many noninfectious processes present with lung infiltrates and fever, falsely attributed to pneumonia [30, 31]. In ventilator associated pneumonia, as many as 30% of clinically suspected cases are not confirmed microbiologically [32], while in surgical ICU patients, Singh et al. [33] reported that only 30% of pulmonary infiltrates were the result of pneumonia. De-escalation is, therefore, problematic in these cases [34] and should be considered cautiously especially in therapeutic emergencies. In the absence of confirmation of infection (for example, negatives cultures in a patient already receiving AT), de-escalation is difficult and the appropriateness can only be evaluated by compliance to the protocols.

The proportions of appropriate AT in ICU patients are usually situated in the range of 70 to 80% of cases [1, 2, 4, 9] and up to 89% in some specific diagnoses [4]. The proportion of documented septic episodes was only slightly greater than 50% in our study and evaluation of appropriateness was based on documented cases. In the study by Kumar [5], appropriateness was also evaluated in non-documented infections by comparing the treatment to local written guidelines.

The absence of a significant link between mortality and appropriateness of AT is somewhat surprising and appears to contradict one the findings of the study: the lack of treatment protocols was an independent risk factor for increased mortality. An explanation for this paradox could be linked to the heterogeneity of the study population involving an insufficient number of patients to reach a significant threshold to observe an effect of inappropriateness. Previous studies demonstrating the importance of appropriateness from AT usually used larger cohorts of patients [1, 2, 5, 35] or analyzed selected populations with a single disease [3, 4, 35, 36]. The role played by young prescribers might also be considered. Inexperienced physicians as mentioned earlier may rather have a tendency to start a broad therapy regime which might explain why no correlation between level of training and appropriateness was found. In addition, the possibility of misdiagnosis cannot be excluded, since appropriateness of AT did not include this criteria. Information about delays in initiating AT would also have been of value in explaining our observations.

Many reports have shown that the use of antimicrobial guidelines was associated with improved appropriate antibiotic use, decreased duration of AT, reduced antibiotic costs and could decrease mortality, as observed in hospital-acquired pneumonia [37, 38]. The use of guidelines could be a surrogate marker for a better quality of care in general in the ICUs, thereby explaining the link between availability of guidelines and prognosis. This point was clearly emphasized in the studies assessing the effects of implementation of the guidelines of the surviving sepsis campaign [27, 39]. In view of the limitations of our work, the observations about the use of guidelines and prognosis should be considered cautiously. However, it is of interest to notice that among the risk factors of mortality identified in our patients, the absence of AT protocols was the only parameter that could be easily modified, as the other risk factors were linked to underlying disease or severity at the time of AT.

In view of our data, almost 20% of antibiotic prescriptions might be unnecessary in patients with suspected infection. In this setting, antimicrobial stewardship programs might be useful [40]. Developing protocols in association with infection control measures could be considered a first step of improving antibiotic use.

Of all indications for AT, the respiratory tract is by far the most important site of infection accounting for 47% of all infections in our cohort, and almost half of these cases corresponded to severe hypoxemic pneumonia. These pulmonary infections are the most frequent reason for AT in ICUs, reported in 43 to 51% of cases [13, 14], and up to 60% in a context of septic shock [4]). The frequency and severity of these cases might justify large-scale diffusion of local protocols concerning this specific issue.

In view of the limited number of publications on this topic, our results should be of interest to clinicians in the field. Our observations show that more than half of the patients admitted in ICU received antibiotics during their stay, half of them on an empiric basis. Half of these treatments may not be justified on the basis of negative microbiologic cultures. In ICUs with written protocols for empiric AT, treatments might be initiated more rapidly at the time of suspicion of infection and out-of-hours. These observations should encourage the establishment of local AT protocols to initiate AT without delay and to stop the abuse of AT. Since pulmonary infections are the most frequent type of infection and as septic shock and MOF are the most life-threatening infections, local guidelines should start by addressing these issues.