MICROBIOLOGY, Dec. 1978, p. 0095-1137/78/0008-0657$02.00/0
Copyright © 1978 AmericanSocietyforMicrobiology Printed in U.S.A.
Multi-Laboratory
Evaluation of
an
Automated
Microbial
Detection/Identification
System
P. B. SMITH,* T. L. GAVAN,2 H. D. ISENBERG,3A. SONNENWIRTH,4W. I. TAYLOR,5 J.A. WASHINGTON Il,6 ANDA. BALOWS'
CenterforDiseaseControl,PublicHealthService, Department of Health,Education,and Welfare,Atlanta,
Georgia 30333;' Cleveland ClinicFoundation, Cleveland,Ohio44106;2 LongIslandJewish-Hillside
MedicalCenter,NewHyde Park,NewYork11040;3JewishHospital ofSt.Louis,St.Louis,Missouri 6311i0; St.Mary ofNazarethHospital, Chicago,Illinois60622;5andMayoClinicandMayo Foundation,
Rochester,Minnesota 559016
Receivedfor publication2August 1978
An automated andcomputerizedsystem(Automicrobic System[AMS])forthe
detection of frequently encountered bacteria in clinical urine specimens was
tested inacollaborativestudyamongsix laboratories. Thesensitivity,specificity,
reliability, andreproducibility of the AMSweredetermined, and thesystemwas
compared with conventional detection and identification systems.Inthisstudy,
pure cultures and mixtures ofpurecultureswereusedtosimulateclinical urine specimens.Withpurecultures, the sensitivity of the AMS inidentifyingthe nine
groupsoforganismsmostcommonly found in urine averaged 92.8%.The
specific-ity averaged 99.4%, and the reliabilspecific-ity ofa positiveresult averaged 92.1%. The
latter value was strongly influenced by a relatively high occurrence of false
positive Escherichia coli results. The AMS wascapable ofdetecting growth of mostorganisms, including those which itwasnotdesignedto identify. However, itidentifiedsomeoftheseincorrectlyas commonurinarytractflora. Reproduci-bility of results, both within laboratories andamongdifferent laboratories, was
high. Fast-growing organisms, suchasE. coli and Klebsiella/Enterobacter
spe-cies,were detected often at cellpopulations well below the AMS enumeration threshold of 70,000/ml. In mixed culture studies, high levels of sensitivity and specificityweremaintained, but when Serratia specieswerepresentinmixtures
with otherorganisms, therewasoftenafalse positivereportof E.coli.Theoverall
performance of the AMS wasconsidered satisfactory under the testconditions used.
A novelsystem for the automated and
com-puterized detection,
enumeration, and identifi-cation oforganisms
inurine, the Automicrobic System (AMS), has recently been described by Aldridge et al. (1). This system represents adrastic
departure
from conventionalmethodol-ogy in microbiology, in that no prior isolation
and purification of
organisms
is required;in-stead,theAMSisdesignedtodetectandidentify specific groups and/or species oforganisms in urine specimens representing either single or
mixedinfections.
During its developmental stages, the AMS wasevaluated with both seeded(simulated) and clinical specimens in a study reported by Son-nenwirth (2). He found about 90% agreement between AMSresults and conventional results withEscherichia coli, Klebsiella/Enterobacter species (the AMS does notdifferentiatethem),
Serratia species, Proteus species, Citrobacter freundii, group D enterococci, andyeasts
(pri-marily Candida and Torulopsis species), but
only about 75%agreementwithStaphylococcus
aureus andPseudomonasaeruginosa. Sonnen-wirth's study also indicated the need for im-provement of (i) reagents, (ii) sensitivity and
specificityfor someorganisms, and (iii) manual
methods used forcomparison andvalidationof results.
The present collaborative study was
under-taken after the manufacturer had instituted some of the recommended improvements and had modified the instruments. A two-stage,
multi-laboratory evaluation of this system was undertaken to (i) enlargeupon the data gener-ated by the twopreliminary reports (1, 2), (ii)
testurinarytractisolatesfromwidely dispersed
geographicareas,(iii) evaluatethe intra-labora-tory andinter-laboratory reproducibility of the system, and (iv) challenge the sensitivity and
specificity of theAMS. Seededspecimens were used in this evaluation. In another evaluation 657
on February 7, 2020 by guest
http://jcm.asm.org/
study, to be reported, clinical specimens from
various hospitals will beused.
MATERIALS AND METHODS Cultures. In the firstphaseofthis study, in which only pure cultures were used, bacterial and yeast cul-tures recently isolated in the various collaborating
laboratories werelyophiized, coded,and distributed
as unknowns to each laboratory by the Center for
DiseaseControl. The selectionofcultures used (see
Table 1) was notbased on the distributionnormally
encountered in clinical urinespecimens, but rather on
the idea of extensively testingeachidentification
ca-pabiity of the AMS system. Additional cultures,
which were not among those that the AMS was
de-signed to identify, were taken from the Center for
Disease Control stockculturecollections,coded, and
distributedasviablesubculturesonagar slants. In the
second phase of the study, in which mixed cultures
wereused, oneculture ofeach of thevarious groups or
species identifiable by the AMSwasselected, coded,
anddistributedas aviableagarslanttoeach
partici-pating laboratory. Ineach laboratory, these cultures
were mixedaccording to specific instructions. In all instances, thecodesforthe unknownswere notbroken untilresults had been obtained.
Equipment. The instrumentation and disposable
components of the AMS, including the Identi-Pak, have beenthoroughly describedbyAldridgeet al. (1). The system is based on theprinciple ofutilizing an array ofselectivemedia topermit significantgrowth
of only one organism, or a group ofclosely related
organisms,inspecific microchambers. Thesystem
per-mits desiredorganismstooutgrowcompetingones, if present, and automaticallycompares initialreadings
withsubsequentones.Aseparate enumeration system,
basedonmost-probable-numbertheory andutilizing
non-inhibitory media, givesapproximate totalcounts
ofall organismspresent. Both the identification and
enumeration systems have adjustable thresholds
which must be exceeded before a positive result is
recorded. In this study, the enumeration threshold
was set at7 x104colony-formingunits(CFU) perml, but lesser populations could be detected because of the separate enumeration and identification systems and because of the highly sensitive selective media employed.
TheIdenti-Paks used in thisstudywereessentially the same as those described by Aldridge et al. (1), except that some selective media formulations had
been changed slightly, following Sonnenwirth's
rec-ommendations.
The manufacturer provided each laboratory with
thecompletesystem,allnecessarymaterials,and
spe-cific instructions foruse.These instructions have also been described(1).The AMS modelemployedhada
capacityof120specimens, containeda tape deck for
recording time-history profiles, and automatically
printed all results after 13 h of incubation.Preliminary results could be obtainedatanytime,if desired.
Media and reagents. All bacteriological media and reagents required in the manual method (see
below)weresuppliedtothecollaboratinglaboratories
fromRegionalMediaLaboratories, Inc., Lenexa,Kans.
Insofar as possible, all laboratories used media and reagents from the same production batches. Media and reagentsused in the routinemethod (see below) were those normally used in the participating labora-toriesand came fromdifferent sources.
Methods. (i) Pureculture study. Each laboratory used three distinct identification schemes on each
unknown specimen: (i) the AMS procedure, as
di-rected by the manufacturer, (ii) a manual method
uponwhich allcollaborators agreed,and(iii) whatever
routine method each of the six laboratoriesnormally used. Theprocedures employedin theAMS and
man-ual methods are described further. In addition, the
Center forDiseaseControllaboratory alsoprovideda
separate "reference" identification for each culture employed.
For the AMS procedure, each coded lyophilized
bacterialculturewas rehydrated withsterile
Trypti-casesoy broth,streaked onto a 5% sheep blood agar
plate, and incubated for 18 to 24 h at 35 to 37°C.
Identical isolated colonies were picked into sterile,
poolednormalurine (suppliedlyophiizedtoall
collab-orators) to form a suspension of about 107 CFU/ml
(by comparison with a turbidity standard), and two
10-folddilutions were madeof this suspension, again
withsterile urine used asdiluent.This final suspension wascalled the "seeded specimen" and served as inoc-ulumfor the AMS and for otherprocedures. Every
fifthspecimen receivedtwoadditional 10-folddilutions
which were used to determine some of the lower
bacterial populations which the AMS could detect.
Culturesofyeasts werehandledsimilarlytothose of
bacteria,exceptthat theoriginal cellsuspensionsfrom
isolated colonieswereadjustedto aturbidityequal to
about 108 CFU/ml. Colony counts were made ofall
culturesfrom the seededspecimen by using eithera
model480 Artek-Fisher Automatic Bacterial Colony
Counterorconventionalcolonycountingmethods.All
Identi-Paks which (i) detected the presence of an
organismbutfailedtoidentifyitor(ii) failedtodetect
an organism when one wasknown to be present by
the manual method wereenteredaseptically through
the plastic membrane of the Identi-Pak by meansof
a sterileneedle andsyringe. Contents of thepositive
controlchamberwerewithdrawn and examinedbythe
manual method.
Themanual methodwasrecognizedtohave certain
deficiencies, but was considered to be a reasonably
adequate system for the identification oforganisms
commonly encountered in urine and to have a
per-formance levelcomparabletothat of the AMS. The
colony characteristics of each bacterial culturewere
observed on the blood agar plate inoculated in the
AMS procedure described above. Selected
well-iso-lated colonieswere then identified by theprocedure shown inFig. 1.It should be noted thatidentification
ofstreptococciwaspresumptive,and thatserological
confirmation was not done. Yeasts were identified from colonies on bloodagar plates by means of the
followingtests: colonial andmicroscopic morphology;
germ tubeformation;pellicleformation;capsule pro-duction (Indiainkstain); urease;dextrose, fermenta-tion; and assimilation of sucrose, maltose, raffinose,
galactose, trehalose, and melibiose. Colony counts
weremade on each culture from the seededspecimen
on February 7, 2020 by guest
http://jcm.asm.org/
Gram stain
(rods)
Oxidase
~1
Enterobacteriaceae Tests
i
Fluorescein, Pyocyanin,
42Cgrowth,
Cetrimide tests
FIG. 1. Manual methodidentificationprocedure. (a) Enterobacteriaceaetests:Triple sugariron,
Christen-sen ureaagar; lysineandornithinedecarboxylase; argininedihydrolase; indole;Simmoncitrate; motility
agar; deoxyribonuclease; phenylalanine deaminase; arabinose, inositol, and adonitolfermentation. (b) If
Acinetobacter,Flavobacterium,etc. aresuspected,useprocedureofindividual choice.
prepared in the AMS procedure. Both blood agar
plates and MacConkey agar plates were used for
countsofbacteria;only blood agarplateswereused for the yeasts.
(ii)Mixed culturestudy.One culturewasselected
from each of the groupsorspeciesoforganismslisted
inTable 1.Ail possible combinations oftwocultures
wereused(e.g.,cultures A andB, A andC,AandD,
Band C, B andD, etc.), and each combinationwas
tested in fourdifferent ratios of the two strains
in-volved. Forexample,P.aeruginosa and E. coliwere
tested in ratios of
107:10i,
106:104, 105:107,and 104:106,respectively. The identification procedures via the
AMS and manual methodswerethe same as
previ-ously described. Yeasts were not employed in this
phase of thestudy.
Datahandling. Alldata,includingtapesand
hand-writtenreports fromtheAMS, manual, and routine
methods,wereforwardedtothe AMS computer
facil-ity of the McDonnell-Douglas Astronautics, East
Corp.Bothindividuallaboratoryandsummary results
werecomplied and returnedtothecollaborating
lab-oratories. The various authors also compiled some
data. In thisstudy,sensitivitywasdefinedasthe ratio
oftruepositive reportstothesumofthetruepositive
andfalsenegative reports,i.e.,theabiityofthe AMS
todetectaparticular organismwhen itwasshownto
be presentbyconventional methods. Specificitywas
definedasthe ratio ofthetruenegativereportstothe
sumof the truenegativeand thefalsepositive reports,
i.e.,theabiityof theAMSnot todetectanorganism when it was notdetected by conventional methods.
Reliabiitywasdefinedasthe ratioofthe truepositive
reports to the sum of the true positive and false
positive reports, i.e., the confidence that could be
placedinanAMSreportthat anorganismwaspresent.
RESULTS
Pureculturestudy.Thesensitivity,
specific-ity, andreliability of the AMS indetectingand
identifyingnine groups oforganisms commonly
found in urineareshown in Table1. The basis for this tabulation wasthe results obtained by
the manualmethod. Thesensitivityof theAMS ranged from 83.3% for Serratia to 98.9% for E.
coli, with a mean of 92.8% and a median of 95.1%. Both the positive control and enumera-tion chambers of the Identi-Pak were affected by testsinwhichvery small populations of or-ganisms (ca. 103/ml) at levels well below the thresholds set for detection and identification wereused.
Specificity values ranged from 98.1% for E. coli to 100% for group D streptococci, with a mean of 99.4% and a median of 99.7%. The positive control specificity is not a significant
figure because of the very small number of "true"negativeresults. The enumeration speci-ficityvalue (83.4%) indicates atendency ofthe AMS to report higher counts on about 17% of thespecimenswith bacterialpopulations evalu-atedmanuallyatlevelsbelow70,000/ml.
Thepercentreliability calculations reflectthe
degree ofconfidence to be placed in apositive AMSresult.Reliabilitypercentages ranged from a low of 71% for E. coli to 100% for group D
streptococci,with a mean of 92.1% and a median of95.4%. The low value for E.coli resulted from anappreciablenumber of Identi-Pak chambers
permitting other organisms tested, not E. coli, to attain detectable cell densities. This same
tendencywasnotedlater in the second phase of the study, in which mixed cultures were used. In addition to testing the AMS with pure cultures of the nine groups or species listed in
(cocci)
Catalase
~1
Bileesculin,
6.5% NaCI broth, CAMPtests
t
Coaugulase
VOL. 8,1978 659
on February 7, 2020 by guest
http://jcm.asm.org/
Table1,which thesystemwasdesignedto iden-tify, 224challengeswere madewithother
orga-nisms whichmay occurinurine (although
infre-quently), but which thesystemwasnotdesigned
toidentify. Thesewereusedtotestthelimits of differential capabiity ofthe AMS, and the
re-sults ofthese challengesare shown in Table 2.
With the exception ofcertain species of
Pseu-domonas, particularlyP.stutzeri, P.
maltophi-lia, P.diminuta,and P.
fluorescens,
mostof the strains tested grew in the Identi-Pak positivecontrolchamber andwere not identified.
How-ever, 45 of the 167 strains which grew were
named,althoughincorrectly. Mostofthe named
iresconsistedofgroupBstreptococciwhich
e incorrectly identifiedas group D
strepto-cocci. Somewhat
disturbing, although
notun-expected,
wasthemisidentification of10strains ofSalmonellaasE. coli.Theagreement ofidentification
by
theAMSwas measured on both an
inter-laboratory
andintra-laboratory
basis.Inter-laboratory
agree-mentisindicated in Table3,whichshows agree-ment ofresults from all sixcollaboratinglabo-ratories,
and offive of the sixlaboratories,
with 152cultures. The manual method resultsagreed
with thereferenceresults inall sixlaboratories
in89.5% of thetestsand infive of the six labo-ratories in 97.5% of thetests.Interestingly,AMS results
agreed slightly
better with thereferenceresults than with the manual method
results,
indicating
some faults in the latter. The AMSTABLE 1. Summaryofpureculturestudyresults:AMSsensitivity, specificityandreliability, basedon
manual methodresults
No.of AMSresults Sensitiv- Specific-
Reliabil-Organism chal- ity ity ity
lenges T+ F- F+ T- (%) (%) (%)
P.aeruginosa 125 119 6 22 1,819 95.2 98.8 84.4
Proteusspecies 116 104 12 3 1,886 89.7 99.8 97.2
C.freundii 95 82 13 3 1,904 86.3 99.8 96.5
Serratia 90 75 15 7 1,902 83.3 99.6 91.5
E. coli 90 88 1 36 1,811 98.9 98.1 71.0
Klebsiella/Enterobacter 110 103 7 5 1,880 93.6 99.7 95.4
Yeasts 348 331 17 4 1,597 95.1 99.8 98.8
GroupDStreptococcus 140 137 3 0 162 97.9 100.0 100.0
S. aureus 144 137 7 9 728 95.1 98.8 93.8
Positive control 1,540 1,406 134 0 13 91.3 100.0 100.0
Enumeration 1,540 1,121 419 81 407 72.8 83.4 93.3
aSensitivity= [(T+) x 100]/[(T+) + (F-)];specificity = [(T-) x
100]/[(T-)
+ (F+)];reliability
=[(T+)x 100]/[(T+)+ (F+)].
bT+=AMS +,
manual
method+; F-=AMS-,manual
method+; F+ =AMS+,manual
method
-;T-=AMS-,manualmethod -.
TABLE 2. AMS responsetochallengewithmiscellaneousorganisms
CorrectAMSresponse Incorrect AMS response
Organism No. Growthin Correct
tested positive Not iden- Incorrect genus, No
control
tified
genus wrong growthspecies
Pseudomonads, other than P. 79 36 30 4 2a 43
aeruginosa
Aeromonas 12 12 il 1 0 0
GroupBstreptococci 29 26 4 0 22b 3
a-Streptococci,notgroup D 30 25 24 1 0 5
S. epidermidis 17 14 12 0 2c 3
Salmonella 12 12 2 lod 0 0
Flavobacterium 12 12 10 2 0 0
Acinetobacter 22 19 18 1 0 3
Providencia il il il 0 0 0
Total 224 167 122 19 26 57
aBothidentified as P.aeruginosa.
b
Ail
identifiedasenterococci. C BothidentifiedasS. aureus.dAll
identified
as E.coli.on February 7, 2020 by guest
http://jcm.asm.org/
andmanual methods produced thesameresults, however, in83% of the participating laboratories (five of six) in93.4% of the challenges made and
in all six laboratories in 81.6% of these chal-lenges. Intralaboratory agreement of results, with 136challengesin eachlaboratory,is shown
inTable4.Obviously,the data were inexcellent
agreement inthevariouslaboratories, and there
is nosignificantvariation in theseresultsamong thelaboratories.
Table 5 presents the sensitivity, specificity,
andreliability data for each laboratory and for
each groupoforganismstested in the pure
cul-turestudy. Thegreatestvariation insensitivity
occurred withC.freundii and Serratia cultures, and itappearsthatrelativelypoorperformance
in two laboratories reduced the overall values
obtained (see Table1).Specificitywasextremely high and uniforminall laboratories. Reliability
wasthe mostvariable oftheseparameters,
par-ticularly
withP. aeruginosa and E. coli. Two laboratories obtained low values forP.aerugi-nosa,thusreducing the overallreliability from
over94%to84.4%(Table1).Similarly,two very low values were obtained for E. coli, with a
resultant low overall reliability (Table1). How-ever,reliabilityof apositiveE. coliresultwas a
common problem among all six laboratories, withnoneof themachieving higher than 83.3%. Therewas noindication fromthe sensitivity or
reliability data thatany onelaboratory
consist-ently performed less well than did other labo-ratories.
The three Identi-Pakswhich wereinoculated witheveryfifthspecimen were used todefmie,to someextent, the lower limits of detection and identification capabilities of the AMS. Some
indicationof this isshown inTable6,where the
AMS results are tabulated in relation to cell
population and
organism.
Certainorganisms,
such as P. aeruginosa, S. auréus, and yeast, were often notdetected at lowcell
concentra-tions
(103/ml), although others,
suchasE. coli andgroup Dstreptococci,
werenotonly
detected butwerealsoidentified
atsuchpopulations.
Atcell populations generally considered to be
in-TABLE 3. Interlaboratory agreement of identification (152 challenges)
Agreement byall Agreement by 5 Comparison 6laboratories of 6 laboratories
No. % No. %
Manualvsrefer- 136 89.5 148 97.4
ence
AMS vs refer- 127 83.6 145 95.4
ence
AMSvsmanual 124 81.6 142 93.4
TABLE 4. Intralaboratory reproducibilityofAMS
results (136 challenges)
Labora-
Labora-
AS AMS No. No.vs
AMS
Positive false false tory refer- manal growth posi-nega-ence
tive
tive
1 124 124 132 5 4
2 127 127 131 2 3
3 124 123 131 6 5
4 128 127 128 5 2
5 128 129 134 0 6
6 122 121 125 1 3
dicative of
probable urinary
tractinfection,
i.e., between 104 and105/ml,
the first threeorga-nisms mentioned abovewerestillnotaccurately identifiedin most tests. The variationobserved
with yeasts was partly due to the variety of
genera andspecies tested, some of which grew
faster than others. In 13 tests with yeast in
populations of 104 to 9 x 104/ml, no growth occurredinthepositive controlchamberofthe
Identi-Pak. In contrast, normally fast-growing
organisms,
such as E. coli, were occasionally detected and identified at populations of less than103/ml
(data notshown).Mixed culture
study.
Theunknowns distrib-uted for thisphase
of the studywere mixedineachcollaborating laboratory accordingto a
pro-tocol which resultedinall culturesbeing tested inpairs and eachpair of cultures being testedin
fourdifferent ratios oforganisms. The detection and identification
capabilities
ofthe AMS for these mixtures are shown in Table 7, which iscomparable to Table 1. The lowest values for both sensitivity and
specificity
were observed for E. coli; the highest values were observed withC.freundii.The meansensitivity for theseorganisms,
evenin the presence ofotherorga-nisms, was96.3%, with a median of 98.3%.The
specificity
ranged
from 96.6% for E.colito100% forC.freundii
and theKlebsiella/Enterobacter species, witha mean of 99.3% andamedian of 99.8%.Another point of investigation in this mixed
culturestudywastheeffect, ifany, ofthe pres-enceofone
particular
organismontheability
of the AMStodetect anotherorganismwhen pres-entsimultaneously; Tables8 to 10contain thesedata. In Table 8, the false positive and false
negativeASM results are tabulated withrespect tobothlaboratory and
organism.
It showsthat E.coli was detected in 22 instances when it was notpresent (false positive),and in 12 instances it was not detected when it was present (false negative),asdeterminedbythe manual method. Thirteen of the22false positiveE. coliresults werefrom onelaboratory-a variation of aboutVOL. 8,1978
on February 7, 2020 by guest
http://jcm.asm.org/
TABLE 5. Pureculturestudy: sensitivity, specificity, and reliability results by organism andlaboratorya
P. aeru- Pro- C. Serra- E. Klebsiella Y Group D S. Determination ginosa teus freundii tia
coli
Enterobacter east Strepto- aureuscoccus Sensitivity
Laboratory1 2 3 4 5 6 Specificity
Laboratory 1 2 3 4 5 6 Reliabiity
Laboratory1 2 3 4 5 6
82.6 94.7 83.3 87.5 100 100
100 89.4 93.7 87.5 100 100
95.2 90.0 84.2 88.2 100 94.7
100 88.2 100 84.6 94.1 89.4
100 80.9 83.3 66.6 94.1 100
94.7 95.0 75.0 90.0 100 83.3
99.3 100 99.7 99.7 97.0 100
99.6 100 100 99.3 99.0 100
97.2 100 100 99.7 98.1 100
97.0 99.6 100 99.6 98.0 100
99.3 99.3 100 99.6 98.7 98.7
100 100 99.3 99.6 97.3 99.6
90.4 100 90.9 93.3 54.5 100
95.0 100 100 87.5 83.3 100
68.9 100 100 93.7 71.4 100
70.3 93.7 100 91.6 76.1 100
92.3 89.4 100 92.3 80.0 80.0
100 100 85.7 90.0 63.6 95.2
96.4 100 96.9
93.3 100 88.8
93.3 94.1 84.8
96.6 100 100
96.5 96.2 100
94.3 100 100
99.6 100
100 100
100
99.5 100 99.6 99.6 100
98.2 100
100 100
100 100
98.3 100 98.2 100 98.0 100
100 96.9 98.4 99.2 99.1 100
94.1 72.7 93.3 96.1 96.7 100
aDefinitions forsensitivity,specificity, andreliabilityaregiven in thetextandthe footnotesforTable1.
TABLE 6. AMSresponse todifferent bacterialpopulations
Rangeofcellpopulations
103/ml 104/ml
Organism
No Positive
Positive
No PositivePositive
growth control cation growth control cationP.aeruginosa il 13 2 1 15 15
Proteusspecies 0 8 5 0 5 8
C.freundii 0 8 6 0 7 9
E.coli 0 0 il 0 0 7
Klebsiella/Enterobacter 0 8 6 0 2 14
Serratia 0 14 1 0 8 9
Yeast 15 8 2 13 15 21
GroupDStreptococcus 1 5 18 0 0 24
S.aureus il 15 0 3 il 12
TABLE 7. Summaryof mixedculturestudyresults:AMSsensitivity,specificity,andreliability, basedon
manualmethodresults
AMS
resuita
Sensitivity
Specificity
Reliability
Organism Challenges F- F+ T- ()(%%
T+ F- F+ T-(%()
P.aeruginosa 119 111 8 1 639 93.2 99.8 99.1
Proteusspecies 129 127 2 2 648 98.4 99.6 98.4
C.freundii 122 122 0 0 655 100.0 100.0 100.0
Serratia 132 125 7 9 646 94.6 98.6 93.2
E.coli 117 105 12 22 629 89.7 96.6 82.6
Klebsiella/Enterobacter 221 215 6 0 500 97.2 100.0 100.0
GroupDStreptococcus 115 114 1 1 649 99.1 99.8 99.1
S.aureus 119 117 2 2 643 98.3 99.6 98.3
Positive control 841 840 1 0 0 99.8 100.0
Enumeration 841 841 0 1 1 100.0 50.0 99.8
aSee
footnote
b,Table1.on February 7, 2020 by guest
http://jcm.asm.org/
EVALUATION OF AUTOMICROBIC SYSTEM 663
seventimes the averageoccurrence of such
re-ports in the other laboratories. There was no
unusual distribution of false negativereports of
E.coliamongthe sixparticipating laboratories. Table-9presentsinmoredetail the conditions
under which the falsepositive AMS resultswere
obtained. It isapparentthateveryfalsereport
of E. coli occurred when Serratiawereactually
present.Similarly, six of the nine falsereportsof
Serratia occurred whenagroupDstreptococcus
waspresent.Theprecisereasonsfor these
asso-ciationsarenotknown, butone reasonmight be
that certain combinations oforganisms act
syn-ergistically to overcome the specific inhibitors
presentin the E. coliorSerratiagrowth
cham-bers of the Identi-Pak.
The
félse
negative reports observed in the mixed culture study are listed in Table 10. E. coliwasmissedmostoften(12 times), andmostof these misses occurred when therewasgrowth
in the Klebsiella/Enterobacterchamber of the
Identi-Pak. C. freundii was most frequently
presentwhen the AMSfailedtodetect another
organism, and most ofthese associations were
with falsenegative Serratia and E. coli results. This couldperhaps be attributed toan
antago-nistic relationship between certain combinations oforganisms, but there isnoadditional evidence
tosupportthis. Even if thisistrue, areciprocal
relationship between C.freundii and Serratiaor
E. coli was not observed, because C. freundii
wasnotmissed during thisphase of the study.
DISCUSSION
Theconcept ofanautomated and
computer-ized system for enumeration, detection, and
identification oforganisms in urine, in the pres-enceof otherorganisms, isadramaticdeparture
from theconceptsinvolved in ingrained,
conven-tional pureculturetechniques. The AMS relies ontheprinciple of utilizing specific compounds
that selectively favor the growth and
metabo-TABLE 8. Mixed culturestudy: AMSfalsepositive andfaIsenegative results
Falsepositive in laboratory no.: Falsenegativeinlaboratory no.: Organism
1 2 3 4 5 6 Total 1 2 3 4 5 6 Total
P.aeruginosa 0 0 0 1 0 0 1 1 5 0 0 0 2 8
Proteusspecies 0 0 0 1 1 0 2 0 0 0 0 0 2 2
C.freundii O O O O O O O O O O O O O O
Serratia 0 4 1 1 2 1 9 1 1 1 1 2 1 7
E.coli 2 0 13 4 1 2 22 3 3 0 0 3 3 12
Klebsiella/Enterobacter O O O O O O 0 3 2 0 0 1 0 6
Group DStreptococcus 0 0 0 1 0 0 1 0 0 0 0 0 1 1
S.aureus 0 0 0 1 0 1 2 2 0 0 0 0 0 2
TABLE 9. Mixed culturestudy: AMSfalsepositive results
Pseudo- Pro- C. Serra- E. Kleb- Entero- Group D S.
Organims present mnonas teus freundii tia
coli
siella bacter co aureusC.freundii/Serratia la
S.aureus/Serratia 8
Klebsiella/Serratia 1
Enterobacter/Serratia 3
Proteus/Serratia 2
P.aeruginosa/Serratia 3
Group D streptococci/Ser- 4
ratia
P. aeruginosa/Group D 4
streptococci
Klebsiella/group D strepto- 2
cocci
C. freundii/group D strep- 2
tococci
P.aeruginosa/Klebsiella 1 i
Proteus/Enterobacter 1 1
E.coli/S.aureus 1 i
GroupDstreptococci/S.au- 1
reus
aFigures indicate number ofoccurrencesof a
false
positive report when the indicated pairs of organisms wereactuallypresent.
VOL. 8,1978
on February 7, 2020 by guest
http://jcm.asm.org/
TABLE 10. Mixed culture study:AMSfalse negative results Organism missed Organiam
alsopresent Pseudo- Pro- C. Serra- E. Kleb- Entero- Group D S. monas teus freundii tia coli siella bacter co aureus
Pseudomonas 1 1
Proteus 1 2
C.freundii 1 5 4 2
Serratia 2 1
E.coli
Klebsiella 2 1 4
Enterobacter 1 1 4
Group D streptococci 1
S. aureus 1
lism ofspecificgroups orspecies ofbacteria, and
combines this withautomatedand computerized
instrumentation.
Thesystemisdesignedtobe usedfor screen-ing urine specimensofthetypes mostcommonly encountered, in which infection isusually
asso-ciated with bacterialcounts of105 or more
or-ganismsperml. Specimens collected by cathe-terization, suprapubic aspiration, or from pa-tients undergoing cystoscopy, in which low
countsmay beclinically significant, should not
be testedintheAMS, butshouldreceivespecial handling.
Publisheddevelopmental studiesonthis sys-tem
(2)
wereencouraginginthat morethan 92%correlation
was obtained with conventional manual results on about 3,400simulated urinespecimens, and, with theexception ofP.
aerugi-nosa, more than 93% correlation was obtained with 1,500 clinical urine
specimens.
Problemswereencountered, however, with both false
pos-itive andfalsenegativeresults andwith various
technological
difficulties.Alaterstudyof theAMS,inwhich a prepro-totype instrumentand both simulatedand
clin-icalspecimenswereused(1),showedhigh levels of correlation
(>90%)
for mostorganisms,
butalso revealed weaknesses in the identification capabilities oftheAMS forP.
aeruginosa,
group Dstreptococci,
andS. aureusin clinical speci-mens.Reproducibility
experimentsinthis samestudy, with twomachines,suggested thatmore
variation was encountered in conventional
man-ual methodsthan in the AMS
procedure.
The multi-laboratory collaborative study re-portedhere hasevaluatedthesensitivity, speci-ficity,andreproducibilityof the AMS. The tech-niques employed,i.e.,
use oflyophiized
purecultures, conventionalmediafromasingle
sup-plier, anda manual
method,
were designed tominimize variation in
experimental
conditions.Nevertheless, some variations did occur. Urine Identi-Paks from the same batch could not be
provided to allparticipants forthe duration of
the study; thus, several production batches of
Identi-Paks were employed. Furthermore,
dur-ing the studysomeof thespecificchamber media
used in the Identi-Paks werefound tobe defi-cient.They werelater modified, thus invalidat-ingresults previously obtained. Additionally, it wasrecognized that"humanerror" could not be
totally eliminated from the manual manipula-tionsinvolvedinsettingupthe varioustests, as
partially
showninTable8.Finally,asmentionedpreviously, the manual method used in all six
laboratories was neither an exhaustive nor an
extremely abbreviated identification system; it
was acompromise between these extremes. The method was reasonably accurate and reflected thelevel of recognition possessed by the
instru-ment.Itstill permittedsomesubjective conclu-sions, andthus somevariations occurred.These variationsundoubtedly hadsomeadverse
influ-ence onthe
experimental
datapresented. Con-versely, theymore closely approximatedcondi-tions which might be encountered ifthe AMS
wereused in a number ofwidely scattered lab-oratories.
We consider the results of this study to be
satisfactorydespite themanyvariablesinherent in it. With pure
cultures,
five of the ninegroup/speciestested showed
sensitivity
ofover 95%, compared with the manual method; onewas93.6%, two were between 85 and 90%, and
only one (Serratia) was between 80 and 85%. Thespecificityof theAMSwasextremely high
forall ofthe
organisms tested, being
atleast 98% orbetter. Thereliabiity
of the AMS showedawider range than did either the
sensitivity
or specificity. The data(Table
1) show that anAMS result of E. coli wasrecorded 124
times,
yet E. coliwereactually
presentonly
88times. This was one of the most disturbingproblems
encountered,
because of both thefrequency
and thesignificance
of E.coli inurinary
tract infec-tions. Thereliabiity
value observed for E. colion February 7, 2020 by guest
http://jcm.asm.org/
was, however, not representative of values ob-tainedfor other organisms.Onlyoneothervalue, that for P. aeruginosa (84.4%),wasbelow 90%.
The limitsof theAMS fordetectionand
iden-tification wereseverelytestedbyincluding var-iousorganisms which are seen only rarely,but
still may occur, in urine specimens (Table 2).
The slow-growing pseudomonads were not de-tected for the most part, whereas most of the
otherorganisms testedweredetected butwere notidentified. The misidentifications ofgroup B
streptococciasgroup Dstreptococci and of Sal-monella as E. coliillustrate limitations in the capabilities of the AMS Identi-Pakto differen-tiateselectivelyamongrelatedorganisms. These limitations (misidentifications) could havesome serious therapeutic consequences, but they should not be overemphasized. Probably the
most commonly encountered problem here would be misidentification ofgroup B
strepto-cocci, becausetheyare not uncommoninurine from females. The reason for this error is not
clear, because
hydrolysis
of esculin andtolerancetohighsaltconcentrationsarerequired for AMS identification of group D streptococci. Most group Bstreptococci will tolerate6.5%salt,but
they do not hydrolyze esculin. Possibly there was sufficient change in turbidity, because of growth of group B streptococci to register a
positive result,but thisdoesnotaccountforthe hydrolysis of esculin. Certainlyconditions exis-tentin a microchamber aredifferent fromthose inconventionaltesttubes,and this mayalso be
a factor. The significance of these data lies in
therecognition that such discrepanciesmay oc-cur, andso
microbiologists
andclinicians mustview themintheproperperspective.
Theproblems mentionedinthetwopreceding
paragraphs
regarding specificity, sensitivity, and reliabilityare influencednotonly by the selec-tive compounds employed inthe specific iden-tification chambers, but also by the threshold levels set for each chamber. These thresholdsare
adjustable,
andvaryfrom chamberto cham-ber, so that only the desiredorganisms
be de-tected. It is appropriate to mention that thisinvestigationwas
designed
toaid infme-tuningthesystem before its use in the clinicalsetting and that future adjustment of both
selective
compounds and identification chamber thresh-oldlevelsmay modify, if not alleviate, some of theproblemsencountered.
Agreement of identification of AMS results was considered tobe very good, both among the sixcollaboratinglaboratories and within each of theselaboratories. Table 3 shows that the man-ual and reference results agreed among all six laboratories in 136of 152 challenges and in five
of sixlaboratoriesin 148of152
challenges.
Thefiguresused for the "AMSversusmanual" com-parison were 124/152
(for
all sixlaboratories)
and 142/152 (forfiveofsixlaboratories). How-ever, a morerealisticmeasuremight betheuse ofthe "manualversusreference"figures
asthedenominators in these
calculations,
thus chang-ing the percents to 91.2 and 95.9, respectively. Thesemoreaccurately reflect theagreementof theAMS and the manual methodresults. Some of thevariationinsensitivity notedamong indi-vidual laboratories (Table5) was probably the resultof batch-to-batch variationinurine Identi-Paks.Thisstudyconfirmed, in six laboratories, the
report by Sonnenwirth (1) that the AMS was
capable ofdetectingandof sometimes identify-ingspecific organismsatlevels of
103
to9 x103
CFU/ml.Thiswas
particularly
notedwith rela-tivelyfast-growing
organisms. Only with slower growingorganisms, suchasStaphylococcus au-reus, yeasts, andpseudomonads,
didsignificant
numbers of tests result in no growth in the Identi-Pakatsuchlow cell concentrations.
Atconcentrations of104to 9 x
104
CFU/m1,
theAMSnearly alwaysdetected the presenceof
organisms
andusuallycorrectly identified them. Onlyyeast culturesstillwere notdetectedin asignificant
numberoftestswhenpresent at thisconcentration. Thus, theabilityof the AMS to
recognizethepresence of
microorganisms
insim-ulated urine specimens at levels below those
considered indicative ofurinary tract infection
wasestablished.
Itshouldbe reemphasizedthat in this study the enumeration threshold was set at 70,000
CFU/mi.
Itmight
beargued that the detection and identification ofsmall cell populations(<104 CFU/ml) bythe AMS isproblematical from aclinical standpoint,because suchpopulationsare
usually regarded as
insignificant.
Thus, detec-tion of low populations by the AMS might befalsely interpreted as representing significant
bacteriuria. This
possibility
is not consideredvery likely, however, because theAMSwill re-port whether counts are above or below70,000
CFU/mi.
Anotherpointofconsiderationis thatpatients withurinarytractinfection oftenhave
extremely
high-count bacteriuria(107
to109
CFU/ml). Inthe present study, populations of
cells in seeded urines rarely exceeded
107
CFU/ml,
so the question might arise as to whether thespecificity
of the AMS would re-main as high withsignificantly
higher count urines. Thiscannotbe answeredonthe basis of results of thisstudy; the answerwill
depend on results ofclinical studies with the AMS.We wish toemphasizethat the AMS
enumer-VOL. 8,1978
on February 7, 2020 by guest
http://jcm.asm.org/
ation and identificationreports on mixed speci-mens do not reflect the proportions in which various organisms are present, and thus give
equal weight to all ofthem, whereas only one mayactuallybepresent in significantnumbers.
Some information may begainedfrom
compar-ing thetimeperiods required for identifyingtwo
organisms,but this is nottotally reliablebecause of the varying growth rates of different orga-nisms in their respective identification cham-bers.
The mixed-culture phase of this study was
intended to explore the effect ofoneorganism
onthe AMS'sabilitytodetect anotherorganism, when bothwerepresentsimultaneously. Onlya
lowlevel of falsereports wereencountered:4.4% false positive and 4.5% false negative. These values compare to 4.4%and4.0%, respectively, in thepureculturestudy. The onlyassociations thatcouldbemaderelatingtothe occurrenceof false results were as follows: (i) a correlation
between false positive reports of E. coli when
Serratiawerealsopresent (Table 9); (ii)a ten-dency forSerratiatobe
reported
when group Denterococci were present
(Table
9); and (iii) apossible, but not clear-cut failure to identify SerratiawhenC.freundiiwasalsopresent
(Ta-ble 10). Thefirstoftheseassociationsseemsthe
clearest, but, because of the small amount of data available, confirmationby testing of fresh urinespecimensis
highly
desirable.Thepresentstudywasdesignedtotestvarious capabilities of the AMS, by usingpure cultures
and simulated mixed cultures of
organisms
found in urinary tract infections. No attempt wasmade to simulate the distributionof
orga-nisms which occur most frequently in urine spec-imens ofpatients with urinary tract infections. Thus,thisstudy is weighted towardsorganisms
lesscommonlyencountered in the usualclinical
practice.Nevertheless, such a study is necessary toproperly definesomeofthelimitsof perform-anceofany device intended for clinical applica-tion.Stockcultures ofsomeEnterobacteriaceae
may notbeasmetabolicallyactive as fresh clin-ical isolates, but even the largest clinical labo-ratories do not encounter sufficient numbers of some organisms to permit a complete study of thistaxonomic family. Thus, this study has
at-tempted to provide a rigid test of the AMS,
undercontrolled conditions,and has shown that
the AMS appears to have potential for use in
clinical microbiology. Clinical studies now in progress shouldfocusonthispotentialand fur-ther define the AMS's capabilities and
limita-tionsinrelationtospecimens frompatients
sus-pectedofhavingurinarytractinfections. In
ad-dition, ourearlyexperiences withdefectivelots
of Identi-Paks show that the manufacturer is
obligated to maintain a high level of quality controlinthelarge-volume production of these products, and the user is obligated to perform his own quality control tests with known
cul-tures atappropriate intervals.
LITERATURE CITED
1. Aldridge,C.,P. W.Jones,S.Gibson,J.Lanham,M. Meyer, R. Vannest, and R. Charles.1977.Automated microbiologicaldetection/identification system. J. Clin. Microbiol. 6:406-413.
2. Sonnenwirth, A. C.1977.Preprototypeof anautomated microbial detectionandidentification system:a devel-opmental investigation. J. Clin.Microbiol. 6:400-405.