Antimicrobial stewardship in the intensive care setting – a review and critical appraisal of the literature

a Department of Medicine, McMaster University, Hamilton, ON, Canada b Department of Clinical Epidemiology and Biostatistics, McMaster University, Hamilton, ON, Canada c Department of Pathology and Molecular Medicine, McMaster University, Hamilton, ON, Canada d Michael G. DeGroote Institute for Infectious Diseases Research, McMaster University, Hamilton, ON, Canada e Hamilton Health Sciences, McMaster University, Hamilton, ON, Canada


Background
For over 50 years, experts have noted extensive inappropriate or unnecessary use of antimicrobials in the range of 30-50% [1][2][3][4][5][6].Inappropriate use of antimicrobials is typically defined as wrong indication, spectrum, route, dose or duration of therapy [1].The emergence of resistance as a consequence of antibiotic use [7,8] (see www.anresis.chfor current resistance patterns in Switzerland) is associated with poor clinical outcomes and increased healthcare costs [9,10].In fact, emergence of resistance continues to pose a challenge as demonstrated by the global spread of New Delhi β-lactamase producing Enterobacteriaceae resulting in strains that are resistant to virtually all currently available antibiotics [11].Furthermore, despite worldwide efforts to promote development of new antibiotics, the pipeline remains dry [12].In addition, antibiotic exposure is a trigger for Clostridium difficile infections and outbreaks [2,3,13].Antimicrobial stewardship programmes (ASPs) continue to be implemented in hospitals and communities around the world as quality improvement and patient safety programmes aimed at curbing inappropriate use of antibiotics with the ultimate goal of optimising patient outcomes while minimising its unintended consequences [1].Given the high density of antibiotic utilisation in the intensive care unit (ICU) setting and the importance of appropriate choice, particularly in septic patients [14], many programmes focus on this setting, assuming the greatest return on investment.A systematic review published in 2011 summarised the evidence on ASP in ICUs published up to 2010 and concluded that although more rigorous studies were required, available evidence did suggest antimicrobial stewardship was associated with improved utilisation of antimicrobials in the ICU setting, improvement in resistance and fewer adverse events [15].In this review, we summarise and assess the quality of evidence supporting the implementation of various stewardship strategies in the ICU setting with a focus on the time period from 2010-2015.

Methods
We searched Medline on April 1 2015 using the search terms "(critical care) AND stewardship" and "(intensive care) AND stewardship".We identified 133 and 193 art-icles, respectively, and also screened review and systematic review articles for additional studies.In this review we summarise 28 of the most relevant articles.We grouped the studies into those assessing the use of audit and feedback, formulary restrictions, procalcitonin, and those with mixed interventions or guidelines and clinical pathways.We adapted the Grading of Recommendations Assessment, Development and Evaluation (GRADE) [16] terminology to assess the quality of evidence for each of the four ASP strategies.A systematic assessment of the risk of bias of each included study and a full GRADE assessment of the overall level of evidence was beyond the scope of this nonsystematic review article.

Audit and feedback
Prospective audit with intervention and feedback is generally regarded as the gold standard stewardship intervention and received an A-1 grading in the Infectious Diseases Society of America (IDSA) guidelines [1].This was based on two randomised controlled trials [16,17] and two observational studies [18,19], none of which were conducted in the ICU setting.We identified nine relevant studies utilising audit and feedback in an ICU environment (table 1).The highest quality study was an interrupted time-series with a control group conducted at a large tertiary care centre in Ontario, Canada, by Elligsen et al., and targeted ICU patients on broad-spectrum antibiotics with audit and feedback on day 3 and 10 of therapy over a 1 year period [20].The ASP pharmacist reviewed patients, conferred with the infectious diseases physician and provided feedback to the team.They reported a reduction in broad-spectrum antibiotic use with no changes in overall use or in their two controls, resulting in a 23% reduction in antimicrobial costs ($95 000).Also, Gram-negative susceptibility to meropenem improved over time in the ICUs.There was no impact on ICU length of stay (LOS) and mortality.The design of this study had the lowest risk of bias; however, the control groups were not optimal (broad-spectrum antibiotics in medical and surgical wards, stress ulcer prophylaxis in the ICU) and it was a single-centre study.
A second time-series analysis was published by Wang et al. from Taiwan [21].They assessed the impact of a blood culture-guided de-escalation of empirical antimicrobials as an add-on to an existing stewardship programme in their 16 adult ICUs.They extended their pre-existing computerised antimicrobial approval system by flagging patients with a positive blood culture for a second review within 48 hours.Antibiotic utilisation, primarily carbapenems, gentamicin and those with activity against vancomycin-resistant Enterococcus, showed a significant reduction.This was reflected by a 19% decrease in antimicrobial costs.However, patient outcomes remained unchanged.Notably, this was an extension of a robust pre-existing ASP, thus, the findings might not be generalisable to a setting where an ASP has not been established.Finally, Bornard et al. published a time-series study from France [22].The intervention included thrice-weekly ASP rounds by an infectious diseases specialist combined with teaching sessions and daily contact with a microbiologist.They found a nonsignificant improvement in appropriate courses of antimicrobials during the intervention (80% vs 73% at baseline), with a higher rate of modification on days 2-4.This study was small (37 and 44 patients, respectively), and the lack of improvement might have been due to a ceiling effect that was attributed to a higher than expected baseline appropriateness of antibiotics.We also identified five uncontrolled quasi-experimental studies [23][24][25][26][27]. Amer et al. from Saudi Arabia targeted five antibiotics and compared the appropriateness of empirical treatment in patients receiving these antibiotics before and after programme implementation [23].Appropriateness of treatment increased from 30.6% to 100%.Prior to the ASP review, 20.9% of patients were treated appropriately, but the acceptance rate for ASP recommendations was 96.3%.The intervention resulted in a 68% reduction in the targeted antibiotics from 1 177.8 to 376.2 defined daily doses / 1000 patient days (DDD/1 000PD).This study was small (n = 24 and n = 49 in the two groups, respectively), and the intervention team was also assessing the appropriateness; thus, it was to be expected that the appropriateness postimplementation was high.[25].The proportion of cultures treated increased for sterile sites and the ratio decreased for nonsterile sites.They noted a significant improvement in appropriate daily documentation for antibiotic courses including the medication, dose, route and duration, as well as de-escalation, if applicable.They also reported a reduction in antimicrobial utilisation and acquisition costs.A strength of this study was the unique goal to improve documentation through education as a component of audit and feedback, and measurement of the proportion of cultures treated as a surrogate for appropriate antibiotic usage.Rimawi et al. reported a significant improvement in guideline-concordant treatment and a reduction in utilisation of extended-spectrum penicillins, carbapenems, vancomycin and metronidazole, and antibiotic costs, but an increase in narrow-spectrum penicillins in a medical ICU in the USA [27].A significant improvement in patient outcomes was reported: reduced mechanical ventilation days, LOS, and hospital mortality.However, the mortality difference was only observed for overall hospital mortality, and not for ICU mortality; therefore, these differences in reported outcomes may not be directly attributable to the ASP intervention.Diaz Granados targeted imipenem and piperacillin/tazobactam in a USA community hospital [24].The acceptance rates for recommendations were comparably low (66.5%);however, they found an improvement in appropriate antimicrobial selection and resistance rates.They also reported a shorter overall and ICU LOS as well as shorter courses of treatment with similar mortality rates.
Limitations included different patient populations in their study periods as well as their approach to quantifying resistance rates, which has neither been validated nor represents standard of practice for measurement.Finally, Leung et al. found a reduction in costs (36.2%) and antibiotic utilisation (38.9%) for antipseudomonal broad-spectrum antibiotics in a Canadian hospital [26].Their model included an ASP pharmacist and physician who conducted daily reviews of ICU patients.No significant differences in C. difficile infection rates or mortality rates were reported.
Finally, the study by Ijo et al. from the USA was the least rigorous from this group since the authors compared patient outcomes (LOS and mortality) during a 4-month ASP audit and feedback intervention with the published literature [28].They found a shorter LOS and similar mortality rates in the 70 patients included compared to the published literature.

Formulary restrictions
Restriction and preauthorisation of specific anti-infectives is recommended by the IDSA guideline as an A-II recommendation to improve drug utilisation and reduce costs.All studies cited in the guidelines were single centre, non-ICU focused, and all but one study targeted parenteral antibiotics.Four of the studies required a verbal approval from an infectious diseases physician prior to release of the antibiotic [29][30][31][32] and three studies utilised clinical pharmacists to enforce restriction criteria [33][34][35].We identified only three recent studies related to restriction and preauthorisation.Sistanizad et al. conducted a controlled quasi-experimental study in a teaching hospital in Iran [36].They restricted carbapenems for the treatment of documented multidrug-resistant organisms when no other therapeutic options were available in a medical ICU using the general ICU as the control group.Consequently, carbapenem use decreased significantly by 60% (from 68.6 to 27.5 DDD/1 000PD) while the overall antibiotic util-isation remained unchanged.Limitations of this study included potential for cross-contamination between the intervention and control unit since physicians were practicing in both settings and, subsequently, carbapenem utilisation also decreased in both units.Guarascio et al. developed an "antifungal bundle" intervention that was implemented in a medical and surgical ICU at a university hospital in the USA in order to limit excessive use of echinocandins, with a clinical pharmacist providing daily reassessment of treatment [37].Historic controls were used for comparison in a matched cohort study design.The reduction in the medical ICU was significant (median DOT 4.0 vs 2.0) with a potential for cost savings of $1 013 per patient.
Finally, Sharma and Barman implemented and assessed the impact of an antibiotic restriction form that outlined appropriate indications for reserved antibiotics in a tertiary care hospital in India [38].They showed a reduction in carbapenem use from 18.8 at the beginning to 10.6 DDD/100 bed days at the end of a 4-month intervention and the antifungal use decreased from 56.1 to 22.1 DDD/100 bed days.

Guidelines, clinical pathways and mixed interventions
A combination of strategies will likely maximise the success of an ASP within an institution.Education, clinical pathways, antimicrobial cycling and integration of healthcare information technology are elements discussed in the IDSA guidelines [1].Overall, we identified nine studies that evaluated guidelines, clinical pathways and/or mixed interventions in an ICU setting (table 2).

Evidenced-based guidelines and pathways
The IDSA guidelines highlighted that evidence-based guidelines developed by a multidisciplinary team incorporating local resistance patterns can improve antimicrobial utilisation as an A-III recommendation [1] based on two randomised controlled trials [39,40] and three quasi-experimental studies [41][42][43].
We identified two recent uncontrolled quasi-experimental studies [44,45] [44].Vancomycin start rates were significantly decreased from 6.9 and 17 to 4.5 and 6.4 per 1 000 patient-days, respectively.No statistical difference in the overall incidence of late-onset sepsis, meningitis, duration of bacteraemia and mortality were found.

Education
Education is considered to be an important element of any intervention and is graded as A-III evidence in the IDSA guidelines.However, education alone is deemed to be marginally effective in modifying prescribing behaviour [1].This recommendation was based on two quasi-experimental studies [46,47].
Meyer et al. provided teaching sessions on changing antibiotic prophylaxis to a single cefuroxime dose prior to central nervous system shunt insertion and found a significant reduction in cefuroxime use in a time-series analysis [48].
Unexpectedly, the proportion of third generation resistant Escherichia coli increased postintervention.Limitations included the absence of clinically meaningful data such as shunt-related infections, and the unexplained change in resistance rates.Chaves et al. provided quarterly education sessions, developed ICU-specific treatment guidelines and implemented a reminder on the medication record for proper documentation [49].Compared with the historical control, adherence to documentation standards significantly improved across all items: start date (72% to 90%), stop date (16% to 63%), indication (58% to 83%).Furthermore, antibiotic use concordance with national guidelines significantly improved from 74% to 89%.

Computerised decision support system
The recommendation of a computerised decision support system (CDSS) received a B-II grading in the IDSA guidelines [1].The majority of the studies discussed in the IDSA guidelines were single centre experiences in the USA [50][51][52][53].In these quasi-experimental studies, the authors demonstrated that CDSS significantly reduced the number of pharmacy interventions and adverse drug reactions, decreased antimicrobial consumption and associated drug costs.The only randomised controlled trial was from Shojania et al., in which national guidelines advocating appropriate vancomycin use were incorporated into the hospital's computerised physician order entry (CPOE) system [54], resulting in an overall reduction in vancomycin use and duration with no impact on patient-important outcomes reported.
More recently, Nachtigall et al. evaluated the impact of a CDSS that guided clinicians through a systematic, evidence-based approach in management of common infections in ICU, while addressing de-escalation strategies and targeted antibiotics following pathogen identification [55].Guideline adherence and antibiotic-free days increased significantly compared with the historical control.Finally, the authors also found a lower mortality in patients where CDSS was adhered to (8% vs 12.3%).Overall, this study demonstrated the sustained effects of a CDSS over a 5-year period; however, the generalisability may be limited owing to the predominantly surgical patient population.

Mixed interventions
In a mixed medical/surgical ICU in the USA, Slain et al. implemented prospective audit and feedback, formulary restriction and preauthorisation, education, pocket cards, and a ventilator-associated pneumonia antibiotic cycling protocol with streamlining and de-escalation in a step-wise fashion [58].Ciprofloxacin and ceftazidime use declined from 148 and 62.5 to 40 and 24.5 DDD/1 000PD, respectively, as did the overall use of antibiotics, but no effect on Pseudomonas aeruginosa resistance patterns was noted.Similarly, Dortch et al. implemented routine bronchoscopy in cases of suspected pneumonia, surgical prophylaxis protocols, treatment guidelines and quarterly antibiotic cycling [59].The proportion of healthcare-associated infections caused by multidrug resistant Gram-negative pathogens decreased from 37.4% to 8.5% over 8 years.Antibiotic cycling appeared to have a significant association with a reduction of multidrug resistant healthcare-associated infections, although this was not the primary endpoint.The potential benefits and risks of antibiotic cycling, which received a C-II grading due to insufficient data in the IDSA guidelines, are beyond the scope of this review [1].

Procalcitonin
Biomarkers, in particular procalcitonin, are emerging as potentially very useful laboratory tests to aid in the diagnosis of bacterial infections such as respiratory infections [57] and sepsis [55,56].Numerous studies have found procalcitonin to be useful in various settings, including the emergency department [60] and ICU (table 3).
In 2013, Prkno et al. conducted a meta-analysis to evaluate the use of procalcitonin in patients with severe sepsis [61].They compared procalcitonin-guided therapy with standard of care, with mortality as the primary outcome.They included seven randomised controlled trials, four of which will be described in more detail [62][63][64][65].Duration of antibiotics was reported in five studies and suggested a significant reduction in the median duration of antimicrobials in the procalcitonin guided group compared to the control group (hazard ratio 1.27, 95% confidence interval [CI] 1.01-1.53])while mortality and ICU LOS remained unchanged.The authors noted a moderate risk of bias across the studies.Although there was no statistical heterogeneity, there was clinical heterogeneity with the use of different procalcitonin algorithms, targeting escalation, de-escalation or a combination thereof, and differences in patient populations.Finally, duration of therapy in the control arms varied.
Hochreiter et al. conducted a single centre, open-label randomised controlled trial in a surgical ICU in Germany [64].They included patients with confirmed or suspected bacterial infection and two or more systemic inflammatory response syndrome criteria comparing a procalcitonin-guided arm with an 8-day standard course.The authors found a significant reduction in the duration of treatment (5.9 vs 7.9 days) and in the LOS in the ICU.The frequency of overriding the protocol was unclear as was the proportion of patients with infections requiring greater than 8 days of therapy.Of note, using an 8-day course as a comparator is questionable since shorter treatment courses can be used in the absence of procalcitonin [66,67].
Using a noninferiority open-label randomised controlled design in seven ICUs in France, Bouadma et al. assessed a procalcitonin-guided protocol to initiate or discontinue antimicrobials [63].The difference in mortality rates was within the 10% non-inferiority margin (21.2% vs 20.4%, risk difference 0.8 [95% CI −4.6 to 6.2] at 28 days).There was also a significant increase in antibiotic-free days from 11.6 to 14.3 days.There was no statistical difference in relapse, superinfection, number of days without mechanical ventilation, length of ICU or hospital stay.However, physicians overruled the procalcitonin algorithm 53% of the time.Furthermore, a 10% noninferiority margin for a mortality outcome is wide and the difference in mortality rates would not have met the noninferiority requirements with a smaller margin.
In another multicentre randomised controlled trial in nine ICUs in France reported by Annane et al., healthcare providers and investigators in the control arm were blinded to procalcitonin levels [62].They included patients who presented with severe sepsis, no clear infection source and negative microbial cultures.The study was stopped early because of low patient enrolment; therefore, the two groups were not balanced and the study was underpowered.They did not find a significant difference between the procalcitonin arm and the control group in the proportion of patients receiving antibiotics on day 5. Furthermore, there was a high rate of protocol deviation, with up to 37% of patients being managed outside of the algorithm recommendations at day 5.
In Denmark, Jensen et al. conducted another multicentre, open-label, randomised controlled trial across nine ICUs [65].In this escalation study, a procalcitonin level >1 mcg/l suggested expanding the antimicrobial coverage and searching for uncontrolled sources of infection.The primary outcome, 28-day survival, was similar.However, days of mechanical ventilation and median ICU LOS were significantly elevated in the procalcitonin arm.Antibiotic utilisation, particularly piperacillin-tazobactam and ciprofloxacin was increased in the procalcitonin arm.This may be explained by the large proportion of surgical patients in this study (41%), as surgical patients have elevated procalcitonin levels for other reasons, which may have prompted clinicians to continue antimicrobial therapy unnecessarily [68].
A more recent randomised controlled trial was published by Shehbabi et al. [69].They evaluated the effect of procalcitonin on antimicrobial prescribing across 11 Australian ICUs in addition to twice-weekly antimicrobial stewardship rounds.90-day all-cause mortality.Of note, the cut-offs used for the procalcitonin algorithm were lower than those that have been applied in other randomised controlled trials, potentially leading to no difference in duration of therapy.Finally, in a German surgical ICU, Hohn et al. conducted a quasi-experimental study evaluating the implementation of a procalcitonin protocol with ASP (antibiotic prescription restricted to senior ICU physician, daily ASP rounds, regular teaching, and local guidelines) [70].There was a reduction in antimicrobial utilisation of 15.4% between 2010-2011 and 2011-2012, driven by a reduction of aminoglycosides, cephalosporins and fluoroquinolones, resulting in a reduction in antimicrobial costs by 43%.
There was no difference in ICU LOS, and mortality rates.
Owing to the bundle approach, it is unclear whether the effect is attributable to ASP, or to the use of procalcitonin.
A more extensive review of the literature for procalcitonin in the ICU setting was beyond the scope of this article, and we refer to the recently published review article by Albrich et al. [71].

Discussion
The evidence for ASPs in the ICU setting is based on quasiexperimental studies with or without times-series analysis and/or control groups and -with the exception of studies on procalcitonin -no randomised controlled trials were identified.While individual-patient randomised controlled trials would not be feasible for the majority of interventions given the hospital-wide implementation of ASP and the risk of cross-contamination, cluster randomised controlled trials with or without a cross-over design would be feasible.The lack of randomisation increases vulnerability to both measured and unmeasured confounders.Also, in the absence of rigorous time-series analyses with an appropriate control group, these studies are at risk of bias as a result of temporal trends [72].While the impact of ASP intervention on appropriateness of antibiotics, utilisation and costs is fairly consistent across the studies, there is currently no convincing evidence that there is an effect on patient-important outcomes or resistance rates.One unpublished and not yet fully analysed study on audit and feedback from 11 academic hospitals with 14 participating ICUs in Ontario, Canada was recently conducted using a quasi-experimental stepped wedge design.Although again a nonrandomised study, the sample size of this study may allow for the detection of changes in patient-important outcomes as well as an impact on the antimicrobial susceptibility patterns.Prospective audit and feedback received an A-I grading in the IDSA guideline based on two non-ICU randomised controlled trials [1].We identified several rigorously conducted quasi-experimental studies that used either timeseries and/or a control group; thus, there is good evidence suggesting that audit and feedback has a positive impact on antibiotic utilisation in the ICU setting.However, given the lack of evidence from randomised controlled trials, our confidence in this finding is low when using GRADE terminology [73].Formulary restrictions and preauthorisation is listed as an A-II recommendation in the IDSA guidelines [1].Despite the high grading of the evidence in the guideline, we found a small number of recent studies on this topic.This might be related to the common concern of a "squeezing the balloon" phenomenon for this type of approach [74]: restriction of certain classes of antibiotics may result in a reduction in use and resistance rates, but without additional measures in place, it may also result in a shift to a higher usage of other antibiotics [32,33,35], thus negatively affecting the resistance rates for those alternative antibiotics.
The previous evidence summarised in the IDSA guidelines [1] suggesting that restriction and preauthorisation impact antibiotic utilisation was corroborated by the more recent studies, but downstream effects due to increased use of other antibiotics was not assessed in these studies.Similar to other strategies, the lack of randomised controlled trials results in a low level of evidence for these strategies.The use of education, evidence-based guidelines and pathways, CDSS and mixed interventions are all endorsed by IDSA [1].Our review of the literature found that the majority were uncontrolled quasi-experimental studies.Existing evidence appears to suggest that any of these interventions, alone or in combination, have a positive impact on process measures such as adherence rate or antimicrobial utilisation.Results on clinically relevant outcomes such as infection rates or resistant patterns were variable.The absence of randomised controlled trials and inclusion of multiple components in the intervention are two recurring challenges when reviewing this group of literature.Given these limitations, we can assign only a low grading.The use of procalcitonin in the ICU setting has been well studied in numerous randomised controlled trials and the findings are promising in terms of reduction of antibiotic use without negatively affecting patient outcomes.While the majority of studies favoured the procalcitonin arm, one large procalcitonin-escalation study found a negative impact on antibiotic utilisation as well as on days on mechanical ventilation and LOS [65].Several limitations were identified: there was significant heterogeneity in protocols, with some studies using procalcitonin to initiate, others to discontinue antibiotics or both.In addition, there is variability in the procalcitonin cut-offs that were used.The populations studied are heterogeneous, as some studies evaluated medical/surgical patients whereas others looked predominately at surgical patients.Only two studies from our review explicitly stated there were ASPs present in the ICUs [69,70]; thus, the added value of procalcitonin within a strong ASP remains unclear.Furthermore, there was significant deviation from the procalcitonin-protocol noted [62,63].Finally, costs remain a concern given the comparably high costs for testing, although a recent cost-minimisation and cost-utility study suggested improved patient quality of life with reduced costs in ICU patients with bacterial infections and sepsis [75].Overall, there is a moderate level of evidence supporting the use of procalcitonin to reduce duration of antibiotics in the ICU setting.
While we found overall convincing data -within the limitations of nonrandomized studies-that most stewardship strategies are potentially effective in reducing and optimizing antibiotic utilisation, the impact on patient-important outcomes has been variable with no convincing evidence at this point.The absence of such evidence does not imply ASP strategies in the ICU setting are not beneficial to patient-important outcomes, but their detection is difficult given the small event rates of failures, and the lag time until effects may affect future patients, for example, due to reduction of the emergence of resistance.Future studies should be large, i.e. multi-centre studies, optimally using a cluster randomised control trial design or at least controlled quasi-experimental studies with time-series analysis over a prolonged period of time in order to be able to identify the impact of ASP on patient-important outcomes and on the emergence of resistance.

Table 1 :
Studies on prospective audit and feedback.
Abx = antibiotics; ASP = antimicrobial stewardship programme; CDI = C. difficile infection; HAI = hospital-acquired infections; ICU = intensive care unit; ID = infectious diseases; LOS = length of stay Review article: Medical intelligence Swiss Med Wkly.2015;145:w14220 Swiss Medical Weekly • PDF of the online version • www.smw.ch within 24 hours of diagnosis of ventilator-associated pneumonia with mandatory use of CCP which was significantly better than voluntary use (44% and 56.3%, respectively).In addition, mandatory CCP shortened time to appropriate therapy and facilitated de-escalation in a greater proportion of patients.
Finally, Wilde et al implemented a computerised clinical pathway (CCP) to facilitate antimicrobial selection for ventilator-associated pneumonia treatment in the three ICUs of their tertiary care institution in USA [57].They achieved 100% compliance and 70.8% of patients received Review article: Medical intelligence Swiss Med Wkly.2015;145:w14220 Swiss Medical Weekly • PDF of the online version • www.smw.chappropriate antibiotics

Table 2 :
Education, evidence-based guidelines and pathways, mixed interventions.
aureus; ICU = intensive care unit; SSTI = skin and soft tissue infection; VAP = ventilator-associated pneumonia Review article: Medical intelligence Swiss Medical Weekly • PDF of the online version • www.smw.ch

Table 3 :
Procalcitonin studies in the ICU setting.
abx = antibiotic; DDD = defined daily dose; ICU = intensive care unit; LOS = length of stay; PCT = procalcitonin, DDD = defined daily dose; RCT = randomised controlled trial Review article: Medical intelligence Swiss Med Wkly.2015;145:w14220 Swiss Medical Weekly • PDF of the online version • www.smw.ch Disclosure statement: DM is a recipient of a Research Early Career Award from Hamilton Health Sciences Foundation (Jack Hirsh Fellowship).