European Journal of Clinical Microbiology and Infectious Diseases

, Volume 25, Issue 10, pp 619–626

Risk factors for bacteremia in patients with limb cellulitis

Authors

    • Internal Medicine ServiceSierrallana Hospital
  • E. Padrón
    • Internal Medicine ServiceSierrallana Hospital
  • M. P. Roiz
    • Microbiology ServiceSierrallana Hospital
  • I. De Benito
    • Microbiology ServiceSierrallana Hospital
  • J. C. Garrido
    • Biochemistry ServiceSierrallana Hospital
  • F. Talledo
    • Internal Medicine ServiceSierrallana Hospital
  • M. J. Rodríguez-Lera
    • Emergency ServiceSierrallana Hospital
  • L. Ansorena
    • Admission ServiceSierrallana Hospital
  • M. B. Sánchez
    • Clinical Pharmacology ServiceUniversity Hospital “Marqués de Valdecilla”
Article

DOI: 10.1007/s10096-006-0186-z

Cite this article as:
Peralta, G., Padrón, E., Roiz, M.P. et al. Eur J Clin Microbiol Infect Dis (2006) 25: 619. doi:10.1007/s10096-006-0186-z

Abstract

The aim of this study was to identify the risk factors for bacteremia in patients with limb cellulitis. Using the administrative and microbiology laboratory databases of a community teaching hospital, a review was conducted of all cases of community-acquired limb cellulitis that occurred during the period 1997–2004 and in which blood cultures had been performed. A comparison of demographical, clinical, and analytical data of patients with bacteremia versus patients without bacteremia was performed by univariate and multivariate analyses. Of 2,678 patients with cellulitis who presented to the hospital’s emergency department, 308 were diagnosed with limb cellulitis and had blood cultures. Of these, 57 (18.5%) had bacteremia. In 24 of the 57 (42.1%) patients with bacteremia, the microorganism isolated in blood cultures was non-group-A β-hemolytic Streptococcus, and in another 14 (24.6%), the microorganism identified was a gram-negative bacterium. Staphylococcus aureus was determined as the cause of bacteremia in just 6 (10.5%) patients and group A Streptococcus in 2 (3.5%). By logistic regression analysis, the following factors were associated with bacteremia: absence of previous antibiotic treatment (OR 5.3, 95% CI 1.4–20.3), presence of two or more comorbid factors simultaneously (OR 4.3, 95% CI 1.6–11.7), length of illness <2 days OR 2.44, 95% CI 1.07–5.56), and proximal limb involvement (OR 6, 95% CI 3.03–12.04). Patients with limb cellulitis who exhibit any of these characteristics are at increased risk of bacteremia. In such patients, it is imperative that blood cultures be performed.

Introduction

Skin and soft tissue infections are commonly seen in clinical practice [1], yet information about the optimal clinical approach to these infections, including the usefulness of an etiological diagnosis and the knowledge of risk factors for bacteremia, is scant [1, 2]. In general, the yield of blood cultures in patients with uncomplicated cellulitis is low and is considered to produce predictable results, usually Staphylococcus aureus and Streptococcus spp. [312], so several authors have recommended that blood cultures not be routinely performed in such patients [3, 4, 9, 11]. Nevertheless, the proportion of bacteremic patients with cellulitis is higher in several series [1337], specifically in patients with lymphedema [13], cirrhosis [14, 15], HIV infection [16], severe soft tissue infections [17, 18], or head and neck cellulitis [2125], and there are a great variety of microorganisms that cause bacteremia in patients with soft tissue infections [1141]. Moreover, a growing incidence of soft tissue infections caused by community-acquired antibiotic-resistant microorganisms, such as methicillin-resistant S. aureus, underscores the relevance of the etiological diagnosis of patients with cellulitis [42, 43]. These facts led us to analyze the incidence of bacteremia among patients with cellulitis and to define the clinical characteristics, the causative microorganisms, and the risk factors for bacteremia.

Patients and methods

Patients

The study was conducted at the Sierrallana Hospital, a 220-bed community hospital. The hospital serves an urban and a rural area with a population of 160,000 inhabitants and has approximately 3,500 admissions and 65,000 presentations to the emergency service annually. Approximately 1,400 blood cultures are performed in the emergency service yearly. To identify all cases with an ICD-9 diagnosis of cellulitis (codes 035, 680, and 681) who attended our hospital emergency service between January 1997 and January 2005, an administrative database was used. The microbiology laboratory database, which contains data of all patients with blood cultures, was then used to select those patients with blood cultures obtained during the previously identified episode of cellulitis. A standardized data collection form was used to review the hospital records (performed by G.P., F.T., and E.P.).

Cellulitis was defined as the presence of signs of cutaneous inflammation of acute onset [10]. Extension of cellulitis to the proximal limb was considered present if the upper arm or the thigh was involved. Trunk extension from the limb was considered present if the cellulitis surpassed proximally the axilla or the groins. Comorbid factors were assessed by using the Charlson comorbidity score [44]. Sepsis, severe sepsis, and septic shock were defined as proposed by Bone et al. [45]. Those patients with hospital-acquired cellulitis, defined as cellulitis that appeared at least 48 h after hospitalization, and those patients with simultaneous limb venous thrombosis or bone or orthopedic implant infection in the limb affected with cellulitis were excluded.

Blood cultures

The recommended practice for blood cultures at our hospital is to obtain 20 ml of venous blood and to inoculate it in equal parts into one aerobic (BacT/ALERT FA Aerobic; bioMérieux, Durham, NC, USA) and one anaerobic blood culture bottle (BacT/ALERT FN; bioMérieux). Blood is extracted by nurses three times from a peripheral vein, using intervals of 30 min. Blood samples obtained from each patient were transported immediately to the microbiology laboratory and loaded into the blood culture instrument (BacT/ALERT microbial detection system, bioMérieux). Cultures were held for 5 days; terminal subcultures from negative culture bottles were not routinely performed unless clinically indicated. Samples from positive culture bottles were further investigated by Gram stain and subculture onto 5% sheep blood, chocolate, and 5% sheep blood/Schaedler agars. These agar plates were incubated in aerobic, 5% CO2, and anaerobic atmospheres at 37°C, respectively. Microorganisms were identified using standard procedures. All catalase-negative samples that were β-hemolytic on 5% sheep blood agar were Lancefield serogrouped using the Slidex Strepto-Kit (bioMérieux) and identified with the API 20 Strep system (bioMérieux). Obvious contaminants, such as isolates of coagulase-negative staphylococci, Bacillus species, diphtheroids, micrococci, and α-hemolytic streptococci, were excluded unless such an organism was isolated from two or more different blood cultures [46, 47]. No method for antibiotic removal was used for blood cultures of patients previously treated with antibiotics.

Statistical analysis

The Student’s t test (two-tailed) was used for the comparison of mean values. Comparisons of categorical data were performed using the two-tailed chi-square test and Fisher’s exact test. Univariate analyses and stepwise logistic regression were performed to identify factors associated with bacteremia. In stepwise regression, factors were added to a base model by forward selection using an F to enter a value of 0.05, followed by a comparison of goodness-of-fit (Hosmer–Lemeshow method). Statistical significance was set at p < 0.05. The statistical analysis was performed using SPSS software, version 12.0 (SPSS Software, Chicago, IL, USA).

Results

Patients

A total of 2,678 patients with cellulitis were evaluated in the emergency service of our hospital during the study period. Four hundred forty-four episodes of non-limb infection were excluded from the analysis. Of the remaining 2,232 episodes of limb cellulitis, blood cultures were performed in 308 (13.8%). General characteristics of patients are shown in Table 1. Mean age was 61 years (range 15–97 years). In 68 (22.1%) cases there was a history of previous cellulitis. Five patients had severe sepsis and none met criteria for septic shock. Immunosuppression was present in seven patients, due to HIV infection in two patients, rheumatoid arthritis in three patients (one treated with methotrexate, one treated with steroids, and another with methotrexate and steroids), and steroid therapy, diffuse B cell lymphoma treated with chemotherapy, and extended myeloma in one patient each. Only one patient had a necrotizing soft tissue infection.
Table 1

General characteristics of the 308 patients with limb cellulitis

Variable

No. (%)

Basal characteristic

 Male

150 (48.7)

 Age >65 years

134 (43.5)

Entry portal

 Any

129 (41.9)

 Trauma

56 (18.1)

 Chronic ulcer on limb

32 (10.4)

 Injecting drug use

8 (2.6)

 Tinea pedis

7 (2.3)

 Animal bite

5 (1.6)

 Diabetic foot

5 (1.6)

 Insect sting

3 (1)

 Psoriasis

3 (1)

Location

 Superior limb

61 (19.8)

 Inferior limb

247 (80.2)

Local predisposing factor

 Entry portal

129 (41.9)

 Edema

77 (25)

 Venous insufficiency

51 (16.6)

 Vascular ulcer

25 (8.1)

 Lymphedema

23 (7.5)

 Lymphadenectomy

20 (6.5)

Comorbid factors

 Obesity

42 (13.6)

 COPD

31 (10.1)

 Diabetes

28 (9.1)

 Renal insufficiency

23 (7.5)

 Alcoholism

17 (5.5)

 Cirrhosis

12 (3.9)

 Cardiac failure

7 (2.3)

 Immunosuppression

6 (1.9)

 Hematological disease

4 (1.3)

 Two or more comorbid factors

26 (8.4)

COPD chronic obstructive pulmonary disease

Microorganisms

Blood cultures were true positive in 57 of the 308 (18.5%) patients. A total of 1,790 blood culture bottles from the 308 patients were sent to the microbiology laboratory (a mean of 5.8 blood culture bottles per patient). In 285 (92.5%) patients, three consecutive blood cultures were performed. In 25 of the 57 (43.9%) true-positive patients, only one blood culture was positive, in 11 patients (19.3%) two blood cultures were positive, and in 21 (36.8%) patients three blood cultures were positive. Table 2 shows the microorganisms isolated in the blood cultures. In 24 of the 57 (42.1%) patients with bacteremia, the microorganism isolated in blood cultures was non-group-A β-hemolytic streptococci, and in another 14 (24.6%), the microorganism identified was a gram-negative bacterium. Organisms considered to be contaminants were isolated from 56 (3.1%) culture bottles. No methicillin-resistant S. aureus and no anaerobes were isolated.
Table 2

Microorganisms isolated in blood cultures from patients with limb cellulitis

Microorganism

No. (%)

Gram-positive

44 (14.3)

 Group C streptococci

10 (3.4)

 Viridans streptococci

9 (3.1)

 Group G streptococci

7 (2.4)

 Group B streptococci

7 (2.4)

 Staphylococcus aureus

6 (2.1)

 Group A Streptococcus

2 (0.7)

 Staphylococcus epidermidis

2 (0.3)

 Staphylococcus lugdunensis

1 (0.3)

Gram-negative

14 (4.5)

 Pseudomonas aeruginosa

3 (0.7)

 Pasteurella multocida

2 (0.7)

 Moraxella spp.

2 (0.7)

 Haemophilus influenzae

2 (0.7)

 Acinetobacter spp.

1 (0.3)

 Alcaligenes xylososidans

1 (0.3)

 Escherichia coli

1 (0.3)

 Klebsiella pneumoniae

1 (0.3)

 Enterobacter aerogenes

1 (0.3)

Totala

58

aIncludes one case of polymicrobial bacteremia caused by S. aureus and P. aeruginosa

Since we observed a very low incidence of S. aureus bacteremia among our patients, we reviewed the records and clinical diagnoses of all patients for whom S. aureus had been isolated in blood cultures during the period of the study in an attempt to discard failures in the detection of patients with cellulitis [5]. Of the 127 patients with S. aureus bacteremia, only six met the selection criteria, and all six had been previously identified.

Risk factors for bacteremia

Patients with bacteremia were statistically significantly older than those without bacteremia (66.7 ± 16.8 years old vs. 59.7 ± 17.9, p = 0.008). Univariate analysis revealed several demographic and clinical factors to be significantly associated with an increased risk of bacteremia (Tables 3 and 4). These included a previous episode of cellulitis, the presence of fever during cellulitis and prior to blood sampling, duration of symptoms less than 2 days, an absence of preculture antibiotic treatment, a temperature greater than 38°C at diagnosis, and previous edema in the limb affected. Of the comorbid factors analyzed, none alone was associated with an increased risk of bacteremia, but the patients with bacteremia more often had two or more comorbid factors. Regarding the extent of the cellulitis, extension to the whole limb, the proximal limb, and the trunk were also significantly more common in the bacteremic patients.
Table 3

Differences in basal and predisposing characteristics between bacteremic and nonbacteremic patients with limb cellulitis

Variable

Nonbacteremic group

Bacteremic group

p value

Total, no. (%)

251 (81.5)

57 (18.5)

 

Age >65 years, no. (%)

103 (41.2)

31 (54.4)

0.05

Female sex, no. (%)

123 (49)

35 (61.4)

0.06

Previous antibiotic, no. (%)

62 (24.7)

3 (5.3)

<0.001

Amoxicillin-clavulanate

35 (13.9)

2 (3.5)

0.02

Ciprofloxacin

8 (3.2)

1 (1.8%)

>0.1

Unknown

7 (2.8)

0

>0.1

Cloxacillin

4 (1.6)

0

>0.1

Erythromycin

2 (0.8)

0

>0.1

Other

6 (2.4)a

0

>0.1

Premorbid conditions

 Charlson score

0.81 ± 1.16

0.53 ± 1.06

0.1

 Obesity, no. (%)

31 (12.4)

11 (19.3)

>0.1

 COPD, no. (%)

25 (10)

6 (10.5)

>0.1

 Diabetes, no. (%)

23 (9.2)

5 (8.8)

>0.1

 Renal insufficiency,  no. (%)

16 (6.4)

7 (12.3)

>0.1

 Alcoholism, no. (%)

14 (5.6)

3 (5.3)

>0.1

 Cirrhosis, no. (%)

8 (3.2)

4 (7)

>0.1

 Cardiac failure, no. (%)

4 (1.6)

3 (5.3)

>0.1

 Immunosuppression,  no. (%)

5 (2.0)

2 (3.5)

>0.1

 Hematological disease,  no. (%)

2 (0.8)

2 (3.5)

>0.1

 Two or more comorbid factors, no. (%)

16 (6.4)

10 (17.9)

0.009

Local predisposing factors

 Entry portal, no. (%)

107 (42.8)

22 (38)

>0.1

 Edema, no. (%)

47 (18.8)

30 (52.6)

<0.001

 Previous cellulitis,   no. (%)

48 (19.1)

20 (35.1)

0.009

 Local surgery, no. (%)

36 (14.4)

8 (14)

>0.1

 Chronic ulcer, no. (%)

24 (9.6)

8 (14.3)

>0.1

 Lymphedema, no. (%)

16 (9.6)

7 (24)

0.03

 Lymphadenectomy,  no. (%)

16 (6.4)

4 (7)

>0.1

 Previous DVT, no. (%)

13 (5.2)

6 (10.5)

>0.1

 Diabetic foot, no. (%)

5 (2)

0

>0.1

 Animal bite, no. (%)

2 (0.8)

3 (5.3)

0.05

 Any local factor, no. (%)

123 (49)

41 (71.9)

0.002

COPD chronic obstructive pulmonary disease, DVT deep venous thrombosis

aAmoxicillin (n = 1), clindamycin (n = 1), clarithromycin (n = 1), piperacillin-tazobactam (n = 1), amoxicillin-clavulanate+levofloxacin (n = 1), amoxicillin-clavulanate+ciprofloxacin (n = 1)

Table 4

Differences in local and general manifestations between bacteremic and nonbacteremic patients with limb cellulitis

Variable

Nonbacteremic group

Bacteremic group

p value

General manifestations

 Previous fever, no. (%)

193 (76.9)

54 (94.7)

0.001

 Temperature >38°Ca,   no. (%)

88 (35.1)

37 (64.9)

<0.000

 Duration of symptoms   <2 days, no. (%)

106 (42.2)

41 (71.9)

<0.000

 Chills, no. (%)

103 (41%)

31 (54.4)

0.07

 Severe sepsis, no. (%)

2 (0.8)

3 (5.3)

0.05

Local manifestations

 Inferior limb location,   no. (%)

196 (78.1)

51 (89.5)

0.03

 Whole limb extension,   no. (%)

20 (8)

19 (33.3)

<0.000

 Proximal extension,   no. (%)

38 (15.1)

29 (50.9)

<0.000

 Trunk extension, no. (%)

4 (1.6)

8 (14)

<0.001

 Local abscess, no. (%)

20 (8.0)

5 (8.8)

>0.1

 Bilateral cellulitis,   no. (%)

9 (3.6)

2 (3.5)

>0.1

aWhen blood cultures were extracted

The comparison of laboratory data is shown in Table 5. No significant differences were detected between bacteremic and nonbacteremic patients, except in the percentage of neutrophils in the leukocyte count and in the albumin concentration in plasma. The following factors were associated with bacteremia by logistic regression analysis: absence of previous antibiotic treatment (OR 5.3, 95% CI 1.4–20.3), existence of two or more comorbid factors simultaneously (OR 4.3, 95% CI 1.6–11.7), duration of illness less than 2 days (OR 2.44, 95% CI 1.07–5.56), and proximal limb involvement (OR 6, 95% CI 3.03–12.04).
Table 5

Comparison of laboratory data for bacteremic and nonbacteremic patients with limb cellulitis

Laboratory data

Bacteremic group

Nonbacteremic group

p value

Leukocyte count, cells/μl

c±/c11,905 ±  6,837

c±/c12,153 ±  5,664

>0.1

Percent neutrophils

c±/c85.1 ± 15.4

c±/c80.2 ± 11.9

0.03

Platelets, cells×1,000/μl

c±/c220.2 ±  117

c±/c225.1 ± 80

>0.1

ESR, mma

c±/c62.3 ± 43.9

c±/c51.3 ± 37.6

>0.1

Creatinine, mg/dl

c±/c1.1 ± 0.4

c±/c1 ± 0.6

>0.1

Albumin, mg/dla

c±/c2.8±0.6

c±/c3.1±0.6

0.01

Cholesterol, mg/dla

c±/c153.3 ±  56.2

c±/c166.1 ± 92.1

>0.1

aPerformed only on hospitalized patients

Evolution of infection

Sixty-eight of the 308 (22.1%) patients did not require hospital admission and were treated as outpatients following their evaluation in the emergency service. The incidence of bacteremia was 21.3% (51/239) among hospitalized patients and 8.8% (6/68) among the ambulatory patients (p = 0.01). Among hospitalized patients, one patient required ICU support and two patients died during the admission, one of whom had diffuse B cell lymphoma treated with chemotherapy and was diagnosed with Pseudomonas aeruginosa bacteremia and the other who had negative blood cultures. Fourteen patients required surgical debridement, ten (4%) in the nonbacteremic group and four (7%) in the bacteremic group (p > 0.1). The antibiotic treatment regimen was changed in 28 (49.1%) patients with bacteremia after the blood culture report was received, in 13 (46.4%) because a bacterium isolated in the blood culture was not covered by the empiric antibiotic regimen, and in 15 (53.6%) because a narrower antibiotic spectrum was desired. No difference was detected in the length of hospitalization between patients with and patients without bacteremia (10.8 ± 6.6 days in the bacteremic group vs. 10.6 ± 7.8 days in the nonbacteremic group).

Discussion

Of the patients with limb cellulitis and blood cultures who presented to our community hospital, 18.5% had true bacteremia. By comparing characteristics of patients with and patients without bacteremia, we identified, through multivariate analysis, several risk factors for bacteremia in patients with cellulitis. As far as we are aware, this is the first study to assess this issue in depth.

Several characteristics of our patients are noteworthy. First, the high incidence of bacteremia that we found contrasts with the low incidence of bacteremia of less than 5% in patients with cellulitis reported by several authors [312], although other studies have reported an incidence of bacteremia similar to or higher than ours [1337]. This great disparity in the incidence of bacteremia reported in different studies may be due to differences in the characteristics of the patients [48, 49]. Other factors, such as differences in the definition of cellulitis [48, 49], the number of blood cultures per patient [50], or the delay before incubation [51], may influence the yield of blood cultures and could contribute to the differences.

A variety of factors may influence the incidence of bacteremia in soft tissue infections, such as premorbid conditions, the clinical characteristics of the acute illness, and the causal microorganisms. Only one study has previously evaluated risk factors for bacteremia in adult patients with cellulitis. This work was a cost-effectiveness study which, in a secondary analysis using univariate analysis, detected that older age, a shorter duration of symptoms, the existence of fever, and an elevated leukocyte count at admission were associated with the presence of bacteremia [3]. Among children, younger age, the presence of high fever, and a left shift in the blood cell count were associated with bacteremia [4].

Several studies that have addressed the characteristics of soft tissue infections in subgroups of patients with specific comorbid factors have found a higher incidence of bacteremia. For example, Manfredi et al. [16] found the incidence of bacteremia in 67 HIV-infected patients with cellulitis to be 25%. In two studies of cirrhotic patients with cellulitis caused by gram-negative organisms, 75% (9/12) of patients were bacteremic [14, 15]. Furthermore, some predisposing local factors could probably increase the risk of bacteremia, as reported by Woo et al. [13], who detected bacteremia in 3 of 10 (33%) patients with previous local lymphedema. Talan et al. [33] detected positive blood cultures among 1 of 14 (7.1%) patients with infected dog or cat bites. Infections of greater severity may also be associated with a higher rate of bacteremia, as addressed in a study of patients with severe soft tissue infections who required intensive care: the incidence of bacteremia was over 45% overall and was 55% among patients with necrotizing fasciitis [17].

Considering all the factors that may influence the development of bacteremia during cellulitis, it is not surprising that rates of bacteremia in cellulitis vary between series. For example, Carratalá et al. [29] found an incidence of bacteremia of 18.7%, comparable to ours, among 251 adults hospitalized for cellulitis in whom both the incidence of comorbid factors and the incidence of gram-negative infections were high. Björnsdóttir et al. [19] reported an incidence of bacteremia of 16% in a recent prospective study that included 81 patients with cellulitis and blood cultures. This contrasts with the results of Perl et al. [3], who, in a large series that assessed the yield of blood cultures in 533 patients with cellulitis, found an incidence of just 2%. In that study, however, the incidence of comorbid factors was low and the extent of the cellulitis was not addressed.

Specific locations of cellulitis are also associated with higher rates of bacteremia, such as periorbital cellulitis in children [2125], head and neck cellulitis in adults (with an incidence of bacteremia of 35% in one study [27]), or decubitus ulcer infections with sepsis, which Galpin et al. [35] found to be associated with bacteremia in 75% of cases. Finally, the causal microorganisms can also influence the rate of bacteremia, and some bacteria cause bacteremic cellulitis more often. In severe infections by group A Streptococcus, a frequency of bacteremia of 60% has been reported [18]. Soft tissue infections by Stenotrophomonas maltophilia are associated with positive blood cultures in 29% of patients [20], and several authors have reported an incidence of bacteremia of around 100% in cellulitis caused by Streptococcus pneumoniae [24, 30], Vibrio vulnificus [26], or Streptococcus iniae [36]. Cases of Haemophilus influenzae cellulitis also are usually bacteremic [25].

It must be noted that the incidence of bacteremia caused by group A streptococci in our study was low, despite group A streptococci being a recognized causal agent of cellulitis [1], which raises the question of whether cellulitis caused by this microorganism is usually nonbacteremic. Moreover, we found a high incidence of non-group-A β-hemolytic streptococcal cellulitis and of gram-negative cellulitis, which can probably be explained by the underlying conditions in our patients. Other studies have described an increasing incidence of cases of cellulitis with bacteremia caused by non-group-A β-hemolytic streptococci [4, 19, 29, 34, 3841] and of gram-negative bacteremia in patients with severe infections or comorbid factors [1329]. The isolation of gram-negative organisms as causal pathogens of cellulitis is of special practical relevance because the empiric antibiotic therapy for cellulitis is usually directed toward gram-positive bacteria. In fact, in our study, antibiotic therapy was changed in 49.1% of the bacteremic patients after the results of blood cultures became available.

Our study has the limitations of a retrospective design: data of the patients’ characteristics are limited to patient records, and some cases of cellulitis may have been missed. Another problem is the limited number of patients with some of the predisposing factors that were analyzed, which could impede the detection of statistically significant differences. It must be noted that, in our study, patients with blood cultures could represent a selected group of cases with more severe cellulitis (13.8% of the total), which may have contributed to the high incidence of bacteremia found. Moreover, because blood specimens for culture were obtained from only a small fraction of the patients who presented to our hospital due to cellulitis and because the study was performed in a community hospital with a relatively low incidence of patients with immunosuppression, any generalization of these findings to other populations should be done with caution.

Previous guidelines have recommended that blood cultures not be performed in patients with cellulitis, due to the low incidence of bacteremia and the predictable causal microorganisms that make antibiotic empiric therapy safe for this infection [112]. In contrast, one author has recommended that blood cultures be performed in patients with (a) cellulitis superimposed on lymphedema, (b) buccal or periorbital cellulitis, (c) an infection whose likely source is saltwater or fresh water, or (d) high fever or chills, signs that are classically associated with bacteremia or severe illness [1]. Perl et al. [3] recommend limiting blood cultures to the most severe cases of cellulitis, i.e. those characterized by acute onset, high-grade fever, significant leukocytosis, immunosuppression, or advanced age of the patient. The Infectious Disease Society of America guidelines state that “blood cultures are not fruitful for the typical case of cellulitis, unless it is particularly severe [49]. Most of these recommendations are not based on studies that specifically analyzed risk factors for bacteremia.”

Our data indicate that several factors are indicators of a higher risk of bacteremia in patients with cellulitis, such as the absence of antibiotic treatment, the coexistence of several comorbid factors, an acute abrupt onset of cellulitis, and proximal limb involvement. Thus, blood cultures are recommended in patients with cellulitis who exhibit any of these characteristics.

Acknowledgment

We thank Luis Martínez (University Hospital “Marqués de Valdecilla”) for critical reading of the manuscript.

Copyright information

© Springer-Verlag 2006