To the Editor—The clinical variables that most commonly trigger blood cultures in daily practice (eg, fever, leukocytosis) correlate poorly with bacteremia.Reference Coburn, Morris, Tomlinson and Detsky ^{1 –} Reference Riedel, Bourbeau and Swartz ³ Published guidelines do not provide specific recommendations for when blood cultures should be drawn.Reference O’Grady, Barie and Bartlett ^{2 ,} Reference Miller, Binnicker and Campbell ^{4 – 6} Liberal blood culture testing may result in unnecessary and potentially harmful consequences: exposure to unnecessary antibiotics, unnecessary removal of venous catheters, overestimation of central-line–associated bloodstream infections (CLABSI) and added healthcare costs.Reference Alahmadi, Aldeyab and McElnay ^{7 ,} Reference Zwang and Albert ⁸

We surveyed prescribers’ knowledge, attitudes, and perceptions regarding blood cultures to help in developing future interventions aimed at optimizing blood culture testing of adult patients. Using the Qualtrics survey system, an 8-item electronic questionnaire was sent to 359 providers who provide clinical inpatient care at The Johns Hopkins Hospital: medical and surgical physician assistants and nurse practitioners (PAs/NPs), medicine house staff, hospitalists, intensivists and infectious disease (ID) physicians. We developed the survey and pilot-tested it among 7 prescribers for readability and relevance of specific items. Answers to questions that used a 5-point Likert scale were condensed into 2 categories: agree/strongly agree and neutral/disagree/strongly disagree. The survey was anonymous and voluntary. No incentives were offered for participation. Differences between groups were assessed with nonparametric tests (Fisher exact and Wilcoxon rank sum) using Stata version 13.0 software (StataCorp, College Station, TX). A 2-sided P value<.05 was considered statistically significant for all tests. The Johns Hopkins University Institutional Review Board acknowledged this study.

Overall, 109 providers (30%) completed the survey. The median number of years of work experience for respondents was 7.5 (interquartile range [IQR], 4–12) for PAs/NPs, 5 (IQR, 3.5–16) for medicine attendings (hospitalists and intensivists), and 13 (IQR, 4–17.5) for ID physicians. Only 50% of PAs/NPs responded that they would order blood cultures for a new fever in a non-ICU patient, but this proportion increased to 83% if the patient was in the ICU (P<.01) (Table 1). For other provider groups, >75% of respondents would obtain blood cultures for this reason regardless of patient location. House staff and PAs/NPs were more likely to obtain blood cultures for new leukocytosis in an ICU patient than for a patient not in the ICU (P<.01 for house staff and P=.02 for PAs/NPs). The proportion of respondents who would order a follow-up blood culture for patients with S. aureus or gram-negative bacteremia in the preceding 24 hours was similar by role, regardless of ICU location. The prescribers more likely to order follow-up blood cultures for gram-negative bacteremia had less years of experience compared to the prescribers unlikely to order blood cultures for this indication (5.2 vs 9.1 years, respectively; P<.01). Single blood cultures were considered appropriate to detect bacteremia in follow-up cultures by 88% of trainees, 56% of medicine attendings, 54% of ID physicians, and 45% of PA/NPs. More than 80% of all respondents agreed that clinicians order blood cultures reflexively in response to signs and symptoms such as fever.

Table 1 Prescribers’ Knowledge, Attitudes and Perceptions About Blood Culturing Practices for Adult Inpatients

Note. ID, infectious disease; PA/NP, physician assistant/nurse practitioner.

Most respondents (>85%) believed that blood cultures are expected as part of a patient’s work up, and many reported that a protocol with indications would improve blood culture ordering practices (>80% of medicine attendings and ID physicians, 72% of PAs/NPs, and 69% of house staff). Fewer respondents felt that order sets or communication among prescribers or between prescribers and nurses would improve blood-culturing practices. Concern for missing an infection was identified by >85% of respondents as a barrier to reducing the number of blood cultures in clinical practice. Respondents with fewer years of clinical experience agreed with this statement (7.3 years for those who agreed versus 10.5 for those who disagreed; P=.04). The consulting service was identified as a barrier to decreasing blood culture testing by 81% of attendings, 79% of PAs/NPs, 54% of ID physicians, and 51% of house staff.

Our study results suggest that decision making around blood cultures is multifactorial and is influenced by the provider’s role, the provider’s years of clinical experience, and patient location (ICU vs non-ICU). In this cohort, PAs/NPs had a lower tendency to order blood cultures than did providers in other roles, and ID physicians were as likely to order blood cultures than medicine attendings. Many providers thought a single set of blood cultures was adequate to detect bacteremia. This may explain trends observed in some units at our hospital, where single sets represent up to 60% of all blood cultures collected (K.C., personal communication).

In general, respondents acknowledged that febrile patients are more likely to yield positive blood cultures and the clear majority of providers indicated that they would order blood culture(s) if a patient developed a new fever. However, collecting blood during temperature spikes was not shown to increase the likelihood of documenting bacteremia.Reference Riedel, Bourbeau and Swartz ³ Clinical prediction rules to increase the positive predictive value of blood cultures have been developed; however, they have not been adopted widely in clinical practice.Reference Eliakim-Raz, Bates and Leibovici ⁹ In pediatric ICU patients, a sepsis screening checklist and a clinical decision algorithm reduced number of blood cultures without adverse events.Reference Woods-Hill, Fackler and Nelson McMillan ¹⁰ Most respondents recognized that blood cultures are ordered to help with antibiotic treatment decisions. Close monitoring of broad-spectrum antibiotic use and antibiotic de-escalation should occur if interventions to limit blood culture testing are implemented.

Our study was performed at a single center and we cannot exclude volunteer bias. However, there was balanced representation from the different groups surveyed and a wide range of years of experience. In summary, more specific guidance with indications for blood cultures may help reduce unnecessary blood cultures, and interventions should include all providers, including consulting physicians.