Secondary Logo

Journal Logo

Predictive Value of Leukocytosis and Neutrophilia for Bloodstream Infection

Walling, Hobart W. MD, PhD; Manian, Farrin A. MD

Infectious Diseases in Clinical Practice: January 2004 - Volume 12 - Issue 1 - p 2-6
doi: 10.1097/01.idc.0000104893.16995.0a
Original Article

Background Bloodstream infections are an important cause of morbidity and mortality in hospitalized adults. We sought to identify laboratory parameters predictive of blood culture-positivity.

Methods We generated a database of patients having positive blood cultures during a 6-month period at our tertiary care community-based medical center. From this patient population, further instances of positive and negative blood culture were identified retrospectively (up to 60 months prior). Differences in laboratory parameters by culture result were studied.

Results Of 766 blood cultures, 278 were positive, 395 were negative, and 93 were contaminated. Compared to patients with negative blood cultures, those with positive blood culture had significantly higher WBC count (15,160 ± 460 μl vs. 10,300 ± 200 μl), percent PMN (84.5 ± 0.6% vs. 77.9 ± 0.6%), absolute neutrophil count (12,950 ± 420 μl vs. 8200 ± 200 μl), and bandemia (8.5 ± 1.3% vs. 4.6 ± 0.9%; P = 0.0178). Calculated odds ratios (OR) for positive blood culture included WBC count ≥ = 12,000: OR = 3.16 (95% CI, 2.28–4.37); %PMN ≥ = 80: OR = 4.16 (95% CI, 2.89–5.98), and band PMN ≥ 5%: OR = 3.25 (95% CI, 1.59–6.64).

Conclusion Presence of leukocytosis, neutrophilia, and bandemia is highly correlated with blood culture positivity (OR 4.26; 95% CI, 2.48–7.32; RR 49.5). Absence of these findings is inversely associated with blood culture positivity (OR 0.086; 95% CI, 0.044–0.17). A quantitative method of risk assessment is provided based on these parameters to expedite diagnosis and treatment of bloodstream infection.

St. John’s Mercy Medical Center, St. Louis, MO 63141.

Address correspondence and reprint requests to Hobart W. Walling, Department of Dermatology, University of Iowa Hospitals and Clinics, Iowa City, IA 52242. E-mail: walling@SLU.EDU.

Bloodstream infection remains an important cause of morbidity and mortality in hospitalized adults. Such infections are implicated as the 13th leading cause of death and have increased in both number and severity in recent decades. 1 Blood cultures are generally ordered based on physicians’ clinical judgment, although their estimates of the chance of bacteremia are often inaccurate. 2 Clinical data obtainable at the bedside, including altered temperature, pulse, respiratory rate, and blood pressure have had disappointing value in predicting severe infection. 3,4 Moreover, bloodstream infection may occasionally be overlooked in an ambulatory setting, leading to inappropriate discharge of seriously ill patients. 5 Only 5 to 15 percent of blood cultures reveal the presence of true blood pathogens, and up to 5.4% of cultures grow contaminating organisms. 6 False-positive cultures may prompt further testing, unneeded therapy, and prolonged hospitalization.

Early prediction of bloodstream infection prior to availability of microbiologic testing would be useful in the decision to obtain blood culture and assess the need for therapeutic intervention. The goal of this study was to identify objective laboratory findings that predict true bloodstream infection and to stratify patients into risk groups according to these findings. Identifying patients with a low risk of bloodstream infection may obviate the need for blood culture and remove the clinical uncertainty inherent in a subsequent false-positive culture. Conversely, identifying patients at high risk of bacteremia may assist in the decision to begin empiric antimicrobial therapy and guide supportive care.

Back to Top | Article Outline


We identified 393 adult inpatients (age 18–97) with positive blood cultures during a 6-month period (1/1/2001 to 6/30/2001) at our tertiary care community-based medical center. We then searched our computerized patient care database (Powerchart) for laboratory data at the time surrounding the blood draw yielding a positive culture. We also searched for subsequent blood cultures (positive and negative) during the same hospitalization and retrospectively searched the database of each identified patient for blood cultures (positive or negative) drawn during the preceding 4 years. Inclusion criteria were a complete blood count (CBC), including white blood cell (WBC) count with differential, drawn no greater than 8 hours prior to or 4 hours after blood culture. Patients’ gender, age, and day of hospitalization were also recorded. For each patient, additional sets of blood cultures drawn less than 36 hours after a prior culture were excluded from analysis. Negative cultures obtained less than 60 hours after a positive culture (time selected arbitrarily) were also excluded. Neutropenic patients (WBC count < 3000 μl) were excluded, as were patients with marked leukophilia (WBC count > 45,000 μl) suggestive of underlying hemoproliferative disorder. To remove potentially confounding variables, patients were not subcategorized by primary diagnosis, fever, use of antibiotics, or other medications (eg, corticosteroids), presence of invasive devices, or outcome of hospitalization. The patients with positive versus negative cultures were well-matched demographically, as each patient generally served as his or her own control.

CBC and automated differentials were performed on a Coulter (Beckman) system. Identification of immature leukocytes or left-shift prompts a manual differential. Normal ranges at our institution are WBC: 4500–9500 μl and %PMN 45–70%. Blood cultures are obtained by nursing staff from peripheral sites after betadine preparation and are subsequently incubated in a Bactec 9240 incubator (Becton-Dickinson). Antibiotic susceptibilities are determined with Vitek systems (BioMerieux) test cards. Any pathogenic organism identified from at least 1 of 2 blood culture bottles was considered a positive blood culture. Coagulase-negative staphylococcus, diphtheroids, and Bacillus species were considered positive only if present in 2 or more culture bottles with identical susceptibility profiles and were categorized as contaminated when present in a single-culture bottle. A negative culture displayed no growth after 3 days of incubation.

Data were collected and categorized. Mean values and standard errors were calculated using Microsoft Excel 97. Statistical analysis consisted of Student t test (calculated using SPSS) for continuous variables and Pearson χ2 test for categorical data, with P < 0.05 considered statistically significant.

Back to Top | Article Outline


During the study period, 445 blood culture-positive episodes yielding 446 microorganisms occurred in 394 patients. Of this total, 37 patients (9.4%) were excluded for neutropenia and 29 (7.4%) were excluded for lack of data. In the remaining 328 patients, 371 blood culture-positive episodes occurred, yielding 391 microorganisms. Of these, 278 (74.9%) were true bloodstream infection, while 93 (25.1%) were classified as contaminated. From the same population of patients, data relating to 395 negative blood cultures were obtained.

Of the 278 true positive cultures, 133 (47.8%) were obtained from males and 145 (52.2%) were obtained from females. Of the total, 188 (67.6%) were Gram-positive bacteria and 80 (28.8%) were Gram-negative bacteria (Table 1). Staphylococcus aureus was the most common isolate. Coagulase-negative staphylococcus was the next most common isolate and was the most common isolate if contaminated cultures are allowed into the analysis. Escherichia coli was the third most common organism overall and was the predominant Gram-negative organism.



We compared the laboratory data of patients with positive blood cultures to those with negative cultures (Table 2). Compared to the culture-negative group, patients with positive blood cultures had significantly higher WBC counts (15,160 ± 460 μl vs. 10,300 ± 200 μl), percent neutrophils (PMN; 84.5 ± 0.6% vs. 77.9 ± 0.6%), and absolute neutrophil count (12,950 ± 420 μl vs. 8200 ± 200 μl; all P < 0.0001; see Fig. 1). Positive blood culture was also associated with a significant increase in percentage of immature (band) neutrophils (8.5 ± 1.3% vs. 4.6 ± 0.9%; P = 0.0178) and with an increased percentage of patients exhibiting bandemia (45% vs. 24.7%; P < 0.001). Data from the contaminated cultures generally mirrored that of the culture negative group.





We next compared data associated with Gram-positive vs. Gram-negative organisms. Gram-negative bacteremia occurred in older patients (66.5 ± 1.8 years vs. 59.9 ± 1.4 years; P = 0.0104) and was associated with higher WBC count (16,700 ± 900 μl vs. 14,200 ± 500 μl) and percentage neutrophils (86.9 ± 0.8% vs. 83.5 ± 0.7%; P < 0.05 in each case), as well as higher absolute neutrophil count (14,500 ± 800 μl vs. 12,100 ± 500 μl; P < 0.01).

We next generated odds ratios (OR) for bacteremia in patients with a clinical indication for obtaining blood culture. The OR for varying levels of leukocytosis, neutrophilia, and bandemia by univariate analysis are shown (Table 3). Increasing WBC count was associated with increasing OR of bacteremia (3.16 at ≥12,000 μl vs. 5.88 at ≥20,000 μl). Neutrophilia was also predictive of bloodstream infection (OR 3.17–4.16). Bandemia was highly correlated with bacteremia, particularly when band cells represented at least 10% of the total neutrophil count (OR 7.89). In multivariate analysis, leukocytosis and/or neutrophilia correlated with increased incidence of bacteremia. Advanced age and female gender were modest risk factors for bacteremia.



We finally analyzed patients in each of 3 categories (Table 4): presence or absence of WBC count ≥ 12,000 μl, percent neutrophils ≥ 80%, and band neutrophils ≥ 5%. When none of these criteria were present, 11.2% of blood cultures were positive (OR = 0.086), constituting the low-risk category. When all 3 criteria were present, 75% of patients had bloodstream infections (OR = 4.25), constituting the high-risk category (RR = 49.5 vs. all negative criteria). When only 1 of the 3 criteria were present, culture-positivity ranged from 25% to 47.5% (OR = 0.43–1.24; RR = 5.0–14.4), with neutrophilia the most important predictor. When 2 of the 3 criteria were present, culture positivity ranged from 53.3% to 61.3% (OR = 1.7–3.92; RR = 19.8–45.6).



Back to Top | Article Outline


In this report, we show that data obtainable on a CBC are predictive of blood culture positivity. Leukocytosis, neutrophilia, and bandemia are reliable indicators of bloodstream infection. We placed patients in risk categories based on WBC count ≥ 12,000 μl, neutrophilia ≥ 80%, and bandemia ≥ 5%. These cutoffs provided the best discrimination between positive and negative cultures in our data set. When none of these criteria are present, the likelihood of bloodstream infection is low and clinicians may choose to defer blood culture in favor of observation. When all 3 are present, the likelihood of bloodstream infection is significant (ie, nearly 50-fold greater than when the above CBC parameters are normal) and blood culture is likely to have high yield. Clinical judgement is of special importance when 1 or 2 of these findings are present.

Although vital signs are an essential part of physical assessment, their deviation from normal may not reliably indicate infection. 3 Fever in the hospitalized patient commonly prompts blood culture, but this variable alone predicts bacteremia in only 2.4-10.2% of cases. 3,7 Although fever in medical patients most commonly arises from microbial infection, it may also occur with drug reaction, malignancy, thromboembolism, and ischemia. 8 Fever is particularly nonspecific in the early postoperative period. 9,10

Leukocytosis is generally considered a sign of microbial infection but has been an inconsistent predictor of bacteremia. Several studies have correlated leukocytosis with bacteremia, 4,7,11 while others have not. 3,5,10,12,13 In 1 study, band count ≥ 1500 mm3 and ESR ≥ 30 mm/h predicted focal bacterial infection but not bacteremia. 3 Others have reported that age over 50, elevated temperature, rigors, intravenous drug abuse, acute abdomen, diabetes mellitus, major comorbidity, hypotension, tachycardia, band count ≥ 1500 mm3 or >20%, elevated alkaline phosphatase, low serum albumin, bacteriuria, immunosuppression, ICU hospitalization, and lack of antibiotics are independent predictors for bacteremia, 3,7,11,14,15,16 although none of these factors was uniformly predictive in each study. Anemia, hyponatremia, and presence of a central venous catheter have been associated with S. aureus bacteremia. 17

Additional laboratory variables recently correlated with bacteremia include plasma procalcitonin, neutrophilic elastase-α1-antitrypsin, and lactoferrin, 4 as well as tumor necrosis factor-α, interleukin-6, phospholipase A2, and complement product C3A. 18,19,20 C-reactive protein has a variable correlation with bacteremia. 18,20,21

Our data must be viewed from the standpoint of several limitations. First, all data were derived from patients in whom blood culture was actually performed and in whom at least a single episode of bloodstream infection was documented. As such, selection bias could make our findings less applicable to patients at low risk for bacteremia. Second, although our study population was heterogeneous, we excluded patients with marked leukopenia (WBC count < 3000) or leukocytosis (WBC count > 45,000) from our analysis and thus may have missed patients with high illness acuity. Although at least 1 study has reported that a low WBC count does not portend a higher rate of bacteremia, 20 others have shown that neutropenia carries a greater risk for infection, particularly fungal or polymicrobial infection. 22 It is certainly reasonable to have a lower clinical threshold to obtain blood cultures in these patients.

Third, our study analyzed patients having positive blood cultures without regard to primary illness, comorbidities, or hospital service. Such variables may have value in predicting bacteremia. 23,24 On the other hand, using only objective data may lend broader clinical applicability to our findings. Although not studied here, other variables that might have predictive value for bacteremia include culture time until positivity, presence and nature of antibiotic therapy, and result of recent prior blood culture.

In summary, our findings show that optimal utilization of blood culture may be furthered by attention to a concurrent CBC. Our stratification based simply on WBC count and differential may allow discrimination between patients with low-risk and high-risk of bacteremia. In low-risk patients, it may be cost-effective to hold a specimen until CBC results are available and to proceed with culture only if predictive findings are present. Conversely, patients identified as high risk (ie, those with WBC count ≥ 12,000 with significant bandemia and neutrophilia) may benefit from timely institution of empiric antibiotic prior to availability of full microbiologic data.

Back to Top | Article Outline


1. National Center for Health Statistics. Month Vital Sta Rep. 1995; 43:5–7.
2. Makadon HJ, Bor D, Friedland G, et al. Febrile inpatients: house officers’ use of blood cultures. J Gen Intern Med 1987; 2:293–297.
3. Mellors JW, Horwitz RI, Harvey MR, et al. A simple index to identify occult bacteria infection in adults with acute unexplained fever. Arch Intern Med. 1987; 147:666–671.
4. Bossink AJW, Groeneveld ABJ, Thijs LG. Prediction of microbial infection and mortality in medical patients with fever: plasma procalcitonin, neutrophilic elastase-(1-antitrypsin, and lactoferrin compared with clinical variables. Clin Infect Dis. 1999; 29:398–407.
5. Epstein D, Raveh D, Schlesinger Y, et al. Adult patients with occult bacteremia discharged from the emergency department: epidemiological and clinical characteristics. Clin Infect Dis. 2001; 32:559–565.
6. Schifman RB, Strand CL, Meier FA, et al. Blood culture contamination: a College of American Pathologists Q-Probes study involving 640 institutions and 497,134 specimens from adult patients. Arch Pathol Lab Med. 1998; 122:216–221.
7. Bates DW, Cook EF, Goldman L, et al. Predicting bacteremia in hospitalization patients. A prospectively validated model. Ann Intern Med. 1990; 113:495–500.
8. Arbo MJ, Fine MJ, Hanusa BH, et al. Fever of nosocomial origin: etiology, risk factors and outcomes. Am J Med. 1993; 95:505–512.
9. Theuer CP, Bongard FS, Klein SR. Are blood cultures effective in the evaluation of fever in perioperative patients? Am J Surg. 1991; 162:615–618.
10. Fanning J, Neuhoff RA, Brewer JE, et al. Frequency and yield of postoperative fever evaluation. Infect Dis Obstet Gynecol. 1998; 6:252–255.
11. Leibovici L, Greenshtain S, Cohen O, et al. Bacteremia in febrile patients. A clinical model for diagnosis. Arch Intern Med. 1991;151:1801–1806.
12. Peduzzi P, Shatney C, Sheagren J, et al. Predictors of bacteremia and gram-negative bacteremia in patients with sepsis. The Veterans Affairs Systemic Sepsis Cooperative Study Group. Arch Intern Med. 1991; 114:332–333.
13. Schwenzer KJ, Gist A, Durbin CG. Can bacteremia be predicted in surgical intensive care unit patients? Intensive Care Med. 1994; 20:425–430.
14. Imperiale TF, Speroff T, Wolinsky E. A predictive index for bacteremia in medical inpatients. Clin Res. 1993; 41:519A.
15. Mozes B, Milantiner D, Block C, et al. Inconsistency of a model aimed at predicting bacteremia in hospitalized patients. J Clin Epidemiol. 1993; 46:1035–1040.
16. Pfitzenmeyer P, Decrey H, Auckenthaler R, et al. Predicting bacteremia in older patients. J Am Geriatr Soc. 1995; 43:230–235.
17. Jensen AG, Wachmann CH, Pulsen KB, et al. Risk factors for hospital-acquired Staphylococcus aureus bacteremia. Arch Intern Med. 1999; 159:1437–1444.
18. Byl B, Deviere J, Saint-Hubert F, et al. Evaluation of tumor necrosis factor-alpha, interleukin-6 and C-reactive protein plasma levels as predictors of bacteremia in patients presenting signs of sepsis without shock. Clin Microbiol Infect. 1997; 3:306–316.
19. Groeneveld AB, Bossink AW, van Mierlo GJ, et al. Circulating inflammatory mediators in patients with fever: predicting bloodstream infections. Clin Diagn Lab Immunol. 2001; 6:1189–1195.
20. Rintala EM, Aittoniemi J, Laine S, et al. Early identification of bacteremia by biochemical markers of systemic inflammation. Scand J Clin Invest. 2001; 61:523–530.
21. Povoa P. C-reactive protein: a valuable marker of sepsis. Intensive Care Med. 2002; 28:235–243.
22. Weinstein MP, Reller LB, Murphy JR, et al. The clinical significance of positive blood cultures: a comprehensive analysis of 500 episodes of bacteremia and fungemia in adults. I. Laboratory and epidemiologic observations. Rev Infect Dis. 1983; 5:35–53.
23. Pittet D, Herwaldt LA, Massanari RM. The intensive care unit. In: Brachman PS, Bennett JV, eds. Hospital infections, 3rd ed. Boston: Little, Brown, 1992;405–439.
24. Bates DW, Sands K, Miller E, et al. Predicting bacteremia in patients with sepsis syndrome. J Infect Dis. 1997; 176:1538–1551.
25. Cockerill 3rd FR, Hughes JG, Vetter EA, et al. Analysis of 281,797 consecutive blood cultures performed over an eight year period: trends in microorganisms isolated and the value of anaerobic culture of blood. Clin Infect Dis. 1997; 24:403–418.
© 2004 Lippincott Williams & Wilkins, Inc.