Cost-Effectiveness of Blood Cultures for Adult Patients with Cellulitis

1483 Cost-Effectiveness of Blood Cultures for Adult Patients with Cellulitis Bezalel Perl, 1 Nathan P. Gottehrer, 2 David Raveh, 1 Yechiel Schlesinger, 1 Bernard Rudensky, 3 and Amos M. Yinnon 1 From the 1 Infectious Diseases Unit, 2 Department of Emergency Medicine, and 3 Clinical Microbiology Laboratory, Shaare Zedek Medical Center and Hadassah-Hebrew University Medical School, Jerusalem, Israel To assess the cost-effectiveness of blood cultures for patients with cellulitis, a retrospective review was conducted of clinical and microbiological data for all 757 patients admitted to a medical center because of community-acquired cellulitis during a 41-month period. Blood cultures were performed for 553 patients (73%); there were a total of 710 blood samples (i.e., a mean of 1.3 cultures were performed per patient). In only 11 cases (2.0%) was a significant patient-specific microbial strain isolated, mainly b-hemolytic streptococci (8 patients [73%]). An organism that was considered a contaminant was isolated from an additional 20 culture bottles (3.6%). The cost of laboratory workup of the 710 culture sets was $36,050. Isolation of streptococci led to a change from empirical treatment with cefazolin to penicillin therapy for 8 patients. All patients recovered. In conclusion, the yield of blood cultures is very low, has a marginal impact on clinical management, and does not appear to be cost-effective for most patients with cellulitis. Erysipelas and cellulitis are common skin infections involving the superficial and subcutaneous tissues, respectively. Most authorities accentuate the importance of determining the specific microbial origin of these infections, to tailor antimicrobial therapy [1]. However, needle aspiration of the lesion of erysipelas or obtaining an aspirate from the edge of cellulitis facilitates isolating an organism in!20% of cases [2, 3], while potentially exposing the patient to superinfection. Therefore, blood has become the most common culture specimen from patients with cellulitis who present to the emergency department. However, the yield of blood cultures for these patients seems to be rather low and only marginally affects treatment. The purpose of this study was to determine the overall yield of blood cultures to develop an evidence-based, cost-effective policy on the role of blood cultures in the management of cellulitis. Methods This retrospective study included all consecutive adult patients who were diagnosed with cellulitis and were admitted to our hospital through the emergency department. Patients were detected according to daily records that are completed each morning on rounds in the emergency department and include patient identification data and clinical diagnoses made by the physicians on rounds. Patients with facial cellulitis were excluded, as were patients with focal infections other than those of the skin (e.g., exit site infection of tunneled central catheters was excluded). The study Received 9 March 1999; revised 5 August 1999. Reprints or correspondence: Dr. A. M. Yinnon, Infectious Diseases Unit, Shaare Zedek Medical Center, P.O. Box 3235, Jerusalem 91031, Israel (yinnon@szmc.org.il). Clinical Infectious Diseases 1999;29:1483 8 1999 by the Infectious Diseases Society of America. All rights reserved. 1058-4838/1999/2906-0022$03.00 period encompassed 41 months from 1 April 1995 through 31 August 1998. In addition, the records of the microbiology laboratory were examined to retrieve information on the number of blood specimens for culture, the number of positive cultures (if any), and the identification of all cultured isolates and antimicrobial susceptibilities. Hospital records of patients with cellulitis who had a positive blood culture were retrieved and reviewed. Relevant clinical and microbiological data were compiled; features of patients with positive cultures were compared with those of patients whose cultures remained sterile, those of patients whose blood cultures had contaminants, and those of patients from whom no blood specimens were obtained for culture. Local site specimens for culture and tissue aspirates are obtained only extremely rarely; therefore, these samples were not included in this study. Blood specimens were inoculated into resin-containing aerobic and anaerobic media and were incubated in the BACTEC 9240 System (Becton Dickinson, Sparks, MD). Blood culture bottles were routinely incubated for up to 5 days, and terminal subcultures were not routinely performed unless clinically indicated. For quality control purposes, our laboratory periodically weighs blood specimens for culture, to help ensure that appropriate blood volume is inoculated into blood culture bottles. We reviewed the clinical microbiology laboratory s database of computerized blood culture results to compare the yield of blood cultures for patients with cellulitis with the yield of blood cultures performed for all other purposes. This database was started in 1990 and includes data for almost 4000 patients with positive blood cultures: excluded were coagulase-negative Staphylococcus, diphtheroids, Bacillus species, and a-hemolytic streptococci, which were considered contaminants if isolated from 1 culture bottle of 1 set when at least 1 other set of blood culture bottles was sterile [4]. We observed a very low incidence of Staphylococcus aureus bacteremia among our patients. To determine possible failure of detection for patients with cellulitis, we reviewed the records and clinical diagnoses of all patients for whom S. aureus was isolated

1484 Perl et al. CID 1999;29 (December) from cultures of blood specimens obtained in the emergency department since January 1997. The costs associated with performing blood cultures were obtained from the clinical microbiology laboratory and included the following: $50 per aerobic or anaerobic culture and $50 for antimicrobial susceptibility testing per positive blood culture. Clinical and microbiological data were entered into a computer program (Epi-Info Version 6.04c; Centers for Disease Control and Prevention, Atlanta, GA). Statistical calculations were performed with the same program and included Student s t test, x 2 analysis, and univariate regression analysis. Statistical significance was set at P!.05. Results From the daily records of the emergency department, we detected 757 adult patients who were diagnosed with cellulitis during the period 1 April 1995 through 31 August 1998. Blood specimens for culture were received in the clinical microbiology laboratory from 553 (73%) of these patients. From some of the patients we obtained 11 blood specimen for culture; a total of 710 blood samples were received (i.e., a mean of 1.3 cultures were performed per patient). Each of these 710 samples was inoculated into an aerobic and an anaerobic culture bottle. In only 11 cases (2.0%) was a significant patient-specific microbial strain isolated. An organism that was considered a contaminant was isolated from an additional 20 cultures (3.6% of patients; table 1). For comparison, during the period 1990 1997, the laboratory received annually a mean number of blood specimens for culture SD of 13,985 4242; of these cultures, a mean proportion SD of 10.5% 0.9% yielded a significant organism (excluding coagulase-negative Staphylococcus, diphtheroids, and Bacillus species) [4]. The cost associated with negative blood cultures and those with contaminants was $34,950 ($50 per culture for 699 cultures); the cost for 11 positive cultures was $1100 ($50 per culture plus $50 for antimicrobial susceptibility testing for 11 cultures). Therefore, the total cost of laboratory workup was $36,050. The isolation of streptococci from blood cultures led to a therapeutic change from cefazolin to penicillin for 8 patients; this treatment was microbiologically more appropriate with, however, a marginal reduction in incurred cost. Hospital records for only 9 of the 11 patients with bacteremia were recovered. The 2 lost records were for a patient with bacteremia due to group A Streptococcus and a patient with bacteremia due to group G Streptococcus. Some clinical and laboratory features of the 9 patients are summarized in table 2. As shown, most patients were elderly and had cellulitis of the lower extremities. No common underlying disease was detected, although at least 3 patients (cases 1, 4, and 6) had an underlying local factor that could have served as the portal of entry. None of the remaining patients with streptococcal bacteremia, due to either group A or group G organisms, had any local or systemic underlying or predisposing illness. Patients with bacteremia due Table 1. Microorganisms isolated from cultures of blood specimens from 553 patients with cellulitis. Organism No. (%) positive cultures Significant organism 11 a (100) Group G Streptococcus 5 (45.4) Group A Streptococcus 3 (27.3) Staphylococcus aureus 1 (9.1) Vibrio vulnificus 1 (9.1) Morganella morganii 1 (9.1) Isolated contaminants b 20 c (100) Coagulase-negative Staphylococcus 12 (60) Diphtheroid 2 (10) a-hemolytic Streptococcus 2 (10) Micrococcus 1 (5) Ochrobactrum 1 (5) Neisseria species 1 (5) Nonhemolytic Streptococcus 1 (5) a 2% of 553 blood cultures. b Isolated from only 1 culture bottle of the set. c 3.6% of 553 blood cultures. to group G Streptococcus tended to have higher mean temperatures at presentation (39.1; SD, 0.8 C) than did patients with bacteremia due to other organisms ( 38.4 0.7 C; P!.025). No differences in clinical course or results of laboratory tests were detected among these patients. However, in cases of bacteremia due to group G Streptococcus, organisms were isolated from only a small percentage of the culture bottles (4 [25%] of 16) as opposed to other organisms that grew in a much higher percentage of culture bottles (16 [100%] of 16; P!.001). The low prevalence of S. aureus bacteremia in our survey was surprising. To rule out inadvertent omission of patients, we assessed all discharge diagnoses of patients who had S. aureus isolated from cultures of blood specimens obtained at admission since 1 January 1997. None of these patients had been diagnosed with cellulitis. To determine possible risk factors for bacteremia, as indicated by 1 positive blood cultures, clinical features of 9 patients with positive cultures were compared with those of 48 patients whose cultures remained sterile, those of 17 patients whose blood cultures had contaminants, and those of 20 patients from whom no blood specimens for culture were obtained (table 3). Univariate regression analysis revealed several variables that could predict the presence of bacteremia: age 145 years; shorter duration of symptoms before physical examination; higher incidence of fever; temperature 38.5 C at admission; and WBC count 113,300/mm 3 at admission. Discussion Cellulitis is an acute spreading infection of the skin that extends deeper than erysipelas to involve subcutaneous tissue [1]. The infection may cause significant local tissue damage and has a propensity to spread via the lymphatics and bloodstream. Group A Streptococcus and S. aureus are considered to be re-

Table 2. Clinical features of patients with cellulitis and bacteremia. Patient no. Age (y), sex Date, mo/y Isolate No. pos/ total no. BCBs Site of infection Underlying condition(s) Highest temp., C Highest WBC count per mm 3 / ESR, mm/h Clinical course and antibiotic therapy 1 62, F 12/1995 Staphylococcus aureus 2/2 Right leg Recurrent SP excision of lipoma 38.4 16,700/NA Clinical improvement after 2 d of treatment with iv cefazolin 2 62, F 5/1996 Group A Streptococcus 4/4 Right leg NIDDM; renal failure 38.6 9900/87 Fever lasted 10 d during iv penicillin therapy 3 82, M 7/1996 Group A Streptococcus 2/2 Right leg IHD, SP CABG 37.5 12,700/114 Clinical improvement after 2 d of treatment with cefazolin 4 69, M 9/1997 Vibrio vulnificus, Clostridium perfringens 4/4 Right hand, 2nd finger 5 70, M 10/1997 Group G Streptococcus 1/4 Left leg and wound 6 50, F 10/1997 Morganella morganii 4/4 Right leg NIDDM; ESRD, receiving hemodialysis 7 69, M 11/1997 Group G Streptococcus 1/4 Left leg 2 mo SP fracture of left tibia; NIDDM HTN 38 10,500/39 Cellulitis and chills 6 h after injury from a fish bone; lymphangitis, axillary lymphadenitis with local bullae, and necrosis treated with iv ceftazidime and gentamicin; afebrile after 48 h Chronic venous insufficiency 39.5 13,400/106 Afebrile after 2 d of cefazolin treatment, which was later changed to penicillin 38.4 13,600/NA Treated with iv cefazolin; complete recovery 39 19,800/30 Afebrile 1 d after treatment with penicillin and metronidazole 8 71, M 6/1998 Group G Streptococcus 1/6 Left leg SP CABG with leg vein grafts 38.1 8500/7 Fever resolved 1 d after iv cefazolin treatment, which was later changed to penicillin 9 47, F 8/1998 Group G Streptococcus 1/2 Right leg Recurrent cellulitis 39.2 15,000/NA Fever resolved 2 d after iv cefazolin treatment, which was later changed to penicillin NOTE. BCB, blood culture bottle; CABG, coronary artery bypass grafting; ESR, erythrocyte sedimentation rate; ESRD, end-stage renal disease; HTN, hypertension; IHD, ischemic heart disease; NA, not available; NIDDM, noninsulin-dependent diabetes mellitus; pos, positive; SP, status post.

1486 Perl et al. CID 1999;29 (December) Table 3. Patients with cellulitis and positive cultures compared with patients with negative cultures, patients from whom no blood specimens for cultures were obtained, and patients whose blood cultures had contaminants. Characteristic Positive cultures (n = 9) Negative cultures (n = 48) No cultures (n = 20) Cultures with contaminants (n = 17) P ( / ) a P ( /n) b Sex Male 5 (55) 26 (54) 8 (40) 9 (53) Female 4 (44) 22 (46) 12 (60) 8 (47) Age, y 65 11 55 18 53 19 62 13!.05!.05 18 45 0 18 (37) 9 (45) 2 (12)!.001!.001 46 65 4 (44) 18 (37) 5 (25) 7 (41) 165 5 (55) 12 (25) 6 (30) 8 (47) Site Left leg 3 (33) 23 (48) 6 (30) 10 (59) Right leg 5 (55) 24 (50) 12 (60) 5 (29) Other 1 (11) 1 (2) 2 (10) 2 (12) Mean duration of symptoms, c d SD 1.2 0.4 3.9 4.4 3.7 3.2 5.3 7.3 Local symptoms Swelling 0 26 (54) 11 (55) 12 (71) Pain 7 (78) 39 (81) 16 (80) 11 (65) Redness 0 36 (75) 18 (90) 10 (59) Systemic symptoms Fever 8 (89) 33 (69) 13 (65) 13 (76)!.001 Chills 3 (33) 5 (10) 0 1 (6) Other 1 (11) 6 (12) 3 (15) 3 (18) Comorbid condition Recent local trauma 2 (22) 12 (25) 4 (20) 1 (6) Diabetes mellitus 2 (22) 8 (17) 1 (5) 2 (12) Recurrent cellulitis 2 (22) 11 (23) 5 (25) 2 (12) Trauma in past d 3 (33) 21 (44) 2 (10) 4 (23) Highest temp, e C 38.5 0.6 37.8 1.1 37.4 1.1 37.6 1.0!.01!.001 WBC f 10 3 /mm 3 13.3 3.5 10.1 4.6 10.3 4.4 12.3 7.7!.025!.005 Treatment Cefazolin 8 (89) 42 (87) 17 (85) 11 (65) Other antimicrobial(s) 1 (11) 6 (12) 3 (15) 6 (35) Duration of admission, g d 8.3 6.3 6.1 3.1 4 2.3 6.4 4.4 h NOTE. Data are no. (%) of patients or mean SD. CABG, coronary artery bypass grafting; SP, status post; temp, temperature. a Comparison of patients with positive blood cultures and those with negative blood cultures. b Comparison of patients with positive blood cultures and those with no blood cultures. c Duration of symptoms before physical examination for patients with positive cultures was significantly shorter than duration for patients in all control groups ( P!.001). d For example, SP vein resection for CABG, venous stripping, SP skin grafting, SP suturing of wound, etc. e As measured in the emergency department on day of admission. f Highest WBC count on day of admission. g Duration of admission was significantly longer for patients with positive cultures than for patients in all control groups (P!.05 to P!.001). h Range, 2 20 days. sponsible for most cases of cellulitis, although other organisms may be involved occasionally [1]. By using aspirates from the advancing edge of cellulitis, skin biopsy, and blood cultures, potential pathogens are isolated from!20% of patients [1 5]. Cultures of specimens from closed lesions of cellulitis, if present, that are obtained by a fine-needle aspiration technique are occasionally positive [2, 3]. However, aspirates are not routinely obtained from infected soft tissue, because of both low yield and concern for superinfection; therefore, the specimen most frequently obtained for isolation of pathogens from patients with cellulitis is blood, and usually it is the only specimen. The importance of blood cultures has been assessed in various clinical situations [6 8]; however, few studies have examined the yield and cost-effectiveness of blood cultures in the setting of cellulitis (the principal purpose of our study). In this retrospective study of 757 patients with cellulitis who were admitted to our medical center during a 41-month period, 1 blood specimen for culture was obtained from 553 patients (73%). A significant organism was isolated in only 11 cases (2%): group G Streptococcus (5 patients), group A Streptococcus (3), S. aureus (1), Vibrio vulnificus (1), and Morganella morganii (1). Empirical treatment of cellulitis in our hospital, as in most medical centers, is with a first-generation cephalosporin. Therefore, of the 11 patients, the 9 (82%) from whom gram-positive organisms were isolated were treated with an appropriate antimicrobial from the outset and would have been treated appropriately even if blood specimens for culture were not obtained. The patient with V. vulnificus bacteremia had a typical history of fish bone injury and associated cellulitis, and the patient with morganella sepsis was receiving long-term treat-

CID 1999;29 (December) Blood Cultures in Cellulitis 1487 ment with hemodialysis which was temporarily administered through an indwelling central catheter. Consequently, these patients with bacteremia due to gram-negative organisms had sufficient unusual features to raise clinical suspicion that would have led to obtaining blood specimens for culture, even if that was not routine practice. Initial empirical treatment for both patients included agents providing appropriate coverage for gram-negative bacilli; in addition, blood cultures were almost twice as likely to be contaminated than positive. The cost of the required laboratory workup of these blood cultures was $36,050; this figure could be considerably higher in other countries. The isolation of streptococci from blood cultures in 8 of the 11 cases allowed for change of empirical treatment with a first-generation cephalosporin to therapy with penicillin; however, the attendant benefits, including economic benefits, are indeed minor compared with the substantial costs of blood cultures. In our study, there were only very few patients with cellulitis other than that of the lower extremities, and there was only 1 immunocompromised patient who received chronic steroid therapy. Comparison of clinical features of patients with bacteremia with those of patients whose cultures remained sterile revealed several indicators for the presence of bacteremia: age 145 years; shorter duration of symptoms before physical examination; higher incidence of fever; temperature 38.5 C at admission; and WBC count 113,300/mm 3 at admission. Therefore it seems that blood culture for most patients with cellulitis is not cost-effective, with the possible exception of those with unusually severe cases (such as elderly patients with acute onset of illness, high-grade fever, and significant leukocytosis and immunocompromised patients). A comparable study of blood cultures for children with cellulitis also showed that blood cultures have a very low yield and are not cost-effective for the younger age group [9]. Our study design has several limitations. First, it is a retrospective study; hence, reports of clinical appearance and documentation of the precipitating cause are limited to patient records. Second, we may have missed some cases of cellulitis, because some patients may have been admitted to our medical center between rounds during which patient lists and diagnoses were made. Third, we may have missed some cases of bacteremia, since some patients may have received antimicrobial treatment before blood specimens for culture were obtained. Finally, blood specimens for culture were not obtained from 27% of the patients admitted to our hospital; the reason for this omission could not be determined by our case-control study, but possibly it indicates the clinicians assessment of a very low likelihood of bacteremia. A similar percentage was reported in a study on cellulitis in children, as described elsewhere [9]. The organisms most commonly associated with cellulitis are group A Streptococcus and S. aureus; however, various other organisms have been reported, including other streptococci, anaerobes, and gram-negative bacilli [10 12]. In our study, group G Streptococcus was isolated from 5 (45%) of the 11 patients with cellulitis and bacteremia. The occasional association of this organism with cellulitis has been reported in other series [13 19]. However, in none of these series was group G Streptococcus the single most frequently isolated organism causing cellulitis. It has been noted that cellulitis associated with group B, C, or G b-hemolytic Streptococcus is often accompanied by severe underlying disease, such as diabetes mellitus, cirrhosis, or malignancy [19, 20]. In our series, 1 patient with group G Streptococcus bacteremia had diabetes mellitus, whereas all patients had an underlying local pathology on their extremities that served as a portal of entry for bacteria. All patients with cellulitis completely recovered. In summary, this study showed that the yield of blood cultures is very low and does not appear to be cost-effective for most immunocompetent patients with cellulitis. The costs associated with blood cultures are substantial, and the results do not significantly alter therapy. In addition, a positive result is almost twice as likely to be a contaminant than to be a truepositive result, which could lead to further tests and unnecessary treatment. Blood cultures for patients with cellulitis should probably be limited to those with unusually severe cases, such as possibly elderly patients with acute onset of illness, highgrade fever, and significant leukocytosis and immunocompromised patients. References 1. Swartz MN. Cellulitis and subcutaneous tissue infections. In: Mandell GL, Bennett JE, Dolin R, eds. Mandell, Douglas and Bennett s principles and practice of infectious diseases. 4th ed. New York: Churchill Livingstone, 1995:909 29. 2. Sigurdsson AF, Gudmundsson S. The etiology of bacterial cellulitis as determined by fine-needle aspiration. Scand J Infect Dis 1989;21:537 42. 3. Sachs MK. The optimum use of needle aspiration in the bacteriologic diagnosis of cellulitis in adults. Arch Intern Med 1990;150:1907 12. 4. Yinnon AM, Schlesinger Y, Gabbay D, Rudensky B. Analysis of five years of bacteremias; importance of stratification of microbial susceptibilities by source of patients. J Infect 1997;35:17 23. 5. Hook EW III, Hooton TM, Horton CA, Coyle MB, Ramsey PG, Turck M. Microbiologic evaluation of cutaneous cellulitis in adults. Arch Intern Med 1986;146:295 7. 6. Niederman MS. Severe community-acquired pneumonia: what do we need to know to effectively manage patients? [editorial]. Intensive Care Med 1996;22:1285 7. 7. Henke PK, Polk HC Jr. Efficacy of blood cultures in critically ill surgical patients. Surgery 1996;120:752 8. 8. Cunney RJ, McNamara EB, Alansari N, Loo B, Smyth EG. The impact of blood culture reporting and clinical liaison on the empiric treatment of bacteremia. J Clin Pathol 1997;50:1010 2. 9. Berger Sadow K, Chamberlain JM. Blood cultures in the evaluation of children with cellulitis. Pediatrics 1998;101:1 4. 10. Gorbach SL. IDCP guidelines: superficial skin and soft tissue infections. Part II. Infect Dis Clin Pract 1997;6:6 11. 11. Brook I, Frazier EH. Clinical features and aerobic and anaerobic microbiologic characteristics of cellulitis. Arch Surg 1995;130:786 92.

1488 Perl et al. CID 1999;29 (December) 12. Baddour LM, Bisno AL. Non group A beta hemolytic streptococcal cellulitis: association with venous and lymphatic compromise. Am J Med 1985;79: 155 9. 13. Auckenthaler R, Hermans PE, Washington JA II. Group G streptococcal bacteremia: clinical study and review of the literature. Rev Infect Dis 1983;5:196 204. 14. Watsky KL, Kollisch N, Densen P. Group G streptococcal bacteremia: the clinical experience at Boston University Medical Center and a critical review of the literature. Arch Intern Med 1985;145:58 61. 15. Vartian C, Lerner PI, Shlaes DM, Gopalakrishna KV. Infections due to Lancefield group G streptococci. Medicine (Baltimore) 1985;64:75 88. 16. Lam K, Bayer AS. Serious infections due to group G streptococci: report of 15 cases with in vitro in vivo correlations. Am J Med 1983;75:561 70. 17. Rolston KVI. Group G streptococcal infections [editorial]. Arch Intern Med 1986;146:857 8. 18. Butt AA, Janney A. Clinical characteristics of group G streptococcal bacteremia. Infect Dis Clin Pract 1998;7:43 8. 19. Skogberg K, Simonen H, Renkonen OV, Valtonen VV. Beta-haemolyticgroup A, B, C, and G streptococcal septicemia: a clinical study. Scand J Infect Dis 1988;20:119 25. 20. Farley MM, Harvey RC, Stull T, et al. A population-based assessment of invasive disease due to group B Streptococcus in non-pregnant adults. N Engl J Med 1993;328:1807 11.