a Division of Infectious Diseases and Hospital Epidemiology, University Hospital Basel, Switzerland
b University of Basel, Switzerland
c Department of Internal Medicine, University Hospital Basel, Switzerland
d Division of Clinical Microbiology, University Hospital Basel, Switzerland
BACKGROUND AND AIMS
The widely used Pitt Bacteraemia Score (PBS) has repeatedly been described as a risk factor for short-term mortality in Staphylococcus aureus bloodstream infection (BSI), but little is known about its overall predictive performance. We therefore aimed to externally validate the PBS in S. aureus BSIs.
We performed a retrospective validation study at the University Hospital Basel. Adult patients with a first episode of methicillin-susceptible S. aureus BSI between January 2008 and December 2013 were eligible for the study. We measured the overall discriminative power of the PBS at day of BSI onset in predicting 30-day all-cause mortality by receiver-operating characteristics analysis. For each PBS cut-off, we calculated the corresponding sensitivity, specificity and predictive values for prediction of 30-day all-cause mortality.
A total of 329 patients were included in the final analysis: The median PBS at BSI onset was 0 (interquartile range, 0–2) with patients suffering from various comorbidities (Charlson Comorbidity Index median 3, interquartile range 1–5). Thirteen percent of patients (43/329) died within 30 days from any cause. At BSI onset, 52% (170/329) of patients had a PBS of zero; the concomitant specificity and positive predictive value for prediction of 30-day all-cause mortality were 0% and 13%, respectively. The overall performance of the PBS in predicting the 30-day all-cause mortality was lower than published, with an area under the curve of 0.711 (95% confidence interval 0.614–0.807; p <0.001).
For short-term mortality, the PBS had a low predictive value in a patient population with methicillin-susceptible S. aureus BSI. There is a need to improve simple clinical scores to better predict mortality, in particular for S. aureus.
Keywords: Bloodstream infection, predictive performance, Pitt Bacteraemia Score, Staphylococcus aureus, validation
Staphylococcus aureus is a leading cause of bloodstream infections (BSIs), associated with a high mortality in spite of adequate antimicrobial and supportive measures [1–3]. Hence, mortality risk scores in S. aureus BSI may have a profound impact on clinical decision support and risk adjustment in observational studies.
The Pitt Bacteraemia Score (PBS) is a widely used severity-of-illness score, which is mainly used to estimate short-term mortality in gram-negative BSIs [4, 5] and is not pathogen-specific. The PBS is calculated at initial patient evaluation from distinct clinical variables (range 0–14 points): temperature of 35.1‒36.0°C or 39.0‒39.9°C (1 point), temperature of ≤35°C or ≥40°C (2 points), mental status (alert, 0 points; disoriented, 1 point; stuporous, 2 points; comatose, 4 points), hypotension (2 points), receipt of mechanical ventilation (2 points) and cardiac arrest (4 points) . In S. aureus BSIs, the PBS has repeatedly been reported to reflect risk for short-term mortality [6–8], but little is known about its overall predictive performance, as studies validating the PBS and modified versions of the PBS have been performed solely in patients with intensive care unit-acquired sepsis and in patients with gram-negative BSIs [9–11]. We therefore aimed to externally validate the PBS in S. aureus BSIs.
Materials and methods
This was a retrospective validation study performed at the University Hospital Basel, an 800-bed tertiary care centre in Northwestern Switzerland. We extracted relevant data from our prospective in-house BSI database, which includes microbiological, laboratory and clinical data of all patients with a positive blood culture.
Patients aged ≥18 years with a first episode of methicillin-susceptible S. aureus BSI between January 2008 and December 2013 were eligible for the study. We excluded patients with a missing PBS at BSI onset. The Ethics Committee of Northwestern and Central Switzerland approved this study (EKNZ BASEC 2016–01515).
A S. aureus BSI was defined as the growth of S. aureus in one or more blood cultures. The time-stamp of the first positive blood culture by the in-house laboratory was defined as the day of BSI onset. Onset of BSI after more than two days of hospitalisation was interpreted as hospital-acquired BSI. Immunosuppression was defined as stated elsewhere . The PBS at BSI onset was calculated retrospectively from clinical information in written/electronic medical records and our in-house BSI database .
During the entire study period, rapid processing of the blood cultures round the clock was ensured by use of one blood culture system (BacT/ALERT; bioMérieux, Durham, NC, USA).
At our institution, beta-lactam antibiotics constitute the first-line empirical treatment for patients presenting with sepsis. Antimicrobials with activity against methicillin-resistant S. aureus (MRSA) are added only in patients who are known carriers or patients with recent hospitalisation in a high-burden region.
For our primary outcome parameter, 30-day all-cause mortality, we measured the overall discriminative power of the PBS at BSI onset by receiver-operating characteristics analysis, with an area under the curve (AUC) of 0.5 indicating a random prediction and a value of 1.0 denoting a perfect prediction; for the validation, only the highest PBS was taken at BSI onset. For each PBS cut-off, we calculated the corresponding sensitivity, specificity and predictive values for prediction of 30-day all-cause mortality. In a univariable logistic regression model, we assessed the calibration of the PBS using the Hosmer-Lemeshow goodness-of-fit statistic  with a p-value ≥0.05 reflecting a good model fit. We analysed all data using SPSS, version 22 (IBM, Chicago, IL, USA).
In the 6-year study period, 7 of 336 eligible patients were excluded because of a missing PBS at onset of S. aureus BSI; the remaining 329 patients were included in the final analysis. The median age of the study population was 67 years (interquartile range [IQR] 51–78 years) (table 1). The median PBS at BSI onset was 0 (IQR 0–2), with patients suffering from various comorbidities (Charlson Comorbidity Index median 3, IQR 1–5). Overall, the most frequent source of BSI was skin/soft tissue (34%; 113/329) and intravascular catheters / foreign materials (27%; 90/329). At BSI onset, 98% of patients (316/323) received adequate empirical antimicrobial therapy. Thirteen percent of patients (43/329) died within 30 days from any cause.
|Age in years, median (IQR)||67 (51‒78)|
|Female gender, n (%)||122 (37)|
|Risk scores and comorbidities|
|Pitt Bacteraemia Score at BSI onset, median (IQR)||0 (0‒2)|
|Charlson Comorbidity Index at BSI onset, median (IQR)||3 (1‒5)|
|Diabetes mellitus type 2, n (%)||81 (25)|
|Any surgery in the last 30 days*, n (%)||85 (26)|
|Intravenous drug use, n (%)||46 (14)|
|Immunosuppression†, n (%)||155 (47)|
|Infection and treatment characteristics|
|Hospital-acquired BSI, n (%)||114 (35)|
|Focus of BSI, n (%)|
|Intravascular catheters / foreign material||90 (27)|
|Respiratory tract||25 (8)|
|Skin and soft tissue||113 (34)|
|Adequate empirical antimicrobial therapy, n (%)||316 (98)‡|
|30-day all-cause mortality, n (%)||43 (13)|
At BSI onset, 52% (170/329) of patients had a PBS of zero; the concomitant specificity and positive predictive value for prediction of 30-day all-cause mortality were 0% and 13%, respectively (table 2). One percent of patients (4/329) exhibited a PBS of ≥7 points, and 3 of these 4 patients died within 30 days from any cause. The PBS cut-off with the highest combined sensitivity and specificity for prediction of 30-day all-cause mortality was ≥2 (sensitivity, 63%; specificity, 76%).
|Score* at BSI onset||Total|
|Death within 30 days|
% (95% CI)
% (95% CI)
% (95% CI)
% (95% CI)
|0||170 (52)||13 (8)||100 (75–100)||0 (0–1)||13 (10–17)||–|
|1||63 (19)||3 (5)||70 (54–83)||55 (49–61)||19 (13–26)||92 (87–96)|
|2||54 (16)||8 (15)||63 (47–77)||76 (70–81)||28 (19–38)||93 (89–96)|
|3||10 (3)||4 (40)||44 (29–60)||92 (88–95)||45 (30–61)||92 (88–95)|
|4||15 (5)||7 (47)||35 (21–51)||94 (91–96)||47 (29–65)||91 (87–94)|
|5||6 (2)||2 (33)||19 (8–33)||97 (94–99)||47 (23–72)||89 (85–92)|
|6||7 (2)||3 (43)||14 (5–28)||98 (96–99)||55 (23–83)||88 (84–92)|
|≥7||4 (1)||3 (75)||7 (1–19)||100 (98–100)||75 (19–99)||88 (84–91)|
In receiver-operating characteristics analysis, the AUC of the PBS in predicting the 30-day all-cause mortality was 0.711 (95% confidence interval [CI] 0.614–0.807; p <0.001) (fig. 1). Hosmer-Lemeshow statistics revealed a good calibration of the PBS with a corresponding insignificant p-value (chi-square goodness-of-fit test = 2.9, p = 0.234).
This is the first study investigating the discrimination and calibration of the PBS in S. aureus BSIs. More than 50% of patients had zero scoring points at BSI onset, which resulted in a low predictive value of the non-pathogen-specific PBS. Our study population was very homogenous: a single pathogen was involved, almost all antibiotics given for empirical therapy were active against S. aureus, MRSA cases were excluded since empirical therapy did not cover MRSA, and vancomycin is considered to be inferior in terms of activity against S. aureus compared to beta-lactams. Therefore, the circumstances for validation of the PBS are in favour of such a score, because confounding variables such as different pathogens with variable susceptibility patterns, sources of infection and classes of antimicrobial therapy have been minimised.
Our results challenge other, mainly small observational, studies indicating that the PBS is –at least in high-risk populations –a good predictor of death in BSI or sepsis: In a Korean study, the PBS had a high discriminative power in predicting mortality in intensive care unit-acquired sepsis (AUC 0.8, 95% CI 0.7–0.9), which was comparable to the performance of the Charlson Comorbidity Index and the Acute Physiology and Chronic Health Evaluation II scoring system . In a prospective study performed on an adult population in New Zealand, a PBS ≥2 was independently associated with death related to S. aureus BSI (overall 30-day mortality 19%) . These results are in line with investigations in emergency room patients with S. aureus BSI and in patients with haematogenous S. aureus meningitis, in which the PBS was an independent predictor of short-term mortality [7, 14].
Variables other than acute severity of illness have been associated with mortality in S. aureus BSI, such as age, presence of comorbidities and source of infection . Inclusion of such strong predictors of all-cause and infection-related 30-day mortality in S. aureus BSI could significantly improve the discriminative power of the PBS. For gram-negative BSIs, the PBS has already been modified by addition of distinct clinical variables, which resulted in an excellent predictive performance of the adapted PBS in external validation (AUC for 28-day mortality 0.9) . In line, modification of the PBS parameters and its associated scoring weights could improve the predictive accuracy of the PBS in gram-positive BSIs.
Our study has limitations. First, we investigated the predictive performance of the PBS solely in S. aureus BSI; however, for infectious agents with lower pathogenicity and concomitant lower case fatality rates, the discriminatory capacity of the PBS might be lower, as even a higher proportion of patients could have a low PBS of 0 or1 points. Second, we did not analyse the predictive power of the different PBS components. Third, we did not compare the predictive performance of the PBS with other well-established clinical scores.
In conclusion, for short-term mortality, the PBS had a low predictive value in a patient population with methicillin-susceptible S. aureus BSI. There is a need to improve simple clinical scores to better predict mortality, in particular for S. aureus.
This work was funded by the Division of Infectious Diseases and Hospital Epidemiology, University Hospital Basel, Basel, Switzerland. No external funding was received. The authors declare no conflicts of interest.
Andreas F. Widmer, MD, MSc, Division of Infectious Diseases and Hospital Epidemiology, University Hospital Basel, Petersgraben 4, CH-4031 Basel, andreas.widmer[at]usb.ch
2 López-Cortés LE, Del Toro MD, Gálvez-Acebal J, Bereciartua-Bastarrica E, Fariñas MC, Sanz-Franco M, et al.; REIPI/SAB group. Impact of an evidence-based bundle intervention in the quality-of-care management and outcome of Staphylococcus aureus bacteremia. Clin Infect Dis. 2013;57(9):1225–33. doi:. http://dx.doi.org/10.1093/cid/cit499 PubMed
3 Skogberg K, Lyytikäinen O, Ollgren J, Nuorti JP, Ruutu P. Population-based burden of bloodstream infections in Finland. Clin Microbiol Infect. 2012;18(6):E170–6. doi:. http://dx.doi.org/10.1111/j.1469-0691.2012.03845.x PubMed
4 Korvick JA, Bryan CS, Farber B, Beam TR, Schenfeld L, Muder RR, et al.Prospective observational study of Klebsiella bacteremia in 230 patients: outcome for antibiotic combinations versus monotherapy. Antimicrob Agents Chemother. 1992;36(12):2639–44. doi:. http://dx.doi.org/10.1128/AAC.36.12.2639 PubMed
5 Paterson DL, Ko WC, Von Gottberg A, Mohapatra S, Casellas JM, Goossens H, et al.International prospective study of Klebsiella pneumoniae bacteremia: implications of extended-spectrum beta-lactamase production in nosocomial Infections. Ann Intern Med. 2004;140(1):26–32. doi:. http://dx.doi.org/10.7326/0003-4819-140-1-200401060-00008 PubMed
6 van Hal SJ, Jensen SO, Vaska VL, Espedido BA, Paterson DL, Gosbell IB. Predictors of mortality in Staphylococcus aureus bacteremia. Clin Microbiol Rev. 2012;25(2):362–86. doi:. http://dx.doi.org/10.1128/CMR.05022-11 PubMed
7 Pedersen M, Benfield TL, Skinhoej P, Jensen AG. Haematogenous Staphylococcus aureus meningitis. A 10-year nationwide study of 96 consecutive cases. BMC Infect Dis. 2006;6(1):49. doi:. http://dx.doi.org/10.1186/1471-2334-6-49 PubMed
8 Hill PC, Birch M, Chambers S, Drinkovic D, Ellis-Pegler RB, Everts R, et al.Prospective study of 424 cases of Staphylococcus aureus bacteraemia: determination of factors affecting incidence and mortality. Intern Med J. 2001;31(2):97–103. doi:. http://dx.doi.org/10.1111/j.1444-0903.2001.00029.x PubMed
9 Rhee JY, Kwon KT, Ki HK, Shin SY, Jung DS, Chung DR, et al.Scoring systems for prediction of mortality in patients with intensive care unit-acquired sepsis: a comparison of the Pitt bacteremia score and the Acute Physiology and Chronic Health Evaluation II scoring systems. Shock. 2009;31(2):146–50. doi:. http://dx.doi.org/10.1097/SHK.0b013e318182f98f PubMed
10 Al-Hasan MN, Lahr BD, Eckel-Passow JE, Baddour LM. Predictive scoring model of mortality in Gram-negative bloodstream infection. Clin Microbiol Infect. 2013;19(10):948–54. doi:. http://dx.doi.org/10.1111/1469-0691.12085 PubMed
11 Al-Hasan MN, Juhn YJ, Bang DW, Yang HJ, Baddour LM. External validation of bloodstream infection mortality risk score in a population-based cohort. Clin Microbiol Infect. 2014;20(9):886–91. doi:. http://dx.doi.org/10.1111/1469-0691.12607 PubMed
12 Osthoff M, Sidler JA, Lakatos B, Frei R, Dangel M, Weisser M, et al.Low-dose acetylsalicylic acid treatment and impact on short-term mortality in Staphylococcus aureus bloodstream infection: a propensity score-matched cohort study. Crit Care Med. 2016;44(4):773–81. PubMed
13 Lemeshow S, Hosmer DW. A review of goodness of fit statistics for use in the development of logistic regression models. Am J Epidemiol. 1982;115(1):92–106. doi:. http://dx.doi.org/10.1093/oxfordjournals.aje.a113284 PubMed
14 Liao CH, Chen SY, Huang YT, Hsueh PR. Outcome of patients with meticillin-resistant Staphylococcus aureus bacteraemia at an Emergency Department of a medical centre in Taiwan. Int J Antimicrob Agents. 2008;32(4):326–32. doi:. http://dx.doi.org/10.1016/j.ijantimicag.2008.04.011 PubMed
Published under the copyright license CC BY-NC-SA: This license allows reusers to distribute, remix, adapt, and build upon the material in any medium or format for noncommercial purposes only, and only so long as attribution is given to the creator. If you remix, adapt, or build upon the material, you must license the modified material under identical terms.