Errors in Interpretation of Gram Stains From Positive Blood Cultures

Microbiology and Infectious Disease / ERRORS IN GRAM STAIN INTERPRETATION Errors in Interpretation of Gram Stains From Positive Blood Cultures Kenneth H. Rand, MD, 1 and Maria Tillan, MD 2* Key Words: Gram stain; Misread Gram stain; Gram positive; Gram negative; Acinetobacter; Bacillus DOI: 10.1309/V4KE2FPM5T8V4552 Abstract We reviewed major errors in Gram stain reports from positive blood cultures to identify patterns and potential clinical impact. During a 23-month period, blood cultures were misread for 57 (0.7%) of 8,253 patients. Of 5,885 read as gram-positive cocci, 6 (0.1%) had only gram-negative organisms by culture, 3 of which were Acinetobacter species. Of 1,959 read as gram-negative bacilli, 25 (1.3%) had only grampositive organisms by culture. Of these, 9 were Bacillus and 2 were Clostridium species. Nonrecognition of mixed Gram stains accounted for 28 errors that most often were associated with a reading of gram-positive cocci. In 4 cases, there were delays of 14 hours to 3 days in starting appropriate antibiotic treatment; 2 deaths occurred, although the erroneous Gram stain report probably was not contributory. Pathologists and laboratory personnel need to be aware of these types of misinterpretations and the potential effects on patient outcome. The Gram stain may be the oldest and most entrenched technique still in use in the microbiology laboratory. Because of its essential simplicity and its widespread familiarity, physicians almost never question its accuracy. Yet, no laboratory test in existence is 100% accurate, and the Gram stain is no exception owing to human interpretive error and the exigencies of the staining properties of certain bacteria. For example, it is well recognized that Bacillus species and certain other gram-positive species often stain gram-negative or gram-variable as cultures age because of cell wall changes with loss of viability. 1 Microbiologists are familiar with many of the problematic areas of the Gram stain, such as underdecolorization and overdecolorization and which species are likely to exhibit interpretive issues. However, we were unable to find any systematic studies in the literature of the incidence or clinical impact of misinterpretation of Gram stains. We reviewed the initial Gram stain readings for all blood cultures reported during a 23-month period, January 1, 2002, through November 30, 2003, because such errors could lead directly to the choice of an incorrect antibiotic or discontinuation of an appropriate one. Materials and Methods Blood cultures were performed with the Bactec 9240 using the Bactec Plus Aerobic/F Medium, the Bactec lytic/10 anaerobic/f medium, and the Bactec Peds Plus/F medium (BD Diagnostic Systems, Sparks, MD). All positive blood cultures were manually recorded in a log book in the Shands Hospital at the University of Florida Clinical Microbiology 686 Am J Clin Pathol 2006;126:686-690 686 DOI: 10.1309/V4KE2FPM5T8V4552

Microbiology and Infectious Disease / ORIGINAL ARTICLE Laboratory (Gainesville). This record included the date, time, Gram stain result, and person to whom the stat result was called. When the culture identification was complete, the final result was entered into the log book alongside the original Gram stain reading. Both of us independently reviewed all positive blood cultures obtained between January 1, 2002, and November 30, 2003. All cases in which the original Gram stain reading that was called to a care provider was inconsistent with the known Gram stain characteristics of the organism(s) based on final culture identification were recorded. All misread Gram stains were found within 24 to 48 hours when the organisms recovered from the subculture plates did not match those described in the initial Gram stain reading. In all cases, the errors were reported as soon as they were recognized, corrected reports were issued, and the proper procedures and interpretation were reviewed with the technologist involved. Because of the retrospective nature of the study, there was no way to systematically determine whether the error was due to an interpretive error, ie, different individuals would interpret the slide differently, or a clerical error, eg, inputting the wrong comment into the computer. Technical errors such as overdecolorization were not recorded systematically. We focused on the types of errors that had the greatest potential for patient harm, ie, errors in which a gram-negative result was reported but the final identification was only grampositive or vice versa. We did not record cases in which grampositive bacilli were reported but only gram-positive cocci ultimately were found (and likewise for the gram-negatives) because such reports were less likely to lead to changes in antibiotic coverage with serious harmful effects on outcome. All cases were recorded in which only gram-positive cocci were reported but gram-negative species (alone or in mixed culture with a gram-positive organism) ultimately were identified by culture; only gram-positive bacilli were reported but gram-negative species (alone or in mixed culture with a gram-positive organism) ultimately were identified by culture; and only gram-negative bacilli were reported but gram-positive species (alone or in mixed culture with a gram-negative organism) ultimately were identified by culture. There were so few blood cultures with gram-negative cocci that no interpretive errors were found. Yeasts were excluded from the analysis. Gram Stain Gram stain of positive blood cultures was performed as described 2 with reagents from Becton Dickinson, Sparks, MD, according to the instructions from the manufacturer. Briefly, slides were air dried, fixed with methanol, and allowed to air dry on a slide warmer. Crystal violet was applied for 45 to 60 seconds, slides were rinsed in tap water, and Gram iodine was overlayed for 1 to 3 minutes. Decolorization was performed with acetone:95% ethanol, 25:75 by volume for less than 10 seconds. Slides were counterstained with safranin for 40 to 60 seconds. Outcome Studies After obtaining approval from our institutional review board for waiver of informed consent, the charts of patients with misinterpreted initial Gram stains were reviewed. Results As shown in Table 1, the Gram stains of 57 (0.7%) of 8,253 positive blood cultures during the 23-month study period were misread. After omitting Bacillus species, which were almost certainly contaminants, there were 46 (0.6%) of 8,253 misread Gram stains. Of the 46, 28 were mixed cultures in which the second (or third) organisms were not recognized in the initial Gram stain reading. Data for the remaining 18 cases are given in Table 2. Chart review revealed that there were no adverse consequences for 12 of 16 patients. For the remaining 4, there were 14- to 72-hour delays in switching to appropriate antibiotics. Two patients died of septic shock, but the delay in reporting correct Gram stain results in case 4 was 14 hours and in case 13 was 48 hours. Empiric gram-negative coverage was ineffective in the former case because slow growth Table 1 Summary of Gram Stains of All Positive Blood Cultures * No. of Positive No. No. of Major No. of Mixed No. of Bacillus Blood Culture Gram Stain Blood Cultures Misread Errors Cultures Species Gram-positive cocci 5,885 22 (0.4) 6 (0.1) 16 (0.3) Gram-positive rods 367 10 (2.7) 7 (1.9) 3 (0.8) Gram-negative rods 1,959 25 (1.3) 5 (0.36) 9 (0.5) 11 (0.6) Gram-negative cocci 42 0 (0) 0 (0) 0 (0) Total 8,253 57 (0.7) 18 (0.2) 28 (0.3) 11 (0.1) * Data are given as number (percentage). A major error was defined as a single organism whose Gram stain result was opposite that found in culture results. See Discussion, last sentence of first paragraph. Am J Clin Pathol 2006;126:686-690 687 687 DOI: 10.1309/V4KE2FPM5T8V4552 687

Rand and Tillan / ERRORS IN GRAM STAIN INTERPRETATION of the organism delayed identification (Capnocytophaga), whereas in the latter case, appropriate treatment failed even after susceptibility information was available. Two systematic errors were observed. First, in 11 cases, Bacillus species were read as gram-negative bacilli. The case shown in Image 1 illustrates well how gram-negative Bacillus species can sometimes appear. Second, in 5 of 13 cases in which Gram stains were read as only gram-positive cocci or gram-positive bacilli, an Acinetobacter species was identified as the sole organism (4 cases) or with another gramnegative bacillus (1 case). As shown in Image 2, Acinetobacter can appear unmistakably as gram-positive. The Table 2 Patient Characteristics, Underlying Diagnoses, Interpretive Error, and Outcome of Misread Gram Stains Drug at Time of Antibiotic Change Case No./ Original Gram Final Erroneous Gram Stain Based on Erroneous Adverse Sex/Age(y) Diagnosis Stain Result Culture Result Interpretation Gram Stain Interpretation Effect Outcome 1/F/75 Diverticulitis with Gram+ rods Bacteroides sp; Ticarcillin disodium/ Ciprofloxacin; None Survived sepsis Escherichia coli clavulanate metronidazole potassium added 2/F/78 Urosepsis, CAP Gram+ rods Klebsiella Ceftriaxone, None None Survived pneumoniae azithromycin 3/F/48 Endocarditis Gram+ rods Acinetobacter Not available Not available Probably none Survived lwoffi; Enterobacter sp 4/M/45 Neutropenic fever, Gram+ rods Capnocytophaga Ciprofloxacin, None 14-h delay until Died; septic shock mucositis vancomycin another bottle 13 d after blood positive for gram culture drawn, rods and cefepime despite imipenem and gentamicin for 5 d begun 4 d started after culture first known to be positive 5/M/54 Respiratory failure Gram+ rods Haemophilus Trimethoprim None; thought to None Survived influenzae sulfamethoxazole be a contaminant 6/M/73 Diabetes, chronic Gram+ rods E coli Not available Not available Not available Survived bronchitis 7/F/38 Line sepsis Gram+ rods Acinetobacter Vancomycin Discontinued line; None Survived baumannii discontinued vancomycin 8/F/60 CAP, heart failure, Gram rods Enterococcus Ceftriaxone, None; Entero- None Survived COPD, diabetes faecalis azithromycin coccus thought to be a contaminant 9/F/3 Stage IV neuro- Gram rods Clostridium Cefepime, piperacillin Gentamicin added 3-d delay in Survived blastoma, neutro- clostridioforme sodium/tazo- starting penic fever, mucositis bactam sodium metronidazole 10/F/52 Fever; subarachnoid Gram rods Corynebacterium Vancomycin, None None Survived bleeding ampicillin 11/M/58 Urosepsis, Gram rods Clostridium sp Vancomycin Gatifloxacin added None Survived lymphoma 12/F/78 Fever, infected Gram rods Staphylococcus Ertapenem Vancomycin None Survived decubitus, ESRD coagulase-negative 13/F/51 ESRD, renal trans- Gram+ cocci A baumannii Vancomycin Ciprofloxacin; 36- to 48-h delay Died; sepsis, multiplant, diabetes, trimethoprim in starting organ failure, 30% fever, cellulitis vs. sulfamethoxazole imipenem body surface Sweet syndrome added bullous lesions, bullous fluid positive for ciprofloxacin-resistant Acinetobacter 14/M/68 Abdominal abscess/ Gram+ cocci Bacillus fragilis Imipenem, None None Survived sepsis bladder amikacin, cancer, post bilateral vancomycin nephrectomy 15/M/64 Myelodysplastic Gram+ cocci A baumannii Gatifloxacin Vancomycin 24-h delay in Survived syndrome, starting fever sepsis cefepime 16/M/44 Fever, sepsis, Gram+ cocci K pneumoniae Imipenem None None Survived pneumatosis coli, 35%-40% burn 17/F/21 Fever, MVC, Gram+ cocci A baumannii Cefazolin, Vancomycin None Survived multiple trauma gentamicin 18/F/66 Ruptured appendix Gram+ cocci Bacteroides sp Amoxicillin/clavula- None None Survived and pelvic abscess nate potassium CAP, community-acquired pneumonia; COPD, chronic obstructive pulmonary disease; ESRD, end-stage renal disease; MVC, motor vehicle crash. 688 Am J Clin Pathol 2006;126:686-690 688 DOI: 10.1309/V4KE2FPM5T8V4552

Microbiology and Infectious Disease / ORIGINAL ARTICLE Image 1 Bacillus species staining gram-negative from a positive blood culture. Although this smear could be interpreted as gram-variable, it is easy to understand how it could be read as gram-negative ( 1,000). Image 2 Acinetobacter staining gram-positive from a positive blood culture ( 1,000). morphologic features of the gram-positive Acinetobacter organism were somewhat variable, most often recorded as gram-positive cocci and less often as gram-positive cocci in pairs or pairs and chains. In 13 cases, mixed gram-positive and gram-negative cultures initially were read as gram-positive cocci only. The gram-positive organisms were coagulase-negative Staphylococcus species (6 cases), Enterococcus species (5 cases), or both (2 cases). In these cases, the original Gram stains were misread only in the sense that an unsuspected organism of opposite staining characteristics also was present. These second organisms may not have been present in high enough quantities to be visualized on Gram stain, but the results of restaining or reviewing the original stain after culture results became available were not recorded systematically. The following gram-negative organisms were found by subculture: Acinetobacter species, 3; Pseudomonas aeruginosa, 3; Enterobacter aerogenes, 2; Serratia marcescens, 2; and Klebsiella pneumoniae, 1. Chart review revealed 4 patients with a clinically recognized abdominal focus of infection or illness, and all had infections with Enterococcus species and an enteric gram-negative organism. In the other cases, the unrecognized gram-negative organism was related to another known site of infection (line infection, 2; sternal wound infection, 1) or had no obvious source for the gramnegative organism (4 cases). Two charts were unavailable. In 14 cases, there were only gram-negative bacilli in the initial Gram stain reading but final cultures grew gram-negative and gram-positive organisms. In 8 of these cases, coagulasenegative staphylococci were found later by culture and were not clinically significant. Two patients had Streptococcus viridans that was not considered pathogenic in 1 case but was recovered from peritoneal fluid along with Serratia species, which were the gram-negative organisms in the blood culture. In this case, infection with the same organisms recurred 3 months later. One patient had a Staphylococcus aureus infection in addition to the gram-negative P aeruginosa, and both organisms were considered pathogenic in a catheter-related bacteremia. One patient had a Streptococcus pneumoniae infection, but the chart could not be obtained to determine the clinical relevance, and the remaining 2 patients had infections with Lactobacillus and Bacillus species that were not pathogens. The last patient had a culture read as gram-positive bacilli but was found to have infection with Clostridium and Bacteroides species by culture. This patient had a small bowel perforation repaired 2 days after the blood culture was obtained. Discussion It is believed that bacterial cells stain gram-positive because the thickness of their cell wall peptidoglycan layer prevents the elution of the crystal violet iodine mordant insoluble complex in an organic solvent. 1,3 In gram-negative cells, the complex diffuses out because of the thin peptidoglycan layer together with damage to the cell wall by the decolorizing solvent, leaving the cells visualized by the safranin counterstain. Am J Clin Pathol 2006;126:686-690 689 689 DOI: 10.1309/V4KE2FPM5T8V4552 689

Rand and Tillan / ERRORS IN GRAM STAIN INTERPRETATION Bacillus and Clostridium species typically are 95% to 100% gram-positive early in their growth phase in broth cultures but become 40% to 50% gram-negative in the late growth phase and 90% to 95% gram-negative in the stationary phase. 1 Although one might expect that blood cultures growing Bacillus species would be in mid growth phase and, thus, gram-positive, the loss of Gram stain retention associated with aging of the culture most likely accounts for the frequency of misreading for these species. Even so, only 11 (7.6%) of 144 blood culture bottles that grew Bacillus species during the study period were read as gram-negative instead of gram-positive. In contrast with the loss of positive Gram stain through overdecolorization or the effects of growth phase on the cell wall of gram-positive organisms, Acinetobacter has been reported to stain gram-positive, despite proper Gram stain technique, 4 and was involved in 5 of our misread cases. In addition, 3 of the mixed cultures read as only gram-positive cocci had Acinetobacter in addition to a gram-positive coccus in culture. During this 23-month period, there were 68 patient blood cultures positive for Acinetobacter species, so 8 (12%) of 68 Acinetobacter species in blood culture were read initially as gram-positive. No explanation is available in the literature for this phenomenon, although it is recognized among microbiologists. 5 Misread Gram stains from positive blood cultures should be recognized by the laboratory in 18 to 36 hours following the erroneous Gram stain report because at that time, there should be sufficient growth on plates that the inconsistent colony morphologic features or presence of a mixed culture should be recognized. However, anaerobes and slow-growing fastidious organisms, such as Capnocytophaga reported herein, may require an additional 24 hours before such errors could be recognized. Clinical microbiology laboratories must maintain a high level of awareness for the possibility that blood culture Gram stains can be misread. Not only can technical and interpretive errors occur but also, as demonstrated in this report, the potential for certain organisms to stain opposite their characteristic manner must be kept in mind, eg, Acinetobacter, Clostridium, and Bacillus species. In 4 of our cases with the morphologic features of enterococci (ie, gram-positive cocci in pairs and chains) in the setting of a potential abdominal source of infection, there also was an enteric gram-negative rod found on culture. The overall 0.7% error rate of misinterpreted Gram stains from positive blood culture bottles is low, but laboratory professionals and pathologists must be aware of the potential types of error that can occur. From the Departments of 1 Pathology and Immunology and Laboratory Medicine, and 2 Medicine, Division of Infectious Disease, University of Florida, Gainesville. Supported in part by the Departments of Pathology and Immunology and Laboratory Medicine (Dr Rand) and the Department of Medicine, Division of Infectious Disease (Drs Rand and Tillan), University of Florida. Presented in part at the 42nd meeting of the Infectious Diseases Society of America, Boston, MA, September 2004. Address reprint requests to Dr Rand: Dept of Pathology, Immunology and Laboratory Medicine, University of Florida College of Medicine, PO Box 100275, Gainesville, FL 32610. * Dr Tillan is currently with the Division of Infectious Diseases, University of Louisville, Louisville, KY. Acknowledgment: We gratefully acknowledge the support of the staff of the Shands Hospital at the University of Florida Clinical Microbiology Laboratory. References 1. Beveridge TJ. Mechanism of Gram variability in select bacteria. J Bacteriol. 1990;172:1609-1620. 2. York MK. Gram stain. In: Isenberg HD, ed. Clinical Microbiology Procedures Handbook. Washington, DC: American Society of Microbiology; 2004. 3. Popescu A, Doyle RJ. The Gram stain after more than a century. Biotech Histochem. 1996;71:145-151. 4. Harrington BJ, Plenzler M. Misleading Gram stain findings on a smear from a cerebrospinal fluid specimen. Lab Med. 2004;35:475-478. 5. Schreckenberger PC, Daneshvar ME, Weyant RS, et al. Acinetobacter, Achromobacter, Chryseobacterium, Moraxella, and other nonfermentative gram-negative rods. In: Murray PR, Baron EJ, Jorgensen JH, et al, eds. Manual of Clinical Microbiology. 8th ed. Washington, DC: ASM Press; 2003. 690 Am J Clin Pathol 2006;126:686-690 690 DOI: 10.1309/V4KE2FPM5T8V4552