Upload
md
View
216
Download
0
Embed Size (px)
Citation preview
Demonstration of the Weighted-Incidence Syndromic Combination Antibiogram: An EmpiricPrescribing Decision AidAuthor(s): Courtney Hebert, MD; Jessica Ridgway, MD; Benjamin Vekhter, PhD; Eric C.Brown, PhD; Stephen G. Weber, MD; Ari Robicsek, MDSource: Infection Control and Hospital Epidemiology, Vol. 33, No. 4, Special Topic Issue:Antimicrobial Stewardship (April 2012), pp. 381-388Published by: The University of Chicago Press on behalf of The Society for Healthcare Epidemiologyof AmericaStable URL: http://www.jstor.org/stable/10.1086/664768 .
Accessed: 17/05/2014 00:56
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .http://www.jstor.org/page/info/about/policies/terms.jsp
.JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range ofcontent in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new formsof scholarship. For more information about JSTOR, please contact [email protected].
.
The University of Chicago Press and The Society for Healthcare Epidemiology of America are collaboratingwith JSTOR to digitize, preserve and extend access to Infection Control and Hospital Epidemiology.
http://www.jstor.org
This content downloaded from 194.29.185.181 on Sat, 17 May 2014 00:56:13 AMAll use subject to JSTOR Terms and Conditions
infection control and hospital epidemiology april 2012, vol. 33, no. 4
o r i g i n a l a r t i c l e
Demonstration of the Weighted-Incidence Syndromic CombinationAntibiogram: An Empiric Prescribing Decision Aid
Courtney Hebert, MD;1 Jessica Ridgway, MD;2 Benjamin Vekhter, PhD;3 Eric C. Brown, PhD;4
Stephen G. Weber, MD;2,3 Ari Robicsek, MD3,4,5
objective. Healthcare providers need a better empiric antibiotic prescribing aid than the traditional antibiogram, which supplies noinformation on the relative frequency of organisms recovered in a given infection and which is uninformative in situations where multipleantimicrobials are used or multiple organisms are anticipated. We aimed to develop and demonstrate a novel empiric prescribing decisionaid.
design/setting. This is a demonstration involving more than 9,000 unique encounters for abdominal-biliary infection (ABI) andurinary tract infection (UTI) to a large healthcare system with a fully integrated electronic health record (EHR).
methods. We developed a novel method of displaying microbiology data called the weighted-incidence syndromic combination anti-biogram (WISCA) for 2 clinical syndromes, ABI and UTI. The WISCA combines simple diagnosis and microbiology data from the EHRto (1) classify patients by syndrome and (2) determine, for each patient with a given syndrome, whether a given regimen (1 or moreagents) would have covered all the organisms recovered for their infection. This allows data to be presented such that clinicians can seethe probability that a particular regimen will cover a particular infection rather than the probability that a single drug will cover a singleorganism.
results. There were 997 encounters for ABI and 8,232 for UTI. A WISCA was created for each syndrome and compared with a traditionalantibiogram for the same period.
conclusions. Novel approaches to data compilation and display can overcome limitations to the utility of the traditional antibiogramin helping providers choose empiric antibiotics.
Infect Control Hosp Epidemiol 2012;33(4):381-388
Affiliations: 1. Department of Biomedical Informatics, Ohio State University, Columbus, Ohio; 2. University of Chicago Medical Center, Chicago, Illinois;3. Center for Health and the Social Sciences and Pritzker School of Medicine, University of Chicago, Chicago, Illinois; 4. Center for Clinical and ResearchInformatics, NorthShore University HealthSystem, Evanston, Illinois; 5. Department of Medicine and Department of Health Information Technology,NorthShore University HealthSystem, Evanston, Illinois.
Received October 3, 2011; accepted December 27, 2011; electronically published March 15, 2012.� 2012 by The Society for Healthcare Epidemiology of America. All rights reserved. 0899-823X/2012/3304-0011$15.00. DOI: 10.1086/664768
Heathcare providers regularly must choose an antibiotic reg-imen for their patients before culture results are available.Recent studies have shown that an incorrect choice can beharmful to patients,1-4 while inappropriate use of broad-spec-trum antibiotics can drive resistance5 and lead to increasedhospital costs.6
An important tool of healthcare providers as well as an-timicrobial stewardship programs is the traditional antibio-gram.7 These antibiograms can provide guidance that is hos-pital specific and reflects local resistance patterns; however,they have several major limitations. (1) They do not providesyndrome or disease-specific advice. A traditional antibio-gram can indicate that 20% of Escherichia coli are resistantto fluoroquinolones but not whether this percentage variesbetween urinary and respiratory isolates. (2) They give nohint as to the likely distribution of organisms in a giveninfection. The information that 20% of E. coli are resistant
to fluoroquinolones has different significance for an infectionusually caused by E. coli than for an infection only occa-sionally caused by this organism. (3) They are particularlyunhelpful for infections that are generally polymicrobial orthat are empirically treated with more than 1 agent. In sum,the perspective of the traditional antibiogram is that of themicrobiology laboratory, asking the question, “Which drugworks for this bug?” This is not the perspective of the treatingphysician, especially the generalist, who is actually asking,“Which regimen works for this syndrome?”
A previously described technique—the combination anti-biogram—provides some help.8-9 This shows the likelihoodthat at least 1 drug in a regimen comprising multiple anti-microbials will “cover” a given organism. While useful whenthe organism is known but susceptibilities are not yet avail-able, it does not give the likelihood that the combination of
This content downloaded from 194.29.185.181 on Sat, 17 May 2014 00:56:13 AMAll use subject to JSTOR Terms and Conditions
382 infection control and hospital epidemiology april 2012, vol. 33, no. 4
table 1. Major Susceptibility Assumptions Made by the Investigators
Organism and antibiotic Assumptions made
Enterococcus speciesErtapenem Not coveredCephalosporins Not coveredCiprofloxacin Not covered unless specifically tested and susceptible
Pseudomonas aeruginosaAmpicillin-sulbactam Not coveredTrimethoprim-sulfamethoxazole Not coveredCefazolin Not coveredCeftriaxone Not coveredErtapenem Not covered
Staphylococcus aureusAmpicillin-sulbactam If resistant to methicillin, then not covered by any beta-lactamPiperacillin-tazobactam If resistant to methicillin, then not covered by any beta-lactamCephalosporins If resistant to methicillin, then not covered by any beta-lactamCarbapenems If resistant to methicillin, then not covered by any beta-lactam
Beta-hemolytic StreptococcusAmpicillin-sulbactam Covered by all beta-lactamsPiperacillin-tazobactam Covered by all beta-lactamsCephalosporins Covered by all beta-lactamsCarbapenems Covered by all beta-lactamsCiprofloxacin Not covered unless specifically tested and susceptible
Gram-negative anaerobic organismsMetronidazole CoveredAmpicillin-sulbactam CoveredPiperacillin-tazobactam CoveredCarbapenems CoveredCephalosporins Not coveredCiprofloxacin Not covered
Gram-positive organismsAmpicillin-sulbactam Covered if covered by ampicillin or penicillinPiperacillin-tazobactam Covered if covered by ampicillin-sulbactamMeropenem Covered if covered by piperacillin-tazobactam
EnterobacteriaceaeAmpicillin-sulbactam Covered if covered by ampicillin or penicillinPiperacillin-tazobactam Covered if covered by ampicillin-sulbactam, except for AcinetobacterMeropenem Covered if covered by piperacillin-tazobactam unless tested and resistant
agents will cover all recovered organisms, and it is not syn-drome specific.
To address these limitations, we have developed a newmethod of displaying microbiological data to guide empirictherapy. We have named the antibiogram we created aweighted-incidence syndromic combination antibiogram(WISCA). The WISCA displays, for a given infection syn-drome, the likelihood that each of a panel of regimens willtreat all relevant organisms recovered in a patient with thatsyndrome, answering the clinician’s question, “Which regi-men works for this syndrome?” In this work, we have useddata from a 4-hospital health system to demonstrate a WISCAand compare it with a traditional antibiogram.
methods
Overview
The objective of this work was to demonstrate the creation ofa WISCA and compare it with a traditional antibiogram, using
a large clinical data set from a 4-hospital health system. Wefocused our analysis on frequently used antibiotic regimens for2 common infectious syndromes—urinary tract infection(UTI) and abdominal-biliary tract infection (ABI)—that arecommunity onset and require hospitalization.
Settings and Participants
The data analyzed in this study were collected at NorthShoreUniversity HealthSystem. NorthShore is a 4-hospital aca-demic health system with more than 75 outpatient clinics andmore than 2,500 affiliated physicians in the metropolitan Chi-cago area. Among the hospitals together, all common surgicaland medical procedures are provided, including comprehen-sive cancer care. Solid organ transplants and allogeneic bonemarrow transplants are not performed.
NorthShore has been fully paperless since 2003, and allinpatient and outpatient entities are connected through thesame comprehensive electronic health record (EHR), which
This content downloaded from 194.29.185.181 on Sat, 17 May 2014 00:56:13 AMAll use subject to JSTOR Terms and Conditions
development of a new syndromic combination antibiogram 383
table 2. Baseline Characteristics of the Study Population
UTI encounters( )n p 8,232
ABI encounters( )n p 997
Age, median (IQR), years 81 (69–87) 64 (51–76)Women 5,887 (72) 496 (50)ER or inpatient visit in past 6 months 4,529 (55) 465 (47)Diabetes mellitus 2,451 (30) 173 (17)Asthma 1,067 (13) 75 (8)CHF 3,143 (38) 172 (17)COPD 1,776 (22) 106 (11)HIV 24 (0.3) ...Chronic liver disease 239 (3) 24 (2)Nursing home resident 1,703 (21) 43 (4)MDRO cultured in the previous yeara 747 (9) 44 (4)In the past 30 days received
Any antibacterial 1,760 (21) 282 (28)TMP-SMX 157 (2) ...Carbapenem 69 (1) 28 (3)Cephalosporin 657 (8) 119 (12)Fluoroquinolone 790 (10) 104 (10)Macrolide 119 (1) 5 (1)Antipseudomonal penicillin 219 (3) 64 (6)
In the past 30–180 days receivedAny antibacterial 3,270 (40) 304 (30)TMP-SMX 428 (5) ...Carbapenem 176 (2) 29 (3)Cephalosporin 1,446 (18) 144 (14)Fluoroquinolone 2,077 (25) 154 (15)Macrolide 437 (5) 27 (3)Antipseudomonal penicillin 554 (7) 74 (7)
note. Data are no. (%), unless otherwise indicated. ABI, abdominal biliaryinfection; CHF, congestive heart failure; COPD, chronic obstructive pulmo-nary disease; ER, emergency room; HIV, human immunodeficiency virus;IQR, interquartile range; MDRO, multidrug-resistant organism; TMP-SMX,trimethoprim-sulfamethoxazole; UTI, urinary tract infection.a Defined as a culture positive for methicillin-resistant Staphylococcus aureus;vancomycin-resistant enterococcus; any extended-spectrum beta-lactamase;Escherichia coli or Klebsiella resistant to ceftazidime; or a carbapenem/cefta-zidime-resistant Pseudomonas, Enterobacter, Acinetobacter, or Citrobacter inthe previous year.
feeds data into the NorthShore enterprise data warehouse(EDW). The EDW contains a comprehensive characterizationof more than 1,000,000 NorthShore patients, including dem-ographics, procedures, laboratory results, medications (in-patient and outpatient), and diagnoses. All study data wereobtained from the EDW.
Study Design and Definitions
The study was designed as a retrospective review of EHRs.Eligible patients were all those adult patients admitted toNorthShore between January 2006 and June 2010 who hada final diagnosis code consistent with UTI or ABI (based onInternational Classification of Diseases, Ninth Revision codesduring admission) and had a positive culture from the pri-mary infection site collected on day 1 through day 2 of hos-pitalization for UTI and days 1–4 of hospitalization for ABI.
Unique patients could be included more than once if theyhad more than 1 eligible encounter during the study period.
Culture and susceptibility data were collected from micro-biology records. For each case, a determination was made ofwhether the infection would have been “covered” by each ofa panel of antimicrobial regimens. An infection was consid-ered covered by a regimen if each relevant organism isolatedfrom the body site of interest was susceptible to at least 1antimicrobial agent in the regimen. Intermediate resistancewas considered resistant. For those cases where data were notavailable on a specific antibiotic (eg, vancomycin suscepti-bilities would not be reported for E. coli), assumptions re-garding susceptibility were made on the basis of expert opin-ion and literature review (eg, E. coli would be considered“resistant” to vancomycin). Further, a determination wasmade regarding which organisms were not “relevant” in a
This content downloaded from 194.29.185.181 on Sat, 17 May 2014 00:56:13 AMAll use subject to JSTOR Terms and Conditions
ta
ble
3.Tr
adit
ion
alA
nti
biog
ram
for
the
Stu
dySi
teov
erth
eE
nti
reSt
udy
Per
iod
Stap
hylo
cocc
us
aure
us
Ent
eroc
occu
s
spec
ies
Bet
a-he
mol
ytic
Stre
ptoc
occu
s
grou
pB
Stre
ptoc
occu
s
pneu
mon
iae
Esc
heri
chia
coli
Kle
bsie
lla
pneu
mon
iae
Pro
teus
mir
abili
s
Pse
udom
onas
aeru
gino
sa
Cit
roba
cter
spec
ies
Ent
erob
acte
r
cloa
cae
Kle
bsie
lla
oxyt
oca
Ent
erob
acte
r
aero
gene
s
Serr
atia
mar
cesc
ens
Mor
gane
lla
mor
gani
i
Am
pici
llin
1396
9769
540
80...
20
10
00
Trim
eth
opri
m-
sulf
amet
hox
azol
e99
......
...75
9085
...90
9095
9891
80
Flu
oroq
uin
olon
e74
69...
9884
9785
8596
9698
9997
91
Cef
azol
in65
......
...92
9690
...47
176
31
0
Cef
tria
xon
e...
...10
0...
9798
97...
9488
9894
9999
Ert
apen
em...
......
...10
010
010
0...
100
9910
010
010
099
Cef
tazi
dim
e...
......
...97
9896
9794
8999
9499
99
Am
pici
llin
-
sulb
acta
m...
......
...82
9397
...66
2991
245
15
Pip
erac
illin
-
tazo
bact
am...
......
...98
9610
097
9492
9692
9999
Mer
open
em...
......
...10
099
100
9498
100
100
100
100
...
no
te.
Dat
aar
epe
rcen
tof
isol
ates
susc
epti
ble.
Th
isan
tibi
ogra
min
clu
des
the
firs
tu
niq
ue
bact
eria
lis
olat
esfr
oman
ybo
dysi
te,
from
all
outp
atie
nt
loca
tion
s(i
ncl
udi
ng
emer
gen
cyro
om)
for
the
enti
rest
udy
peri
od.
Elli
pses
indi
cate
that
ther
ear
ele
ssth
an10
0se
nsi
tivi
tyte
sts
don
efo
rth
isco
mbi
nat
ion
ofm
edic
atio
nan
dor
gan
ism
over
the
stu
dype
riod
.Th
e14
mos
tfr
equ
entl
yis
olat
edor
gan
ism
s,ex
clu
din
gco
agu
lase
neg
ativ
eSt
aphy
loco
ccus
,ar
esh
own
.
This content downloaded from 194.29.185.181 on Sat, 17 May 2014 00:56:13 AMAll use subject to JSTOR Terms and Conditions
development of a new syndromic combination antibiogram 385
table 4. Weighted-Incidence Syndromic CombinationAntibiogram
Infectionscovered, %
Antimicrobial regimen UTI ABI P a
Trimethoprim-sulfamethoxazole 55 ... ...Ciprofloxacin (�MTZ)b 62 37 !.001Cefazolin (�MTZ)b 62 47 !.001Ceftriaxone (�MTZ)b 71 59 !.001Ertapenem 71 63 !.001Ceftazidime (�MTZ)b 76 65 !.001Ampicillin-sulbactam 83 68 !.001Ampicillin and gentamicin (�MTZ)b 84 81 .10Cipro � MTZ � vancomycin ... 84 ...Ceftazidime (�MTZ)b � vancomycin 88 93 !.001Ertapenem � vancomycin ... 88 ...Piperacillin-tazobactam 89 88 .70Piperacillin-tazobactam � vancomycin 91 93 .04Meropenem 91 91 .72Meropenem � vancomycin 93 96 !.001
note. Listed are 15 common antimicrobial regimens and thepercentage of the time that a regimen would cover all recoveredorganisms in an individual patient’s infection. This was cal-culated by taking the number of patients covered by a regimenas the numerator and all patients for which susceptibilities areknown or imputed for this agent as the denominator. ABI,abdominal-biliary infection; Cipro, ciprofloxacin; MTZ, met-ronidazole; UTI, urinary tract infections.a P values compare coverage of a given regimen between UTIand ABI.b Metronidazole included in regimen only for ABIs.
particular body site (eg, Staphylococcus epidermidis in a peri-toneal culture). Nonrelevant organisms were removed fromthe analysis. See Table 1 for a list of the major assumptionsthat were made. Demographic, clinical, laboratory, and fullmedication history data were obtained on each patient fromthe EDW.
The UTI and ABI WISCA was created using the data above.The WISCA is constructed to provide stratification by diseasestate (ie, there is a different antibiogram for different diseasesyndromes, hence “syndromic”). The rows characterize reg-imens rather than individual antimicrobial agents, and someof these regimens include multiple agents (hence “combi-nation antibiogram”). The percentage of infections coveredwas calculated by dividing the number of cases for whomeach relevant organism was susceptible to at least 1 of theagents in a given regimen (“covered” cases) by the numberof cases (not organisms) for which there were susceptibilitydata for the full antibiotic regimen. This percentage is sen-sitive not only to the susceptibilities of the organisms recov-ered from a given infection type but also to their relativedistribution (hence “weighted incidence”).
For comparison, we tabulated a traditional antibiogram.This was constructed by including all first isolates, from all
body sites, cultured from outpatients (including emergencyroom) over the entire study period and all sites. We choseto include only outpatient cultures because our WISCA pa-tient population included only community-onset infections,so it would not be represented well by an inpatient antibio-gram. As with the WISCA, a patient could be included morethan once if they had multiple encounters during the studyperiod. Institutional review board approval was obtained forthis study.
Statistics
Comparisons of coverage between the UTI and ABI WISCAswere analyzed using the Fisher exact test.
results
During the study period, there were 997 encounters for ABI(901 unique patients) and 8,232 encounters for UTI (6,039unique patients). Patient characteristics are shown in Table2.
Traditional Antibiogram
The traditional antibiogram is shown in Table 3. It includes62,308 unique isolates from 36,897 patients. This antibiogramshows which antibacterial regimens are most likely to covereach included bacterial isolate. Looking at the data in thisway gives us some useful information. For example, in thispopulation, there is relatively high sensitivity of E. coli tofluoroquinolones (84%), ampicillin-sulbactam (82%), andeven trimethoprim-sulfamethoxazole (75%). There is almostno resistance for the broadest-spectrum antibiotics, such asertapenem and meropenem. Of note, there are many missingcells, implying that susceptibility results were not availablefor this “bug-drug” combination. This may be because theantibiotic is known to lack coverage (ampicillin-sulbactamfor Pseudomonas) or because this antibiotic was not specifi-cally tested for, but susceptibility may be assumed (pipera-cilllin-tazobactam would be assumed to cover 13% of Staph-ylococcus aureus because ampicillin-sulbactam covers 13%).
WISCA
A WISCA for each syndrome is shown in Table 4. Whencompared with the traditional antibiogram, the WISCA givesus new information. Although 84% of E. coli is covered bya fluoroquinolone in the traditional antibiogram, only 62%of the UTI patients would have been adequately covered bya fluoroquinolone. The lower coverage reflects the contri-bution of other organisms (eg, enterococci) that are not wellcovered by ciprofloxacin. This percentage drops even furtherwhen we look at the ABI group. Only 37% are covered inthis group (even with anaerobic coverage added). The dif-ference between these 2 groups can be explained by the prev-alence of E. coli in the 2 infections. E. coli makes up 49% ofall isolates recovered from UTI patients but only 11% of thosefrom patients with ABI (data not shown).
This content downloaded from 194.29.185.181 on Sat, 17 May 2014 00:56:13 AMAll use subject to JSTOR Terms and Conditions
386 infection control and hospital epidemiology april 2012, vol. 33, no. 4
table 5. Urinary Tract Infection (UTI) Weighted-Incidence Syndromic Combination Antibiogram Illustration forSeveral Antimicrobial Regimens
UTIs covered, %
Ciprofloxacin Ampicillin-sulbactam Piperacillin-tazobactam
In all encounters 62 83 89In individuals 165 years of age 61 83 89In individuals with a recent ER or inpatient visita 55 77 85In individuals with COPD 52 78 86In nursing home residents 36 73 81In individuals with an MDROb in the past year 28 54 71In individuals with FQ exposure in the past 30 days 20 57 68In individuals with none of the above 82 93 95
note. COPD, chronic obstructive pulmonary disease; ER, emergency room; FQ, fluoroquinolone; MDRO, multidrug-resistant organism.a In the past 6 months.b Defined as a culture positive for methicillin-resistant Staphylococcus aureus; vancomycin-resistant enterococcus; anyextended-spectrum beta-lactamase; Escherichia coli or Klebsiella resistant to ceftazidime; or a carbapenem/ceftazidime-resistant Pseudomonas, Enterobacter, Acinetobacter or Citrobacter in the previous year.
Another interesting difference is in the susceptibility ofceftriaxone. In the traditional antibiogram, it shows excellentcoverage for most gram positives and gram negatives. How-ever, in the WISCA, ceftriaxone alone would cover only 71%of UTIs and 59% of ABIs. One reason for this is that cef-triaxone does not cover enterococci, which make up 10% ofthe UTI isolates and 12% of the abdominal-biliary isolates(often as part of a polymicrobial infection). A similar patternis seen with ertapenem. Alone, it covers only 63% of casesof ABI, but in combination with vancomycin, the likelihoodof coverage rises to 88%.
Table 5 demonstrates how a WISCA can be stratified onthe basis of patient characteristics. While overall, a regimencomprising ciprofloxacin alone would have adequately cov-ered only 62% of UTI, in patients with few risk factors, itwould have been adequate 82% of the time, perhaps sufficientcoverage probability to be a good empiric choice in an oth-erwise healthy patient. On the other hand, piperacillin-tazobactam adequately covers UTI in most scenarios but maynot be sufficient empiric therapy alone in a patient who hadrecently received a fluoroquinolone (only 68% chance ofcoverage).
discussion
Selection of empiric antimicrobial therapy is a very commonchallenge for clinicians, and uncertainty in this process leadsto errors of both overtreatment and undertreatment.1,2,10 Pro-viding guidance for empiric antimicrobial decision makingis a key function of an antimicrobial stewardship program.While the traditional antibiogram is a mainstay of this pro-cess, it suffers important limitations, including minimal stra-tification, no information regarding which organisms are ex-pected in a given infection, no guidance in situations wheremultiple antimicrobials are used or multiple organisms areanticipated, and an unhelpful focus on bug-drug combina-
tions instead of on syndrome-regimen combinations. To beuseful at the time empiric therapy is selected, an antibiogramwould integrate data regarding likely organisms and likelysusceptibilities, providing information about the likelihoodthat a given regimen would “cover” a specific infection syn-drome, whether a case had monomicrobial or polymicrobialinvolvement.
We have demonstrated that such an antibiogram—aWISCA—can be generated with relative simplicity. The crea-tion of a WISCA requires knowledge of only which body sitewas affected, which organisms were recovered from a patient,and what the susceptibilities of the isolates were. While weincorporated diagnosis codes into our system of determiningwhich body site was affected, in principle this could be doneusing only data on the body site from which a specimen wascollected (eg, urine, sputum, peritoneal fluid). Thus, most lab-oratories with the capacity to generate a traditional antibiogramcould create a WISCA as well. This holds true for automatedthird-party surveillance vendors (eg, CareFusion-MedMined,Hospira-TheraDoc, Premier-SafetySurveillor).
This study is a proof of concept and, as such, is not in-tended to establish definitively all the assumptions that mustgo into creating a WISCA. Such assumptions include whichregimens are appropriate for managing a given syndrome,which unknown susceptibilities can be imputed from whichknown susceptibilities (eg, if E. coli is susceptible to cefazolin,can it be assumed susceptible to piperacillin-tazobactam?),and which organisms are not “relevant” in a given infection(eg, should enterococci be removed from the ABI WISCA?)These issues need to be the subject of further work and willprobably require some local decision making at each centerwhere WISCAs are used. In the creation of our WISCA andtraditional antibiogram, we allowed patients to contributemore than once if they had more than 1 encounter. Rec-ommendations for antibiogram creation suggest only 1 isolate
This content downloaded from 194.29.185.181 on Sat, 17 May 2014 00:56:13 AMAll use subject to JSTOR Terms and Conditions
development of a new syndromic combination antibiogram 387
per patient per antibiogram.11 However, antibiograms gen-erally span only 1 year, and our antibiogram included 5 yearsand 4 hospitals. In this scenario, we thought it was reasonableto allow multiple encounters for 1 patient. It is not clear atthis time how a WISCA would change clinical practice. It ispossible that physicians might see the overall lower percentagesusceptible (as compared with the traditional antibiogram)and opt for a broader-spectrum agent on a more regular basis.To avoid this, physicians would need to be educated to un-derstand the implications of the WISCA. However, in someways, the WISCA might allow a physician more comfort inchoosing a narrower-spectrum agent (eg, ampicillin-sulbac-tam) when they see how it compares with a broader-spectrumagent (piperacillin-tazobactam) when confronted with a low-risk patient (see Table 5). This is especially true if WISCAsare stratified by patient characteristics.
There are limitations to the WISCA. As with a traditionalantibiogram, it represents only positive cultures, with poten-tial resultant biases. Also, some organisms included in theanalysis may represent colonization, another potential sourceof bias that WISCA shares with the traditional antibiogram.Perhaps more importantly, the WISCA represents a changefrom current practice and will require adjustment in howdata are handled by antibiogram compilers. Certain infectionswill not be amenable to WISCA creation because positivecultures are very rare (eg, simple cellulitis). Other syndromeswill be a challenge for smaller centers because positive cul-tures are infrequent (eg, pneumonia, meningitis). However,common conditions with relatively frequent positive cultures(eg, UTI, surgical site infection, diabetic foot infection, ABI)account for much antimicrobial misuse,10 and improvingstewardship around this use has substantial potential forbenefit.
The central insight in our proposed methodology is asfollows. In a traditional antibiogram, the unit of analysis isthe organism, and the outcome of interest is susceptibility.This does not yield information that is very useful for se-lecting empiric therapy before any microbiology data areavailable. In the WISCA, the unit of analysis is the patient,and the outcome of interest is coverage by a regimen. Inaddition to the advantages demonstrated above (handling ofrelative frequencies of the organisms recovered in a giveninfection, handling of combinations of organisms and anti-microbials), a key benefit of this “framing” is that it allowsmore sophisticated modeling techniques to be applied to theoutcome “coverage of a patient by a regimen.” For example,from the data set in this study, we have found that withmultivariable modeling techniques, we can accurately predicta probability that each regimen will “cover” a patient, giventheir specific clinical characteristics (data not shown). Sucha tool can enable highly specific empiric choices—broad spec-trum where needed and narrow spectrum where not—thathave the potential to reduce unnecessary antimicrobial use.To create this personalized WISCA, we capitalized on dataavailable from structured fields in a sophisticated EHR. Many
centers are in the early stages of EHR adoption, and thepersonalized WISCA is outside the technical capacity of mostcenters at present.12 However, this capacity is expected tomarkedly expand in coming years,13 and an antibiogram fullypersonalized—and continuously updating—could become areality.
In conclusion, we have presented a new form of antibio-gram—the WISCA—as an improved tool for aiding empiricantimicrobial decision making, a critical point in the stew-ardship process. This technique overcomes several major lim-itations of the traditional antibiogram and sets the stage formore sophisticated guidance tools.
acknowledgments
Financial support. This work was funded by a Merck Investigator Initiatedgrant to A.R. and a Centers for Disease Control and Prevention FoundationGet Smart in Healthcare Settings award to A.R.
Potential conflicts of interest. All authors report no conflicts of interestrelevant to this article. All authors submitted the ICMJE Form for Disclosureof Potential Conflicts of Interest, and the conflicts that the editors considerrelevant to this article are disclosed here.
Address correspondence to Courtney Hebert, MD, 3190 Graves Hall, 333West 10th Avenue, Columbus, OH 43210 ([email protected]).
Presented in part: 48th Annual Meeting of the Infectious Diseases Societyof America; Vancouver, Canada; October 21–24, 2010.
references
1. Lee S, Kim Y, Chung DR. Impact of discordant empirical therapyon outcome of community-acquired bacteremic acute pyelo-nephritis. J Infect 2011;62:159–164.
2. Micek ST, Welch EC, Khan J, et al. Resistance to empiric an-timicrobial treatment predicts outcome in severe sepsis associ-ated with gram-negative bacteremia. J Hosp Med 2011;6:405–410.
3. Zilberberg MD, Shorr AF, Micek ST, et al. Hospitalizations withheathcare-associated complicated skin and skin structure infec-tions: impact of inappropriate empiric therapy on outcomes. JHosp Med 2010;5:535–540.
4. Paul M, Shani V, Muchtar E, Kariv G, Robenshtok E, LeiboviciL. Systematic review and meta-analysis of the efficacy of ap-propriate empiric antibiotic therapy for sepsis. Antimicrob AgentsChemother 2010;54:4851–4863.
5. Fridkin SK, Edwards JR, Courval JM, et al. The effect of van-comycin and third-generation cephalosporins on prevalence ofvancomycin-resistant enterococci in 126 U.S. adult intensivecare units. Ann Intern Med 2001;135:175–183.
6. Mauldin PD, Salgado CD, Hansen IS, Surup DT, Bosso JA.Attributable hospital cost and length of stay associated withhealth care–associated infections caused by antibiotics-resistantgram-negative bacteria. Antimicrob Agents Chemother 2010;54:109–115.
7. Dellit TH, Owens RC, McGowan JE, et al. Infectious DiseasesSociety of America and the Society for Healthcare Epidemiologyof America guidelines for developing an institutional programto enhance antimicrobial stewardship. Clin Infect Dis 2007;44:159–177.
This content downloaded from 194.29.185.181 on Sat, 17 May 2014 00:56:13 AMAll use subject to JSTOR Terms and Conditions
388 infection control and hospital epidemiology april 2012, vol. 33, no. 4
8. Christoff J, Tolentino J, Mawdsley E, Matushek S, Pitrak D,Weber SG. Optimizing empirical antimicrobial therapy for in-fection due to gram-negative pathogens in the intensive careunit: utility of a combination antibiogram. Infect Control HospEpidemiol 2010;31:256–261.
9. Fox B, Shenk G, Peterson D, Spiegel C, Maki D. Choosing moreeffective antimicrobial combinations for empiric antimicrobialtherapy of serious gram-negative rod infections using a dualcross-table antibiogram. Am J Infect Control 2008;36:S57–S61.
10. Hecker MT, Aron DC, Patel NP, Lehmann MK, Donskey CJ.Unnecessary use of antimicrobials in hospitalized patients: cur-
rent patterns of misuse with an emphasis on the antianaerobicspectrum of activity. Arc Intern Med 2003;163:972–978.
11. Hindler JF, Stelling J. Analysis and presentation of cumulativeantibiograms: a new consensus guideline from the Clinical andLaboratory Standards Institute. Clin Infect Dis 2007;44:867–873.
12. Jha AK, DesRoches CM, Kralovec PD, Joshi MS. A progressreport on electronic health records in U.S. Hospitals. Health Aff(Milwood) 2010;29:1951–1957.
13. Department of Health and Human Services. Medicare and Med-icaid programs; electronic health record incentive program; pro-posed rule. Fed Regist 2010;75:1844–2011.
This content downloaded from 194.29.185.181 on Sat, 17 May 2014 00:56:13 AMAll use subject to JSTOR Terms and Conditions