A Multi-Laboratory Study of the BIOMIC Automated Well Reading Instrument versus

Transcription

1 JCM Accepts, published online ahead of print on 13 March 2013 J. Clin. Microbiol. doi: /jcm Copyright 2013, American Society for Microbiology. All Rights Reserved A Multi-Laboratory Study of the BIOMIC Automated Well Reading Instrument versus MicroScan WalkAway in the Reading of MicroScan Antimicrobial Susceptibility and Identification Panels Robert C. Fader * 1, Emily Weaver 1, Rhonda Fossett 2, Michele Toyras 3, John Vanderlaan 4, David Gibbs 5, Andrew Wang 5, Nikolaus Thierjung 5 1 Scott & White Healthcare, Temple TX, 2 Texas Health Arlington Memorial Hospital, Arlington TX, 3 Marquette General Hospital, Marquette MI, 4 Grand River Hospital, Kitchener, Canada, 5 Giles Scientific Inc, Santa Barbara CA. Corresponding author *Robert C. Fader, Ph.D. Scott & White Healthcare 2401 South 31 st Street Temple, TX rfader@sw.org Running Title: BIOMIC /MicroScan WalkAway comparison 1

2 24 ABSTRACT This study compared the BIOMIC automated well reader results to the MicroScan WalkAway results for reading MicroScan antimicrobial susceptibility and identification panels at four different sites. Routine fresh clinical isolates and quality control (QC) organisms were tested at each study site. A total of 46,176 MicroScan panel drug-organism combinations were read. The BIOMIC Category Agreement for 3117 Gram negative bacteria was 98.4%, with 1.4% minor and 0.2% major discrepancies. The BIOMIC Category Agreement for 5233 Gram positive bacteria was 98.7%, with 0.9% minor, 0.3% major and 0.1% very major errors. Essential agreement, defined as BIOMIC results that were within ± one two-fold dilution of the MicroScan results, was 99.3% for Gram negative and 98.3% for Gram positive bacteria. BIOMIC reading of MicroScan identification panels provided an overall agreement (first and second choice organism match) of 99.5% with 846 Gram negative isolates and 99.5% with 430 Gram positive isolates. These results suggest that the BIOMIC automated reader can provide accurate reading of MicroScan panels and has the capability of a visual panel read for manual adjustment of results. 2

3 40 Introduction The BIOMIC automated well reading instrument (Giles Scientific, Inc., Santa Barbara, CA) was originally developed as a new method for automated reading of disk diffusion susceptibility plates (1,2) with extrapolation to an estimated minimum inhibitory concentration through regression analysis. It has since been adapted for automated reading of yeast disk diffusion tests (3, 4) and can also be used to interpret colonies on chromogenic agars (5). The instrument has compared favorably with MIC determinations and category interpretations with the Vitek system and with the gradient diffusion (E test) method (6, 7). The versatility of the BIOMIC could be further expanded if 96-well microtiter plates for identification (ID) and antimicrobial susceptibility testing (AST) could also be accurately read in the instrument. The purpose of this study was to evaluate the accuracy of the BIOMIC automated well reader in interpreting the results from a variety of AST/ID panels read initially on the MicroScan WalkAway (Siemens, Sacramento, CA) instrument. Four different health care institutions participated by testing both clinical isolates and quality control strains. 3

4 56 Methods and Materials 57 Participating Laboratories The following laboratories participated in the study: Scott & White Memorial Hospital, Temple, TX; Texas Health Arlington Memorial Hospital, Arlington, TX; Marquette General Hospital, Marquette, MI; and Grand River Hospital, Kitchener, Canada. Test Organisms. The participating laboratories tested MicroScan panels with a total of 1911 clinical isolates in antimicrobial susceptibility tests, 1276 clinical isolates in identification tests and 309 tests with QC organisms. All clinical isolate test and quality control panels were read as part of the normal laboratory routine with sequential clinical isolates to eliminate an organism selection bias. Study sites and isolates represented a broad clinical and geographic range. Organism sources consisted of a broad spectrum of clinical human specimens and panel manufacturer recommended QC strains. Test Panels Commercially manufactured MicroScan test panels and materials were provided by each study site as part of their routine test procedures. Each study site employed the MicroScan Lab Pro Version 3.01 for results interpretation. Standard panel manufacturer guidelines were followed for all AST and ID methods in this study. Breakpoint panels have antibiotics with less than 4 dilutions with antimicrobial concentrations based on the category breakpoint values for susceptible and resistant established by the Clinical and Laboratory Standards Institute 4

5 (CLSI)(8). MIC panels have antibiotics with 4 dilutions although some antibiotics may have fewer than 4 dilutions on the panel. All panels required overnight incubation (minimum of 16 hours) All MicroScan identifications were part of combination ID/AST panels. They included Neg Breakpoint Combo Type 30, Neg Urine Combo Type 35, Neg Urine Combo Type 45, Pos Breakpoint Combo Type 20, Pos Breakpoint Combo Type 23 and Pos Combo Type 29. Comparative Reader Method All MicroScan test panels were read in each laboratory with their routine manufacturer recommended equipment and methods. Each panel was first incubated and read on the MicroScan Walkaway system. Following completion of reading on the Walkaway, the panel was then removed from the Walkaway and read on the BIOMIC System within 1 hour. Photographs of the panels and printed test results were provided to Giles Scientific Inc. by , fax or mail, according to the Study Protocol. No patient demographics were transmitted to Giles Scientific Inc. BIOMIC automated well reader A standard commercially available BIOMIC V3 microbiology system with the BIOMIC 2011 Software was provided to each participating site by Giles Scientific Inc. This system consisted of the standard BIOMIC V3 reader cabinet containing LED visible with a high resolution color digital camera, a personal computer with Windows operating system, and BIOMIC 2010 clinical microbiology software. Additionally, each system included the 5

6 96 97 BIOMIC automated well reader software module. AST and ID test panels were read on the BIOMIC after being read on the MicroScan reader employed routinely in that laboratory 98 Quality Control QC testing of all panels was performed using MicroScan recommended ATCC strains for each specific panel type. Data Analysis All test results were read, interpreted and recorded in the BIOMIC software and included minimum inhibitory concentration (MIC), Susceptible/Intermediate/Resistant (SIR) category calls, genus/species identification, and panel image. This information was transferred to Giles Scientific for analysis by secure internet upload to the Giles Scientific website. Printed paper copies of MicroScan test results were provided to Giles Scientific and were manually entered into a master BIOMIC database. Data was verified by Giles Scientific by using the Data Verification Procedure (9) described as follows: 1. MicroScan MIC results were recorded manually into the BIOMIC software, which uses Microsoft Access database format. 2. All test results were divided into two groups: one containing exact MIC matches, the other containing MIC discrepancies. 3. For the test results group with MIC discrepancies, each test result was verified independently by two persons to check for possible transcription errors. 6

7 For the test results group with exact MIC match between MicroScan and BIOMIC, we adopted ANSI/ASQC Z1.4 sampling plans (2) for quality control. Given the lot size range of 35, ,000, sampling size code N is used under general inspection. It requires a sample size of 800, acceptance on 0 nonconformity and rejection on 1 nonconformity. This corresponds to AQL (Acceptable Quality Level) of 0.015% and LTPD (Lot Tolerance Percent Defective) of 0.37%. It gives a 95% confidence level that the data have 0.015% or below defective human errors (7 or less bad test results). For the 5% tail probabilities, the upper bound for the possible defective data is 0.37% or 174 test results. BIOMIC Expert System The BIOMIC Expert System monitors test results that are unusual or unlikely and detects resistance by using combinations of test results from several antibiotics. The Expert System applies CLSI recommendations (8) for category call adjustments based on detection of certain resistance mechanisms such as oxacillin resistance and detection of ESBL mediated-resistance. The Expert System may present a warning message onscreen to the technologist after a test is read, change an interpretation (S to R) on the results, or suppress a drug result on the printout. Category and Essential Agreement Definitions Category agreement means the category call of susceptible, intermediate, or resistant as determined by CLSI guidelines (8) was achieved between both instruments. Essential agreement means that BIOMIC results were within ± one two-fold dilution of the MicroScan results. Category Call Error Definitions 7

8 Minor errors occurred when either system called an antibiotic interpretation as Intermediate but the other system called it Resistant or Susceptible. Major errors (MEs) were defined as a MicroScan category call of Susceptible when the BIOMIC category call was Resistant. Very major errors (VMEs) occurred when the BIOMIC system resulted in a Susceptible category call but the MicroScan system interpretation was Resistant. Downloaded from on January 13, 2019 by guest 8

9 Results A total of 46,176 MicroScan panel antibiotic-organism combinations were read (Table 1). For MIC panels and antibiotics with 4 or more dilutions, the BIOMIC Category Agreement for 3117 Gram negative bacteria was 98.4% with 1.4% minor, 0.2% MEs and no VMEs (Table 1A). The BIOMIC Essential Agreement was 99.3%. For 5,233 Gram positive organisms, the BIOMIC Category agreement was 98.7% with 0.9% minor, 0.3% MEs and 0.1% VMEs. The BIOMIC Essential Agreement for Gram positive bacteria was 98.3%. Data analysis by antibiotic is listed in Table 2. Only 5 VMEs were noted: 2 with erythromycin and 1 each with ampicillin, ceftriaxone and penicillin. All VMEs occurred with Staphylococcus species. For MicroScan AST Breakpoint panels, the BIOMIC Category Agreement with 28,993 drug- Gram negative organism combinations was 98.2% with 1.4% minor, 0.3% MEs and 0.1% VMEs (Table 1B). BIOMIC Category Agreement with 8833 drug-gram positive organism combinations tested with MicroScan AST Breakpoint panel was 96.5% with 2.7% minor, 0.6% MEs and 0.2% VMEs. After BIOMIC Expert Rule application, this Category Agreement increased to 98.3% and the VMEs decreased from 22 to 8. The 19 VMEs for Gram negative organisms and the 22 VMEs for Gram positive organisms were distributed among the various antibiotics (Table 3). When VMEs were evaluated by organism identification (Table 4), the majority of VMEs in Gram negative bacteria occurred with Pseudomonas (11/19), while most VMEs in Gram positive bacteria were noted with Staphylococcus (20/22). 162 Organism Identification Comparison 9

10 A comparison of BIOMIC automated well reading of MicroScan ID panels provided an overall agreement (first and second choice organism match) of 99.5% with 846 Gram negative isolates and 99.5% with 430 Gram positive isolates. BIOMIC provided a first choice organism match in 98.1% of Gram negative isolates and 98.4% of Gram positive isolates (Table 5). Eight of the 17 discrepant identifications of Gram negative organisms occurred with Pseudomonas species while 3 occurred with the Proteus/Morganella group. All 7 of the Gram positive misidentifications occurred with Staphylococcus species. Intra-lab Instrument Reproducibility Instrument reproducibility was evaluated by reading MicroScan panels 3 successive times in the BIOMIC instrument. Three sites selected patient isolates while one site selected QC panels for reproducibility studies. The overall intra-laboratory reproducibility was 99.0%. Downloaded from on January 13, 2019 by guest 10

11 Discussion This study has examined the ability of the BIOMIC automated well reader to provide accurate reading of 96-well overnight MicroScan panels. The BIOMIC reading of MicroScan AST/ID panels correlated highly to the panel manufacturer s automated instrument reading as performed routinely in each participating laboratory. Minimal numbers of VMEs were noted in the evaluation. Of the VMEs that were noted with Gram positive bacteria in the analysis of breakpoint panels, 20 of 22 occurred with Staphylococcus. Of the 20 VMEs noted before BIOMIC Expert System application, nine were identified with cephalosporin antibiotics and two with amoxacillin-clavulanate whose SIR category call would normally be determined by the oxacillin result, which had 100% accuracy in the study. Consequently, after the application of the BIOMIC Expert System, the number of VMEs in Gram positive bacteria dropped to 6. No VME occurred with Gram negative bacteria and full MIC panels. With Gram negative breakpoint panels, Pseudomonas was responsible for 11 of 19 VMEs. The carbapenem antibiotics imipenem (3/1178) and meropenem (3/424) were the most common VMEs observed, while the remainder of the VMEs were distributed between six different genera. The BIOMIC interpretation of biochemical reactions for organism identification also resulted in a high degree of accuracy with over 98% of first choice and 99.5% either first choice or second choice identification with both Gram positive and Gram negative organisms. For Gram negative isolates, the majority of discrepancies occurred with Proteus/Morganella (3 of 68), 11

12 Acinetobacter (2 of 11) or Pseudomonas spp. (8 of 92). The BIOMIC identified three P. aeruginosa as Alcaligenes xylosoxidans although P. aeruginosa was the second choice organism. All discrepancies with Gram positive organisms occurred with Staphylococcus spp. Misidentification of S. aureus as a member of the coagulase negative Staphylococci occurred in 3 of 209 readings. One S. epidermidis was identified biochemically as S. aureus. The remainder were species discrepancies among the coagulase negative Staphylococci. A weakness of the study design was the inability to adjudicate discrepancies since the data was transferred to Giles Scientific Inc. in the form of photographs and MicroScan printouts for analysis. Another potential weakness of the study was the time difference from the final MicroScan reading to the time of the BIOMIC reading. However, all panels were read within 1 hour of being removed from the MicroScan WalkAway and all MicroScan ID/AST panels required full overnight incubation and were not rapid panels. An interesting capability of the BIOMIC is the ability of the microbiologist to see all test well results on a video screen, to make interpretive adjustments as needed, and to save test panel images for supervisory review at a later time. MicroScan automated readers (AutoScan and WalkAway) do not enable users to view the actual image of the panel on screen but only provides a graphic representation of the panel with positive and negative biochemical reactions and markers to indicate growth in antibiotic-containing wells. In the case of skipped wells or questions about biochemical test reactions, the MicroScan panel must to be removed from the 12

13 instrument for a visual examination. In contrast, the BIOMIC instrument allows the visual examination to be made in situ to enable manual adjustments of susceptibility results or biochemical reactions for identification In conclusion, the BIOMIC may be useful as a primary automated reader in laboratories where disk diffusion susceptibility tests and identification panels are currently read manually. In those laboratories, it would permit the use of 96-well MicroScan ID/AST panels for testing organisms such as the Non-Enterobacteriaceae where disk diffusion standards have not been established (8). It could also be used as a backup automated well reader in instances where the primary antimicrobial susceptibility testing instrument is incapacitated. One notable advantage of the BIOMIC system is the versatility it has to interpret different methodologies, whether they are different susceptibility methods or identification panels. This is an asset important in times when budget constraints could limit the ability of a laboratory to purchase different automation platforms. The instrument can also eliminate the factor of reader variability in manually determining the results of various testing methods but still allows for user control. 13

14 References Broekema NM, Van TT, Monson TA, Marshall SA, Warshauer, DM Comparison of cefoxitin and oxacillin disk diffusion methods for detection of mecamediated resistance in Staphylococcus aureus in a large-scale study. J. Clin. Microbiol. 47: Korgenski EK, Daly, JA Evaluation of the BIOMIC Video Reader System for determining interpretive categories of isolates on the basis of disk diffusion susceptibility Results. J. Clin. Microbiol. 36: Azevedo AC, Bizerra FC, da Matta DA, de Almeida LP, Rosas R, Colombo, AL In vitro susceptibility of a large collection of Candida strains against fluconazole and voriconazole by using the CLSI disk diffusion assay. Mycopathologia 171: Hazen KC, Baron EJ, Lopes Colombo A, Girmenia C, Sanchez-Sousa A, del Palacio A, de Bedout C, Gibbs, DL, The Global Antifungal Surveillance Group Comparison of the susceptibilities of Candida spp. to fluconazole and voriconazole in a 4-Year global evaluation using disk diffusion. J. Clin. Microbiol. 41: Baron EJ, D'Souza H, Qi Wang A, Gibbs DL Evaluation of the Biomic V3 Microbiology System for identification of selected species on BBL CHROMagar Orientation Agar and CHROMagar MRSA Medium. J. Clin. Microbiol. 46:

15 Berke I, Tierno Jr. P M Comparison of efficacy and cost-effectiveness of BIOMIC VIDEO and Vitek antimicrobial susceptibility test systems for use in the clinical microbiology laboratory. J. Clin. Microbiol. 34: Sautter R, DeWeese D Comparison of two gradient diffusion susceptibility methods, BIOMIC and Etest, for the determination of susceptibility on routine clinical isolates: 166 routine isolates, 7 species, and 4 drugs, 638 drug combinations. Abstract C336, American Society for Microbiology Annual Meeting. 8. Clinical Laboratory Standards Institute. Performance Standards for Antimicrobial Susceptibility Testing: Twentieth Informational Supplement. CLSI document M100-S21. Clinical Laboratory Standards Institute, 950 West Valley Road, Suite 2500, Wayne, PA, ANSI/ASQ Z Sampling Procedures and Tables for Inspection by Attributes, American National Standard Institute. Acknowledgment This study was supported by Giles Scientific Inc. Santa Barbara, CA 15

16 Table 1: Summary of BIOMIC Reading of MicroScan Panels Table 1A: MIC Well Readings Only (Drugs with 4 dilutions) (See Table 2) Total # of Drug/Organism Combinations Tested Essential Agreement Category Agreement Minor % Major % Very Major Gram Negative Bacteria % 98.4% % 7 0.2% - Gram Positive Bacteria % 98.7% % % 5 0.1% Table 1B: Breakpoint Readings Only (Drugs with < 4 dilutions) (See Table 3) Total # of Drug/Organism Combinations Tested Category Agreement without Expert System Category Agreement with Expert System Minor % Major % Very Major Gram Negative Bacteria % 98.1% % % % Gram Positive Bacteria % % % % 98.3% % % 8 0.1% % % 16

17 274 17

18 Table 2: BIOMIC Reading of MicroScan Panels - Essential Agreement and Category Agreement by Drug (MIC Well Readings Only a ) Gram Negative Bacteria Drug Total # of Drug/Organism Combinations Tested Essential Agreement b Category Agreement Minor % Major % Very Major Amikacin % 100.0% Ampicillin % 95.4% 6 3.4% 2 1.1% Cefazolin % 100.0% Cefepime % 100.0% Cefotaxime % 97.4% % Cefotetan % 98.0% 1 2.0% Ceftazidime % 98.9% 5 0.6% 5 0.6% Ceftriaxone % 98.9% 2 1.1% Cefuroxime-parenteral % 98.0% 1 2.0% Cephalothin % 94.1% 3 5.9% Gentamicin % 100.0% Imipenem % 100.0% Meropenem % 96.2% 2 3.8% Piperacillin % 94.2% 3 5.8% Piperacillin-tazobactam % 100.0% Ticarcillin-clavulanate % 94.2% 3 5.8% Tobramycin % 100.0% Total % 98.4% % 7 0.2% % 18

19 Table 2 continued Gram Positive Bacteria Drug Total # of Drug/Organism Combinations Tested Essential Agreement b Category Agreement Minor % Major % Very Major Amoxicillin-clavulanate % 100.0% Ampicillin % 99.3% 1 0.3% 1 0.3% Azithromycin % 87.5% % Cefaclor % 100.0% Cefazolin % 100.0% Cefepime % 100.0% Cefotaxime % 93.8% 1 6.3% Ceftriaxone % 92.0% % 1 0.5% Cefuroxime-parenteral % 100.0% Chloramphenicol % 100.0% Ciprofloxacin % 98.8% 1 1.3% Clindamycin % 99.4% 1 0.3% 1 0.3% Daptomycin % 98.6% 5 1.4% Erythromycin % 97.8% 7 1.6% 1 0.2% 2 0.4% Gatifloxacin % 99.6% 1 0.4% Gentamicin % 100.0% Imipenem % 97.7% 1 2.3% Levofloxacin % 99.6% 1 0.4% Linezolid % 99.5% 1 0.5% Meropenem % 100.0% % 19

20 275 Moxifloxacin % 98.7% 2 0.7% 2 0.7% Oxacillin % 99.7% 1 0.3% Penicillin G % 99.4% 3 0.5% 1 0.2% Piperacillin-tazobactam % 100.0% Quinupristin/dalfopristin % 97.7% 5 2.3% Tetracycline % 98.8% 2 0.8% 1 0.4% Trimethoprimsulfamethoxazole % 100.0% Vancomycin % 98.6% 8 1.2% 1 0.2% Total % 98.7% % % 5 0.1% a MIC readings are defined as drugs with 4 dilutions. Drugs with <4 dilutions were excluded b Essential agreement is defined as BIOMIC results within ± 1 two-fold dilution of the MicroScan results Downloaded from 20 on January 13, 2019 by guest

21 276 Downloaded from 21 on January 13, 2019 by guest

22 Table 3: BIOMIC Reading MicroScan Panels - Category Agreement by Drug (Breakpoint Readings Only a ) Gram Negative Bacteria Drug Total # of Drug/Organism Combinations Tested Category Agreement without Expert System Minor % Major % Very Major % Category Agreement with Expert System b Amikacin % 3 0.3% % % Amoxicillin-clavulanate % % 4 0.4% 1 0.1% % Ampicillin % % 2 0.2% % Ampicillin-sulbactam % % 2 0.2% % Aztreonam % % 2 0.2% % Cefazolin % 1 0.1% % Cefepime % % 3 0.3% % Cefotaxime % 6 1.1% 1 0.2% % Cefotetan % 3 0.9% 6 1.8% % Cefoxitin % % 2 0.5% % Cefpodoxime % 4 0.8% % Ceftazidime % % % Ceftizoxime % % Ceftriaxone % 5 0.5% 3 0.3% 1 0.1% % Cefuroxime-parenteral % % 2 0.2% 4 0.4% % Cephalothin % % 1 0.1% 1 0.1% % Chloramphenicol % 8 1.9% % Ciprofloxacin % % 1 0.1% 1 0.1% % Ertapenem % % Gatifloxacin % % 3 0.5% % Gemifloxacin % 3 1.2% % Gentamicin % 2 0.3% 1 0.1% % Imipenem % 7 0.6% 3 0.3% % Levofloxacin % % 2 0.2% % 22

23 Meropenem % 3 0.7% % Moxifloxacin % % Nitrofurantoin % % % Norfloxacin % 3 0.6% 2 0.4% 1 0.2% % Piperacillin % % 3 0.3% % Piperacillin-tazobactam % 6 0.6% 4 0.4% % Tetracycline % % 5 0.4% % Ticarcillin-clavulanate % % 4 0.3% 2 0.2% % Tobramycin % 1 0.1% % 1 0.1% % Trimethoprim % 1 0.4% % Trimethoprimsulfamethoxazole % % % Total % % % % 98.1% Gram Positive Bacteria Drug Total # of Drug/Organism Combinations Tested Category Agreement without Expert System Minor % Major % Very Major % Category Agreement with Expert System Amoxicillin-clavulanate % % 2 0.4% 99.6% Ampicillin % 100.0% Ampicillin-sulbactam % 7 1.5% 2 0.4% 98.5% Azithromycin % 100.0% Cefazolin % % 1 0.2% 1 0.2% 96.2% Cefepime % % 1 0.7% 5 3.3% 100.0% Cefotaxime % % 2 1.3% 1 0.7% 97.3% Cefoxitin % 100.0% Ceftriaxone % % 2 0.7% 2 0.7% 99.7% Cephalothin % 2 3.6% 100.0% Chloramphenicol % % 93.9% Ciprofloxacin % 1 0.3% 2 0.6% 99.1% 23

24 Clindamycin % 2 1.3% 98.7% Daptomycin % 1 0.5% 1 0.5% 98.9% Ertapenem % 7 5.0% 100.0% Erythromycin % 2 1.0% 99.0% Gatifloxacin % 100.0% Gentamicin % 3 1.0% 99.0% Gentamicin HLAR % 100.0% Imipenem % % 3 0.7% 98.8% Levofloxacin % 4 1.0% 1 0.2% 98.8% Linezolid % 5 1.2% 3 0.7% 98.1% Meropenem % % 98.5% Moxifloxacin % 100.0% Nitrofurantoin % 8 1.6% 98.4% Norfloxacin % 100.0% Ofloxacin % 100.0% Oxacillin % 100.0% Piperacillin-tazobactam % 3 2.1% 3 2.1% 100.0% Quinupristin/dalfopristin % 5 1.2% 1 0.2% 1 0.2% 98.3% Rifampicin % % 3 0.5% 95.7% Streptomycin HLAR % 100.0% Tetracycline % 8 1.9% 98.1% Trimethoprimsulfamethoxazole % % 97.2% Vancomycin % 100.0% Total % % % % 98.3% a Breakpoint readings are defined as drugs with < 4 dilutions. Drugs with 4 dilutions were excluded b To compare category agreement, raw MIC readings from both BIOMIC and MicroScan were interpreted by the BIOMIC software using the latest interpretive CLSI guideline (M2100-S21). To compare category agreement with the Expert System, the same set of expert rules were applied to the raw interpretive results using the BIOMIC software. 24

25 Table 4: BIOMIC Reading MicroScan Breakpoint Panels - Category Agreement by Organism Gram Negative Bacteria Total # of Drug/Organism Combinations Tested Category Agreement without Expert System Minor % Major % Very Major % Category Agreement with Expert System Acinetobacter sp % % 94.2% Aeromonas sp % % 87.5% Citrobacter sp % % 1 0.1% 97.3% Enterobacter sp % % 2 0.1% 98.3% Escherichia sp % % % 1 0.0% 98.7% Klebsiella sp % % 6 0.1% 1 0.0% 97.8% Morganella sp % 6 3.0% 2 1.0% 96.4% Proteus sp % % 9 0.4% 1 0.0% 97.8% Providencia sp % 1 0.6% 1 0.6% 98.8% Pseudomonas sp % % % % 93.7% Salmonella sp % 1 0.7% 100.0% Serratia sp % 7 3.5% 1 0.5% 93.5% Shigella sp % 100.0% Yersinia sp % 100.0% Yokenella sp % 100.0% Total % % % % 98.1% 25

26 277 Gram Positive Bacteria Total # of Drug/Organism Combinations Tested Category Agreement without Expert System Minor % Major % Very Major % Category Agreement with Expert System Enterococcus sp % % Staphylococcus % % a sp. Streptococcus sp % 100.0% Total % % % % 98.3% a The category agreement count for Staphylococcus sp. with the Expert System applied are 56 Minor, 26 Major and 6 Very Major Errors. Downloaded from 26 on January 13, 2019 by guest

27 Table 5: BIOMIC Reading MicroScan Panels - Identification Agreement by Organism Gram Negative Bacteria Identification by MicroScan No. Tested Achromobacter xylosoxidans 1 Acinetobacter baumannii/calcoaceticus Discrepancy a Organism Identification by BIOMIC 7 1 Pseudomonas fluorescens/putida 278 Second Choice Match b Acinetobacter lwoffii 4 1 Acinetobacter No baumannii/calcoaceticus Aeromonas hydrophila 1 Alcaligenes sp. 3 Burkholderia cepacia 1 Citrobacter amalonaticus 1 Citrobacter freundii 16 Citrobacter koseri (diversus) 7 Escherichia coli Escherichia vulneris, Yes Salmonella/Arizona Escherichia fergusonii 2 1 Yersinia ruckeri No Enterobacter aerogenes 10 Enterobacter cloacae 30 Klebsiella oxytoca 20 Klebsiella pneumoniae 98 Morganella morganii 11 2 Proteus mirabilis Yes Pantoea agglomerans 2 Yes 27

28 Proteus mirabilis 49 Proteus penneri 2 1 Proteus vulgaris Yes Proteus vulgaris 4 Providencia rettgeri 2 Providencia stuartii 4 Pseudomonas aeruginosa 86 3 A. xylosoxidans Yes Pseudomonas 3 2 Pseudo. aeruginosa Yes fluorescens/putida Pseudomonas stutzeri 3 2 Pseudomonas No fluorescens/putida Ralstonia pickettii 2 Salmonella enterica 3 Serratia marcescens 5 Stenotrophomonas maltophilia 5 Total Gram-negative First Choice Match: 98.1% (830/846) Gram positive Bacteria Identification by MicroScan No. Tested Enterococcus avium 1 Enterococcus faecalis 96 Enterococcus faecium 27 Enterococcus raffinosus 1 Micrococcus sp. 1 Discrepancy First and Second Choice Match: Organism Identification by BIOMIC 99.5% (842/846) Second Choice Match 28

29 Staphylococcus aureus Staphylococcus capitis, S. epidermidis, S. simulans Yes 279 Staphylococcus auricularis 1 Staphylococcus capitis 1 Staphylococcus epidermidis 61 1 S. aureus Yes Staphylococcus haemolyticus 5 Staphylococcus hominis 1 Staphylococcus hominis 2 10 Staphylococcus intermedius 3 2 S. simulans No Staphylococcus lugdunensis 1 Staphylococcus saprophyticus 5 Staphylococcus simulans 1 1 S. epidermidis Yes Staphylococcus urealyticum 1 Staphylococcus ureolyticus 1 Staphylococcus warneri 2 Streptococcus agalactiae 1 Streptococcus mitis/oralis 1 Total Gram-positive First Choice Match: 98.4% (421/430) First and Second Choice Match: 99.5% (428/430) a Discrepancy is defined as a different first choice of the organism identification result b Second choice match is noted when the 2nd choice of the organism identification by BIOMIC matches the first identification by MicroScan 29