Electronic Alerts with Automated Consultations Promote Appropriate Antimicrobial Prescriptions
Electronic Alerts with Automated Consultations Promote Appropriate Antimicrobial Prescriptions
Moonsuk Kim 0 1
Kyoung-Ho Song 0 1
Chung-Jong Kim 0 1
Minkyo Song 1
Pyoeng Gyun Choe 1
Wan Beom Park 1
Ji Hwan Bang 1
Hee Hwang 1
Eu Suk Kim 0 1
Sang- Won Park 1
Nam Joong Kim 1
Myoung-don Oh 1
Hong Bin Kim 0 1
0 Department of Internal Medicine, Seoul National University Bundang Hospital , Seongnam , Republic of Korea, 2 Department of Internal Medicine, Seoul National University College of Medicine, Seoul, Republic of Korea, 3 Department of Preventive Medicine, Seoul National University College of Medicine, Seoul, Republic of Korea, 4 Center of Medical Informatics, Seoul National University Bundang Hospital , Seongnam , Republic of Korea
1 Editor: Giuseppe Vittorio De Socio, Azienda Ospedaliera Universitaria di Perugia , ITALY
Data Availability Statement: All relevant data are
within the paper and its Supporting Information files.
Funding: The authors have no support or funding to
Competing Interests: The authors have declared
that no competing interests exist.
To promote appropriate antimicrobial use in bloodstream infections (BSIs), we initiated an intervention program consisting of electronic alerts and automated infectious diseases consultations in which the identification and antimicrobial susceptibility test (ID/AST) results were reported.
We compared the appropriateness of antimicrobial prescriptions and clinical outcomes in
BSIs before and after initiation of the program. Appropriateness was assessed in terms of
effective therapy, optimal therapy, de-escalation therapy, and intravenous to oral switch
There were 648 BSI episodes in the pre-program period and 678 in the program period. The
proportion of effective, optimal, and de-escalation therapies assessed 24 hours after the
reporting of the ID/AST results increased from 87.8% (95% confidence interval [CI] 85.5–
90.5), 64.4% (95% CI 60.8–68.1), and 10.0% (95% CI 7.5–12.6) in the pre-program period,
respectively, to 94.4% (95% CI 92.7–96.1), 81.4% (95% CI 78.4–84.3), and 18.6% (95% CI
15.3–21.9) in the program period, respectively. Kaplan-Meier analyses and log-rank tests
revealed that the time to effective (p<0.001), optimal (p<0.001), and de-escalation (p =
0.017) therapies were significantly different in the two periods. Segmented linear regression
analysis showed the increase in the proportion of effective (p = 0.015), optimal (p<0.001),
and de-escalation (p = 0.010) therapies at 24 hours after reporting, immediately after
program initiation. No significant baseline trends or changes in trends were identified. There
were no significant differences in time to intravenous to oral switch therapy, length of stay,
and 30-day mortality rate.
This novel form of stewardship program based on intervention by infectious disease specialists and information technology improved antimicrobial prescriptions in BSIs.
In the treatment of bloodstream infections (BSIs), administration of appropriate antimicrobials
without delay is crucial for improving clinical outcomes [
]. However, in previous studies,
23% to 46% of empirical antimicrobials for BSIs were considered inappropriate [
after the identification and antimicrobial susceptibility test (ID/AST) results were reported, 8%
to 19% of the antimicrobials prescribed remained inappropriate [
There have been several trials of interventions to improve the appropriateness of antimicro
bial therapy in BSIs, some of which led to improvements in antimicrobial prescribing and
clinical outcomes [
]. However, many of these interventions were highly labor-intensive. In
planning sustainable interventions, the available resources, circumstances of individual
hospitals, and strategies to increase efficiency should be considered.
To promote appropriate antimicrobial use in BSIs, we initiated an intervention program in
August 2011 which consisted of electronic alerts and automated infectious diseases consultations, and evaluated the impact of the program on the appropriateness of antimicrobial prescriptions.
Materials and Methods
Hospital setting and program design
Our institution is a tertiary teaching hospital with 900 beds. We planned an intervention pro
gram taking advantage of our health information system for electronic medical records (EMR)
in our hospital. The key feature of this program is a pop-up message, which gives the ID/AST
results with a note about whether a consultation with an infectious disease specialist (IDS) is
needed. Attending physicians can issue the consultation according to the automated process.
The only action needed to make a consultation is to click on the message. However, the consul
tation is not mandatory and can be canceled (Fig 1). Pop-up messages are generated for all
categories of ID/AST results. Before the initiation of this intervention program, no additional
process of reporting was available, and attending physicians had to open the test results section
of the EMR to view the results. This program was initiated on August 19, 2011.
We designated the period before the initiation of the program as the “pre-program period”,
and the period of the program as the “program period”. Excluding the month of initiation of
the program (August 2011), the total duration of the study was 2 years, comprising a 1-year
pre-program period (from August 2010 to July 2011) and a 1-year program period (from
September 2011 to August 2012).
2 / 14
Fig 1. Flow diagram of electronic alerts and automated consultations.
During the study period, all episodes of BSI in adult ( 18 years) patients were reviewed. A BSI
episode was defined as growth of pathogenic bacteria or fungi from one or more sets of blood
culture samples in a patient with systemic inflammatory response syndrome. When present,
common skin commensals, such as Corynebacterium species or coagulase-negative
staphylococci were considered pathogens only when two or more sets of blood cultures were positive
within 48 hours, and when clinically related infection existed.
BSI episodes were included in the analysis only when blood cultures were performed for
hospitalized patients or those in the emergency department. Because the program started to
operate after the reporting of the ID/AST results, the episodes involving patients discharged
from the hospital before the reporting were excluded. We considered that the impact of the
program on antimicrobial prescription and clinical outcomes could not be assessed
appropriately in cases involving patients who died within 24 hours after the reporting and patients with
polymicrobial infection, defined as the isolation of two or more pathogens from the same
blood culture set. Furthermore, the appropriateness of the use of antimicrobial agents for
polymicrobial infections was not estimated solely on the basis of the ID/AST results. Therefore, the
episodes that involved early death and polymicrobial infections were excluded from the study.
3 / 14
BSI onset was defined as the time when the initial blood culture was ordered. If there was a
new positive culture two weeks after BSI onset, that episode was considered a separate one.
However, if an obviously different infection developed, it was counted as a separate episode even when it developed within two weeks of the initial episode.
Primary outcomes were the time to appropriate antimicrobial administration and the fre
quency of appropriate antimicrobial administration after the ID/AST results were reported.
The appropriateness of antimicrobial use was assessed in terms of effective therapy, optimal
therapy, de-escalation therapy, and intravenous to oral switch. Secondary outcomes were
length of stay, and all-cause and infection-related mortalities 30 days from BSI onset.
Infection-related mortality was defined according to the following criteria, adapted from Harbarth
], (1) positive blood culture at the time of death, (2) a persistent focus of BSI with associated
clinical signs of sepsis, and (3) death within two weeks of the documentation of BSI without an
Definitions of variables
1. Reference time: the time when the ID/AST results were reported (S1 Fig).
2. Type of acquisition: (1) community-onset BSI: BSI that occurred <48 hours after admission
(or emergency department visit, if the patient was admitted via the emergency department);
(2) healthcare-associated BSI: BSI that occurred <48 hours after admission with risk of
healthcare exposure [
]; (3) hospital-onset BSI: BSI that occurred 48 hours after
3. Empirical therapy: antimicrobial in use at the reference time. Antimicrobials were classified
into ineffective, effective and optimal, and effective but non-optimal. Superfluous
broadspectrum antimicrobial therapy was considered effective but non-optimal.
4. Effective therapy: Effective therapy was defined as the administration of an active
antimicrobial for a confirmed pathogen for each ID/AST result. Time to effective therapy was defined
as the interval from the reference time to effective therapy.
5. Superfluous broad-spectrum antimicrobials: if the antimicrobials of interest were used as
empirical therapy and the ID/AST results revealed that their spectra were excessive, we
classified them as superfluous. For gram-negative BSI, the antimicrobials of interest were
piperacillin/tazobactam, ceftazidime, cefepime, carbapenem, and polymyxin E. Glycopeptide and
oxazolidinone were the antimicrobials of interest for gram-positive BSI (daptomycin and
anti-MRSA cephalosporins were not available during the study period).
6. Optimal therapy: Optimal therapy was defined as the administration of an effective
antimicrobial which was not one of the superfluous broad-spectrum antimicrobials and adequate
in relation to the site of infection, route of administration, and dosing. Likewise, time to
optimal therapy was defined as the interval from the reference time to optimal therapy.
7. De-escalation therapy: De-escalation therapy was defined as the discontinuation of use of a
broad-spectrum antimicrobial and the administration of an antimicrobial with a narrower
spectrum. We assumed the time of discontinuation of an antimicrobial to be the time of the
next scheduled administration of the antimicrobial. The antimicrobial with a narrower
spectrum was not necessarily the one with the narrowest spectrum of those available.
4 / 14
8. Intravenous to oral switch: Only the cases involving the switch to antimicrobials that were
active in vitro were included in the analysis. We considered the switch as appropriate in cases
of (1) absence of clinical indication for intravenous therapy, such as CNS infection and
endocarditis, (2) normalization of body temperature, (3) decrease in the level of inflammatory
markers, (4) improvement in the patient’s symptoms and signs, and (5) normal
gastrointestinal absorption (records of normal oral intake without vomiting and diarrhea) [
To compare general characteristics between the two periods, the chi-square test or Fisher's
exact test was used for categorical variables, and Student’s t-test or the Mann-Whitney median
test for continuous variables. Kaplan-Meier analysis and log-rank tests were performed to
compare the time to effective therapy, optimal therapy, de-escalation therapy, and intravenous to
oral switch. All data were censored at 10 days (240 hours) from the reference time. We
considered the data as censored when a patient died within 10 days after reporting, or if the patient
was discharged from the hospital within 10 days and was not monitored after this period.
Segmented linear regression analyses were performed to evaluate the trends of outcomes
over time before and after the initiation of the intervention program [
]. Analyses were
conducted using monthly data of the proportion of patients on appropriate therapy at 24 and 48
hours after the reporting of the ID/AST results. The Durbin-Watson statistic was used to detect
the presence of autocorrelation.
All analyses were performed using SAS software version 9.4 (SAS Institute Inc.). All tests were two-tailed and adopted an alpha level of 0.05.
The study was approved by the Institutional Review Board of Seoul National University Bundang Hospital (IRB No. B-1412/278-113). The study was conducted with a waiver of informed consent from individual patients. Patient data were de-identified and anonymized before analysis.
A total of 879 episodes in the pre-program period and 899 episodes in the program period
were identified as BSIs of clinical significance. After that, 231 episodes in the pre-program
period and 221 episodes in the program period were excluded according to the exclusion
criteria. Therefore, 648 episodes in the pre-program period and 678 in the program period were
included in the analysis (S2 Fig). The corresponding number of patient was 566 and 591,
The time from BSI onset to the reporting of the ID/AST results was shortened in the pro
gram period by 9.0 hours (95% confidence interval [CI] 5.0–12.9 hours, p<0.001). The
automated consultations were canceled by the attending physicians in 22.0% of all episodes in the
program period (149 of 678). In the remaining 529 episodes, in which the consultations were
not canceled, the mean interval from automated consultation forwarding to signed reply was
6.8 hours (95% CI 5.8–7.8).
Clinical characteristics of bloodstream infection episodes
The clinical characteristics of the episodes in the two periods are compared in Table 1. There were no significant differences in median age, median age-adjusted Charlson comorbidity index , sex distribution, comorbidity, service department in charge, type of acquisition,
5 / 14
(n = 648)
primary site of infection, and classification of empirical therapy (S3 Fig). However, the
proportion of ICU episodes was larger in the program period (7.1% versus 11.7%, p = 0.005).
There were also no significant differences in the distribution of pathogens between the two
periods except for cases of coagulase-negative staphylococci (2.5% versus 5.0%, p = 0.015) and
Escherichia coli (37.8% versus 31.7%, p = 0.020). The proportion of extended-spectrum
betalactamase (ESBL) producers among the isolates of E. coli was larger in the program period
(32.2% versus 44.2%, p = 0.008). The pathogen distribution is shown in Table 2.
Fig 2. Kaplan-Meier curve of the time to effective therapy. *Identification and Antimicrobial susceptibility test.
Time to effective therapy
We compared the time to effective therapy between the two periods in two steps according to
the target group using Kaplan-Meier methods as shown in Fig 2. First, the analysis was
performed for all episodes. The proportion of effective therapy at 24 and 48 hours after the
reporting of the ID/AST results increased from 87.8% (95% CI 85.5–90.5) and 91.8% (95% CI 89.7–
93.9) in the pre-program period to 94.4% (95% CI 92.7–96.1) and 97.4% (95% CI 96.1–98.6) in
the program period. Subsequently, the analysis was performed only for the episodes for which
empirical treatment was classified as ineffective. The proportion of effective therapy at 24 and
48 hours increased from 38.8% (95% CI 30.4–47.2) and 58.7% (95% CI 50.1–67.2) to 74.8%
(95% CI 67.9–81.7) and 88.1% (95% CI 82.9–93.3). The median time to effective therapy was
shortened in the program period by 30.1 hours (from 40.7 hours [95% CI 24.8–48.0] to 10.6
hours [95% CI 7.7–16.3], p<0.001). The results of the log-rank tests confirmed the presence of
significant differences between the two periods in both steps (p<0.001).
Time to optimal therapy
The proportions of optimal therapy at 24 and 48 hours increased after the reporting of the ID/
AST results from 64.4% (95% CI 60.8–68.1) and 74.3% (95% CI 70.9–77.6) in the pre-program
period to 81.4% (95% CI 78.4–84.3) and 91.4% (95% CI 89.3–93.5), respectively, in the
program period for all episodes (Fig 3). For non-optimal episodes, the proportion of cases that
received optimal therapy at 24 and 48 hours increased from 31.0% (95% CI 26.0–36.0) and
50.2% (95% CI 44.8–55.6) to 63.0% (95% CI 57.8–68.1) and 83.0% (95% CI 78.9–87.0),
respectively. The median time to optimal therapy was shortened in the program period by 27.5 hours
(from 46.6 hours [95% CI 40.8–55.2] to 19.1 hours [95% CI 15.4–20.6], p<0.001) (Fig 3). The
results of the log-rank test revealed the presence of significant differences between the two
periods in both steps (p<0.001).
Time to de-escalation therapy
Among the effective episodes, the proportions of de-escalation therapy at 24 and 48 hours after
the reporting of the ID/AST results increased from 10.0% (95% CI 7.5–12.6) and 17.0% (95%
8 / 14
Fig 3. Kaplan-Meier curve of the time to optimal therapy. *Identification and Antimicrobial susceptibility test.
CI 13.8–20.3) in the pre-program period to 18.6% (95% CI 15.3–21.9) and 26.6% (95% CI
22.8–30.4) in the program period, respectively (Fig 4). Among the superfluous broad-spectrum
episodes, the proportion of de-escalation therapy at 24 and 48 hours increased from 28.2%
(95% CI 21.6–34.7) and 47.8% (95% CI 40.5–55.1) to 57.1% (95% CI 49.6–64.5) and 81.5%
(95% CI 75.6–87.3), respectively. The median time to de-escalation therapy was shortened in
the program period by 27.9 hours (from 48.3 hours [95% CI 42.0–65.9] to 21.6 hours [95% CI
18.8–24.2], p<0.001) (Fig 4). The results of the log-rank test revealed the presence of
significant differences between the two periods in both steps (p = 0.017 and <0.001, respectively).
Time to intravenous to oral switch
The proportion of switch therapies performed on days 2, 4, and 8 increased from 8.5% (95% CI
6.4–10.7), 14.1% (95% CI 11.4–16.8), and 22.2% (95% CI 18.9–25.4), respectively, in the
preFig 4. Kaplan-Meier curve of the time to de-escalation therapy *Identification and Antimicrobial susceptibility test.
9 / 14
program period to 9.1% (95% CI 7.0–11.4), 16.1% (95% CI 13.3–18.9), and 24.4% (95% CI
21.1–27.7), respectively, in the program period. However, Kaplan-Meier analysis and the
logrank test showed that there was no statistically significant difference between the times to
intravenous to oral switch in the two periods (p = 0.447).
Results of segmented linear regression analysis
The results of segmented linear regression analyses are shown in Table 3. The monthly propor
tion of patients on effective, optimal, and de-escalation therapies at 24 and 48 hours from the
reference time increased immediately after the implementation of the program. There were no
significant baseline trends and no significant changes in trends after the implementation of the
program. The Durbin-Watson statistic revealed no significant autocorrelation.
Length of stay and mortality rate
The number of episodes in which the outcome of the admission treatment was death was 85 in
the pre-program period and 89 in the program period. These cases were excluded from the
estimation of length of stay. There was no significant difference in the median length of stay
between the two periods (16.6 [interquartile range (IQR) 8.8–29.8] versus 16.0 [IQR 9.5–28.9],
p = 0.715). There was also no significant difference in all-cause 30-day mortality rate between
the two periods (68/648 versus 67/678, p = 0.713). The infection-related 30-day mortality rate
decreased (39/648 versus 27/678), but there was no statistical significance (p = 0.089).
Subjects in the program period whose consultations were canceled
When we compared the 22.0% (149/678) whose automatically generated consultations were
canceled to the remaining 78.0% (529/678), log-rank tests revealed there were no significant
differences in time to effective, optimal, de-escalation, and intravenous to oral switch therapies
(S1 Table). Length of stay (16.4 [IQR 8.9–35.3] versus 16.0 [IQR 9.7–27.2], p = 0.500) and
30-day mortality rate (54/475 versus 13/136, p = 0.592) were not significantly different between
the two groups.
However, the proportion of subjects on effective (87.2% [130/149] versus 75.0% [397/529],
p = 0.002) and optimal (60.4% [90/149] versus 46.7% [247/529], p = 0.003) antimicrobials at
the reference time was higher in the canceled subgroup. The proportion of subjects on
superfluous broad-spectrum antimicrobials at the reference time was not significantly different
(24.8% [37/149] versus 25.1% [133/529], p = 0.939). The proportion of cases in which
infectious diseases consultations were issued between 24 hours before the BSI onset and the
10 / 14
reference time was higher in the canceled subgroup (51.0% [76/149] versus 45.0% [238/529])
without statistical significance (p = 0.193).
This study showed that the intervention program in our institution promoted appropriate
antimicrobial prescription in the treatment of BSI. Kaplan-Meier analyses and log-rank tests
revealed that the probabilities of receiving effective, optimal, and de-escalation therapies were
higher in the program period. Furthermore, time to effective, optimal, and de-escalation
therapies was shortened in the program period. Segmented linear regression analysis revealed that
the proportion of effective, optimal, and de-escalation therapies increased after the
implementation of the program. No significant trends or changes in trends were identified before or after
the implementation of the program. We did not detect improvements in time to intravenous to
oral switch, reduction in length of stay and 30-day mortality rate.
The program consisted of alerts and automated consultations. Alerts coupled with antimi
crobial stewardship intervention [
], and routine consultations [
] have been proposed as
effective forms of antimicrobial stewardship program (ASP) for limited subjects. Our study
included a complete set of BSIs and involved a novel form of stewardship program which could
improve the appropriateness of antimicrobial prescription.
Intervention by IDS is a key element in optimizing antimicrobial therapy, and previous
studies have shown that IDS services and consultation improved antimicrobial prescription
and clinical outcomes [
]. Our program was also based on interventions conducted by
IDS. However, automatically generated consultations were canceled in 22.0% of the subjects in
the program period. We could not evaluate the reasons for cancelation and it was impossible to
clearly separate the effects of electronic alerts and automated consultations. The cancelation
could be related to the consultation to IDS before the reporting of the ID/AST results.
However, there was no statistically significant difference in the proportion of consultation to IDS
between the canceled 22% and the remaining 78%.
Another remarkable feature of our program was the use of information technology
resources. Interventions based on electronic health records (EHRs) and clinical decision
support systems (CDSSs) have been predicted to increase the efficiency of ASP [
]. In our
study, the electronic alerts may have had a synergistic effect in improving antimicrobial
prescription, and the use of electronic medical records may have enabled faster and more accurate
communication. The mean time to signed replies in the automated consultations of only 6.8
hours (95% CI, 5.8 to 7.8) achieved here might not have been possible by human effort alone.
There has not been any unified definition of appropriate antimicrobial use [1,8]. Other
terms such as “adequate use” or “optimal use” have been employed with similar meaning.
Usually these words were applied only to activity in vitro. Interventions to improve
antimicrobial use should not just aim for early initiation of active antimicrobial therapy. By
developing comprehensive definitions, we aimed to simultaneously initiate active antimicrobial
therapy and narrow the coverage. We have shown that our program achieved both aims.
Furthermore, the impact of our program in avoiding unnecessary antimicrobial use might be
underestimated because we did not include the cases regarded as contaminants in the
However, our approach had no significant impact on the intravenous to oral switch. We
suspected that this was due to the relatively low cost of hospitalization and the lack of economic
incentives to reduce the duration of hospitalization in our health care system. This reasoning
could also be applied to the issue of length of stay. With respect to the extended average length
of stay in hospitals, the Republic of Korea had ranked the second among OECD member
11 / 14
countries recently [
]. Since a considerable proportion of the total episodes (40.4%) were
categorized as hospital-onset, the program’s impact on length of stay could not be substantial.
In the program period, the interval from the BSI onset to the reporting of the ID/AST
results decreased. We assumed that this finding was the result of placing dedicated staffs in
our laboratory since June 2011. Although the decrease in the reporting time was not directly
related to the program, it might influence the overall time to appropriate antimicrobial
therapy. Despite this subsidiary change and the promotion of appropriate antimicrobial
prescriptions, there was no significant improvement in the 30-day mortality rate after the initiation
of the program. The possibility of improvement in the mortality rate might be small because
approximately 80% of the study subjects were already on effective antimicrobials at the
reference time. In addition, the impact of our intervention on the mortality rate might be limited
owing to the considerable portion of older individuals with a high comorbidity index. First, it
is possible to infer that the critical time point may be some time before the reference time of
our study [
]. Nonetheless, considering that our program promoted the administration of
effective antimicrobials and de-escalation therapy concurrently, both the absence of an
increase in all-cause mortality rate and the tendency of improvement in the infection-related
mortality rate can be considered meaningful. Mandatory consultations, as in the cases of
Staphylococcus aureus BSI [16,17], may enhance the impact of our program on clinical outcomes.
Despite its limitations, the combination of electronic alerts and automated consultations
aimed at BSI treatment was found to improve antimicrobial prescription in the current study.
Our approach offers an efficient form of stewardship intervention and demonstrates the
possibility of exploiting informatics. However, additional efforts to identify more effective and
efficient forms of stewardship program using information technology are needed. Also studies to
establish the most appropriate time for intervention in terms of efficiency and practicality are
warranted. The point in time when the results of the Gram stain are reported may represent
the appropriate time for an intervention. However, an earlier intervention will require more
resources and intensive work. The introduction of novel diagnostic tools such as
matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS) and S.
aureus PCR [
] offer further opportunities for stewardship programs for BSI.
S1 Fig. Schematic diagram of the time course of the program. aEmergency department.
bIdentification and Antimicrobial susceptibility test.
S2 Fig. A flow chart of describing selection of the subjects included in the analysis.
S3 Fig. Classification of empirical therapy.
S1 Table. Log-rank p-values according to the subgroups.
We express our gratitude to Center of Medical Informatics in Seoul National University Bundang Hospital for the cooperation in the development and maintenance of the program.
12 / 14
The results of this study were presented as a slide presentation at the 55th Interscience
Conference on Antimicrobial Agents and Chemotherapy (ICAAC 2015) in San Diego, California,
U.S.A, September 17–21, 2015.
Conceptualization: MK KHS CJK MS ESK HBK.
Formal analysis: MK KHS.
Investigation: MK KHS CJK ESK HBK.
Methodology: MK KHS CJK MS ESK HBK.
Project administration: MK KHS ESK HBK.
Software: KHS HH ESK HBK.
Supervision: PGC WBP JHB SWP NJK MDO HBK.
Validation: MK KHS CJK ESK HBK.
Visualization: MK KHS MS.
Writing - original draft: MK KHS.
Writing - review & editing: MK KHS CJK MS PGC WBP JHB HH ESK SWP NJK MDO HBK.
13 / 14
1. McGregor JC , Rich SE , Harris AD , Perencevich EN , Osih R , Lodise TP , et al. A systematic review of the methods used to assess the association between appropriate antibiotic therapy and mortality in bacteremic patients . Clin Infect Dis . 2007 ; 45 : 329 - 37 . doi: 10 .1086/519283 PMID: 17599310
2. Paul M , Shani V , Muchtar E , Kariv G , Robenshtok E , Leibovici L . Systematic review and meta-analysis of the efficacy of appropriate empiric antibiotic therapy for sepsis . Antimicrob Agents Chemother . 2010 ; 54 : 4851 - 63 . doi: 10 .1128/AAC.00627-10 PMID: 20733044
3. Retamar P , Portillo MM , López-Prieto MD , Rodríguez-López F , de Cueto M , García M V , et al. Impact of inadequate empirical therapy on the mortality of patients with bloodstream infections: a propensity score-based analysis . Antimicrob Agents Chemother . 2012 ; 56 : 472 - 8 . doi: 10 .1128/AAC.00462-11 PMID: 22005999
4. Harbarth S , Garbino J , Pugin J , Romand JA , Lew D , Pittet D. Inappropriate initial antimicrobial therapy and its effect on survival in a clinical trial of immunomodulating therapy for severe sepsis . Am J Med . 2003 ; 115 : 529 - 535 . doi: 10 .1016/j.amjmed. 2003 . 07 .005 PMID: 14599631
5. Arbo M , Snydman D. Influence of blood culture results on antibiotic choice in the treatment of bacteremia . Arch Intern Med . 1994 ; 154 : 2641 - 5 . Available: http://www.ncbi.nlm.nih.gov/pubmed/7993147 PMID: 7993147
6. Byl B , Clevenbergh P , Jacobs F , Struelens MJ , Zech F , Kentos A , et al. Impact of infectious diseases specialists and microbiological data on the appropriateness of antimicrobial therapy for bacteremia . Clin Infect Dis . 1999 ; 29 : 60 - 6 ; discussion 67- 8 . doi: 10 .1086/520182 PMID: 10433566
7. Rintala E , Kairisto V , Eerola E , Nikoskelainen J , Lehtonen OP . Antimicrobial therapy of septicemic patients in intensive care units before and after blood culture reporting . Scand J Infect Dis . 1991 ; 23 : 341 - 346 . PMID: 1882200
8. Pulcini C , Botelho-Nevers E , Dyar OJ , Harbarth S. The impact of infectious disease specialists on antibiotic prescribing in hospitals . Clin Microbiol Infect . 2014 ; 20 : 963 - 72 . doi: 10 .1111/ 1469 - 0691 .12751 PMID: 25039787
9. Bouza E , Sousa D , Muñoz P , Rodríguez-Créixems M , Fron C , Lechuz JG . Bloodstream infections: a trial of the impact of different methods of reporting positive blood culture results . Clin Infect Dis . 2004 ; 39 : 1161 - 9 . doi: 10 .1086/424520 PMID: 15486840
10. Harbarth S , Rutschmann O , Sudre P , Pittet D. Impact of methicillin resistance on the outcome of patients with bacteremia caused by Staphylococcus aureus . Arch Intern Med . 1998 ; 158 : 182 - 189 . PMID: 9448557
11. Friedman ND , Kaye KS , Stout JE , McGarry SA , Trivette SL , Briggs JP , et al. Health care-associated bloodstream infections in adults: A reason to change the accepted definition of community-acquired infections . Ann Intern Med . 2002 ; 137 : 791 - 797 . doi: 10 .1097/ 00019048 -200206000-00016 PMID: 12435215
12. Hamilton-Miller JMT . Switch therapy: the theory and practice of early change from parenteral to nonparenteral antibiotic administration . Clin Microbiol Infect . 1996 ; 2 : 12 - 19 . Available: http://www.ncbi. nlm.nih.gov/pubmed/11866805 PMID: 11866805
13. Shardell M , Harris AD , El-Kamary SS , Furuno JP , Miller RR , Perencevich EN . Statistical analysis and application of quasi experiments to antimicrobial resistance intervention studies . Clin Infect Dis . 2007 ; 45 : 901 - 7 . doi: 10 .1086/521255 PMID: 17806059
14. Charlson M , Szatrowski TP , Peterson J , Gold J . Validation of a combined comorbidity index . J Clin Epidemiol . 1994 ; 47 : 1245 - 51 . Available: http://www.ncbi.nlm.nih.gov/pubmed/7722560 PMID: 7722560
15. Pogue JM , Mynatt RP , Marchaim D , Zhao JJ , Barr VO , Moshos J , et al. Automated alerts coupled with antimicrobial stewardship intervention lead to decreases in length of stay in patients with gram-negative bacteremia . Infect Control Hosp Epidemiol . 2014 ; 35 : 132 - 8 . PMID: 24442074
16. Jenkins TC , Price CS , Sabel AL , Mehler PS , Burman WJ . Impact of routine infectious diseases service consultation on the evaluation, management, and outcomes of Staphylococcus aureus bacteremia . Clin Infect Dis . 2008 ; 46 : 1000 - 8 . doi: 10 .1086/529190 PMID: 18444816
17. Tissot F , Calandra T , Prod'hom G , Taffe P , Zanetti G , Greub G , et al. Mandatory infectious diseases consultation for MRSA bacteremia is associated with reduced mortality . J Infect. Elsevier Ltd; 2014; doi: 10 .1016/j.jinf. 2014 . 05 .004
18. Lesprit P , Landelle C , Brun-Buisson C. Clinical impact of unsolicited post-prescription antibiotic review in surgical and medical wards: a randomized controlled trial . Clin Microbiol Infect . 2013 ; 19 : E91 - 7 . doi: 10 .1111/ 1469 - 0691 .12062 PMID: 23153410
19. Forrest GN , Van Schooneveld TC , Kullar R , Schulz LT , Duong P , Postelnick M . Use of electronic health records and clinical decision support systems for antimicrobial stewardship . Clin Infect Dis . 2014 ; 59 Suppl 3 : S122 - 33 . doi: 10 .1093/cid/ciu565 PMID: 25261539
20. Kullar R , Goff D a. Transformation of Antimicrobial Stewardship Programs Through Technology and Informatics. Infect Dis Clin North Am . 2014 ; 28 : 291 - 300 . doi: 10 .1016/j.idc. 2014 . 01 .009 PMID: 24857394
21. Health at a Glance 2015 [ Internet] . OECD Publishing; 2015 .
22. Pollack L a , Srinivasan A. Core elements of hospital antibiotic stewardship programs from the Centers for Disease Control and Prevention . Clin Infect Dis . 2014 ; 59 Suppl 3 : S97 - 100 . doi: 10 .1093/cid/ ciu542 PMID: 25261548
23. Avdic E , Carroll KC . The role of the microbiology laboratory in antimicrobial stewardship programs . Infect Dis Clin North Am. Elsevier Inc ; 2014 ; 28 : 215 - 35 . doi: 10 .1016/j.idc. 2014 . 01 .002
24. Bauer K A , Perez KK , Forrest GN , Goff D A . Review of rapid diagnostic tests used by antimicrobial stewardship programs . Clin Infect Dis . 2014 ; 59 Suppl 3 : S134 - 45 . doi: 10 .1093/cid/ciu547 PMID: 25261540