Hostname: page-component-7b7567c8c-7pwfn Total loading time: 0 Render date: 2026-11-14T10:26:31.810Z Has data issue: false hasContentIssue false
  • English
  • Français

The SCANNED rule: derivation and validation of a two-step rule for blood culture stewardship in patients presenting with suspected infection

Published online by Cambridge University Press:  14 July 2026

Zhenghong LiuOpen the ORCID record for Zhenghong Liu[Opens in a new window],
Maehanyi Frances Rajendram,
Li Yan Julia Jaffar,
Xiang Yin Tan,
Myriam Hamtiaux,
Siew Yee Thien,
Frank Scheuermeyer and
Robert StenstromOpen the ORCID record for Robert Stenstrom[Opens in a new window]
Zhenghong Liu*
Affiliation:
Singapore General Hospital, SingaporeSt Paul’s Hospital, Canada
Maehanyi Frances Rajendram
Affiliation:
Singapore General Hospital, Singapore
Li Yan Julia Jaffar
Affiliation:
Singapore General Hospital, Singapore
Xiang Yin Tan
Affiliation:
Singapore General Hospital, Singapore
Myriam Hamtiaux
Affiliation:
University of Ottawa Faculty of Medicine, Canada,
Siew Yee Thien
Affiliation:
Singapore General Hospital, Singapore
Frank Scheuermeyer
Affiliation:
St Paul’s Hospital, Canada
Robert Stenstrom
Affiliation:
St Paul’s Hospital, Canada
*
Corresponding author: Zhenghong Liu; Email: liu.zh@duke-nus.edu.sg
Rights & Permissions [Opens in a new window]

Abstract

Background:

Blood culture over-ordering is a recognized diagnostic stewardship challenge, contributing to contamination, inappropriate antimicrobial escalation, and laboratory burden. Validated tools to guide culture ordering at initial patient assessment are lacking.

Objective:

We aim to derive and externally validate a two-step diagnostic stewardship rule for blood culture ordering in the emergency department (ED) patients with suspected infection.

Methods:

We derived the Stepwise Culture Assessment with NEWS and NLR in ED (SCANNED) rule using consecutive blood cultures from a Singapore ED (April–July 2025) and validated it in Vancouver, Canada (2016–2019). The rule uses age-stratified National Early Warning Score (NEWS) thresholds (Step 1), followed by neutrophil-lymphocyte ratio (NLR) ≥5 for remaining patients (Step 2). Thresholds were selected via exhaustive grid search, optimizing culture reduction while constraining the false-negative rate to <5%.

Results:

We included 1,224 derivation and 808 validation encounters. Blood culture positivity was 9.6% (117/1,224) and 13.8% (112/808), respectively. The optimal rule (NEWS ≥5 if age <75 or NEWS ≥3 if age ≥75; NLR ≥5 for remainder) reduced cultures by 18.8% (95% CI 16.6–20.9%) in derivation and 15.1% (95% CI 12.8–17.6%) in validation, missing <5% of bacteremia cases in both cohorts. Performance was consistent across infection sources, including unclear/occult infections (17.5% reduction, 3.7% false-negative rate).

Conclusion:

The SCANNED rule represents a practical diagnostic stewardship intervention that can reduce unnecessary blood cultures by 15–19% while maintaining >95% sensitivity for bacteremia. It requires no electronic health record integration and can be applied at triage. Prospective validation is needed before clinical implementation.

Information

Type
Original Article
Information
Infection Control & Hospital Epidemiology , Volume 47 , Issue 9 , September 2026, pp. 908 - 914
Check for updates
Copyright
© The Author(s), 2026. Published by Cambridge University Press on behalf of The Society for Healthcare Epidemiology of America

Background

Bloodstream infections are a leading cause of high-acuity presentations to emergency departments (EDs),Reference Goto and Al-Hasan1 and blood cultures are essential for pathogen-specific therapy and antimicrobial stewardship. However, the incidence of true positive blood cultures in the ED is only 1.4–12.2%, with positive results impacting treatment decisions only 0.18–2.8% of the time.Reference Schinkel, Boerman and Bennis2,Reference Mountain, Bailey, O’Brien and Jelinek3 When cultures return negative, these results also rarely alter the initial treatment course; treatment decisions using empiric antibiotic guidelines are usually made in the ED before culture results are available.Reference Girón-Ortega JA, Fernández-Guerrero and De Oca Arjona4

Timely antimicrobial administration and performing blood cultures prior to that are quality measures in many centers, potentially leading to unnecessary blood cultures. Inappropriate blood cultures done in the ED have downstream consequences; blood culture contamination rates are higher when done in the ED, and are associated with a prolonged length of stay, inappropriate antibiotic use, bacterial resistance, and premature removal of indwelling lines.Reference Fabre, Klein and Salinas5–Reference Bates, Goldman and Lee7 Beyond direct costs, over-ordering perpetuates a culture of diagnostic inertia: patients with negative blood cultures frequently continue empiric antimicrobial therapy without reassessment, with one prospective multicentre cohort reporting inappropriate antibiotic treatment in 40% of culture-negative patients at 48 hours.Reference Girón-Ortega JA, Fernández-Guerrero and De Oca Arjona4

Previous diagnostic stewardship efforts to develop blood culture decision rules have produced scores with good predictive performance, including the Shapiro Score (area under curve (AUC) 0.68),Reference Quintero Montealegre, Flórez Monroy and Garzón Herazo8,Reference Shapiro, Wolfe, Wright, Moore and Bates9 the ID-Bacter Score (AUC 0.74),Reference Quintero Montealegre, Flórez Monroy and Garzón Herazo8 and the BAROC Score (AUC 0.76).Reference Pehlivan, Douillet and Jérémie10 In theory, these scores could achieve significant reductions in blood culture use (10% for BAROC, 26.7% for Shapiro), while maintaining acceptable false negative rates of 2–5%. Reference Quintero Montealegre, Flórez Monroy and Garzón Herazo8–Reference Takeshima, Yamamoto and Noguchi11 A previous study of 149 clinicians found that most were comfortable with a 5% false negative rate,Reference Lautenbach, Localio and Nachamkin12 suggesting these performance characteristics fall within clinically acceptable ranges. However, despite their theoretical promise, these scores have not been widely implemented,Reference Eliakim-Raz, Bates and Leibovici13 possibly due to practical barriers.

Many scores require blood test results, conflicting with ED workflows where providers order all tests at initial IV cannulation, often after triage.Reference Quintero Montealegre, Flórez Monroy and Garzón Herazo8,Reference Shapiro, Wolfe, Wright, Moore and Bates9 Others require extensive data entry,Reference Pehlivan, Douillet and Jérémie10 electronic health record integration,Reference Schinkel, Boerman and Bennis2 or subjective clinical assessments.Reference Shapiro, Wolfe, Wright, Moore and Bates9,Reference Pehlivan, Douillet and Jérémie10 Furthermore, existing rules do not leverage common early warning scores such as quick sequential organ failure assessment (qSOFA), systemic inflammatory response syndrome (SIRS) and National Early Warning Score (NEWS), already widely deployed for sepsis risk stratification.Reference Usman, Usman and Ward14

To our knowledge, no existing rules combine triage-available clinical variables with a routine laboratory biomarker in a two-step sequential framework. We therefore aimed to derive and externally validate a two-step diagnostic stewardship rule integrating triage-available clinical variables with a routine laboratory biomarker to guide blood culture ordering in patients presenting with suspected infection.

Methods

Study design and setting

This was a retrospective prediction model derivation and external validation study conducted at two independent EDs.

Derivation site

Model derivation was performed at Singapore General Hospital (SGH), a 1,800 bed tertiary medical center in Singapore with an annual ED census of approximately 130,000 patients. Consecutive blood cultures drawn between 20th April 2025 and 20th July 2025 were included.

SGH ED serves a predominantly elderly population with multiple co-morbidities, including transplant and oncology patients. During the study period, blood culture ordering decisions were typically made by trainees in consultation with fully qualified Emergency Physicians (EP), without standardized department or hospital protocols. Patients undergoing blood culture collection generally had their cultures drawn by junior doctors and were admitted for follow-up of culture results. All consecutive patients above 18 years old for whom blood cultures were drawn were included, without exclusions.

Validation site

External validation was performed at St Paul’s Hospital (SPH), a 500-bed tertiary academic medical center in Vancouver, Canada (∼80,000 annual ED visits). The validation cohort comprised 808 consecutive cultures from 2016to 2019 among adults with suspected sepsis, defined as suspected infection and ≥2 SIRS criteria at triage per institutional protocol.Reference MacRedmond, Hollohan, Stenstrom, Nebre, Jaswal and Dodek15 Exclusion criteria included septic shock, injection drug use, immunocompromise (HIV not on anti-retroviral therapy; active chemotherapy, corticosteroids); indwelling venous catheters; and hospitalization or surgical procedure in the previous 2 weeks. Of 1,038 patients screened at SPH, 230 were excluded: 78 (33.9%) were excluded for septic shock, 91 (39.6%) for injection drug use, 42 (18.3%) for immunocompromised status, 8 (3.5%) for indwelling venous catheters, and 11 (4.8%) for recent hospitalization or surgery.

Reporting of this study followed the Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD) reporting guidelines.Reference Collins, Reitsma, Altman and Moons16

Outcomes

The primary outcome was blood culture positivity, defined as growth of a clinically significant pathogen. In the derivation cohort, final microbiology reports were reviewed and criteria from Bates et al.Reference Bates17 were applied to distinguish true positives from contaminants. In the validation cohort, a microbiologist assessed blood culture results to determine positivity when it was unclear if a positive culture was due to contamination.

Secondary outcomes included the proportion of blood cultures that would be obtained under the rule, the percentage reduction in cultures compared to culturing all patients, the false negative rate and the distribution of cultures between Step 1 and 2.

Predictors

The final diagnosis was also extracted and categorized broadly based on ICD code groupings into one of “respiratory”, “intra-abdominal”, “urinary”, “skin and soft tissue”, and “source unclear” to allow for preplanned subgroup analysis.

Sample size

No formal sample size calculation was performed. The derivation sample included all consecutive eligible encounters during the 3-month study period (N = 1,224), with 117 positive cultures providing approximately 23 events per candidate predictor in Step 1 (4 predictors) and 19 events per candidate predictor in Step 2 (6 predictors).

Missing data

Complete-case analysis was used for each variable. The validation cohort had complete data for all variables. Patients with missing data for a specific predictor were excluded from analyses involving that predictor but retained for other analyses.

Statistical analysis and derivation of clinical rule

To narrow down candidates for our rule, we first performed a univariate comparison of clinical scores and laboratory values between patients with positive blood cultures and those without. Continuous and ordinal variables were summarized using median and interquartile range (IQR). Group comparisons were conducted using the Wilcoxon rank-sum test (two-sided). For each comparison, we additionally reported rank-biserial correlation (r) as an effect size, where positive values indicate higher values in the blood-culture-positive group.

Based on univariate results and clinical considerations, we selected NEWS, age, and NLR as rule components. Age stratification was incorporated because older adults often exhibit blunted physiologic responses to infection- with a resulting lower NEWS score despite significant illness severity.Reference Tupper, Redfern and Harrison21,Reference Clifford, Dy-Boarman, Haase, Maxvill, Pass and Alvarez22

We defined the following cutoffs: age in 5-year increments (65–80 years), NEWS thresholds of 3–5 (aligned with existing sepsis protocolsReference Inada-Kim23), and NLR increments of 0.5 (range 5–10 based on literatureReference Jiang, Liu and Yu20,Reference Gurol, Ciftci, Terzi, Atasoy, Ozbek and Koroglu24 ). We specified candidate rules as follows: Step 1, to be used in triage, recommended blood cultures based on age-stratified NEWS thresholds. Patients not meeting Step 1 criteria proceeded to Step 2, which is assessed when the blood test results are out, which recommended cultures if NLR exceeded a prespecified threshold. We performed an exhaustive grid search across all clinically plausible threshold combinations.

For each candidate rule, we calculated: (1) proportion of encounters in which cultures would be obtained; (2) percentage reduction in cultures; (3) proportion of culture-positive cases missed (false-negative rate); and 4) proportion of cultures triggered in Step 1 versus Step 2. The final rule was selected as the candidate achieving the greatest culture reduction while maintaining false-negative rate of <5%.

After deriving the two-step SCANNED (Stepwise Culture Assessment with NEWS and NLR in ED) rule, internal validation of the SCANNED rule was performed in the derivation cohort using nonparametric bootstrapping with 2000 resamples drawn with replacement. In each resample we recalculated all performance metrics and summarized uncertainty using percentile-based 95% confidence intervals.

External validation was performed by applying the fixed, prespecified rule to the independent SPH cohort and computing the same performance metrics.

We also evaluated rule performance stratified by infection source category. To assess whether inclusion of immunocompromised patients in derivation influenced rule performance, a sensitivity analysis was performed excluding patients with neutropenia (absolute neutrophil count (ANC) <1.5 x 10 ^ 9/L). The derivation data set lacked structured fields for immunocompromised status or indwelling vascular access; neutropenia was selected as the most objective available proxy.

All statistical analysis was conducted using R v4.5.2 (R Foundation, Vienna, Austria).

Ethical approval

Ethics review was deemed exempt by the SingHealth Centralised Institutional Review Board due to the use of deidentified data. The ethics review board of Providence Health Care and the University of British Columbia approved this study for the SPH cohort.

Results

Derivation and validation cohorts

The derivation cohort comprised 1,224 encounters from SGH with a median age of 71 years (IQR 61–81), while the external validation cohort comprised 808 encounters from SPH with a median age of 48 years (IQR 31–62). Sex distribution was similar across cohorts SGH 703 (57.4%) male vs SPH 462 (57.1%). Blood culture positivity was 117 (9.6%) in SGH and 112 (13.8%) in SPH. Admission rates were 1,208 (98.7%) in SGH and 567 (70.1%) in SPH (Table 1). In the derivation cohort, CRP was missing in 27.7% (339/1,224) and white blood cell count in 6.6% (81/1,224).

Table 1.

Baseline characteristics of derivation and validation cohorts

A table comparing baseline characteristics of derivation and validation cohorts.
Table 1. Long description

The table compares baseline characteristics of derivation and validation cohorts across various categories. It has 16 rows and 4 columns. The columns are labeled as Characteristic, SGH derivation (N = 1,224), and SPH validation (N = 808). The rows are grouped into Demographics, Clinical variables, Laboratory variables, and Outcomes. Row 1: Age, years, median (IQR), SGH derivation 71 (61-81), SPH validation 48 (31-62). Row 2: Male, n (%), SGH derivation 703 (57.4%), SPH validation 462 (57.1%). Row 3: Female, n (%), SGH derivation 521 (42.6%), SPH validation 342 (42.3%). Row 4: Transgender, n (%), SGH derivation 0 (0.0%), SPH validation 5 (0.6%). Row 5: Initial temperature, °C, median (IQR), SGH derivation 37.6 (36.7-38.4), SPH validation 38.2 (37.7-38.9). Row 6: NEWS score, median (IQR), SGH derivation 3 (1-5), SPH validation 4 (3-6). Row 7: qSOFA score, median (IQR), SGH derivation 0 (0-1), SPH validation 0 (0-1). Row 8: SIRS score (excluding WBC), median (IQR), SGH derivation 1 (0-2), SPH validation 2 (2-3). Row 9: NLR, median (IQR), SGH derivation 7.4 (3.8-13.9), SPH validation 8.6 (4.5-14.6). Row 10: CRP, mg/L, median (IQR), SGH derivation 50 (13-131), SPH validation 69 (23-144). Row 11: WBC, x 10^9/L, median (IQR), SGH derivation 10.3 (7.4-14.2), SPH validation 11.0 (7.2-14.6). Row 12: Lactate, mmol/L, median (IQR), SGH derivation 2.1 (1.4-3.4), SPH validation 1.5 (1.1-2.1). Row 13: Blood culture positive, n (%), SGH derivation 117 (9.6%), SPH validation 112 (13.8%). Row 14: Admitted, n (%), SGH derivation 1,208 (98.7%), SPH validation 567 (70.1%). Row 15: Suspected infection source, n (%), Respiratory, SGH derivation 364 (29.7%), SPH validation 153 (18.9%). Row 16: Suspected infection source, n (%), Unclear/occult, SGH derivation 586 (47.9%), SPH validation 504 (62.3%). Row 17: Suspected infection source, n (%), Intra-abdominal, SGH derivation 98 (8.0%), SPH validation 52 (6.4%). Row 18: Suspected infection source, n (%), Skin/soft tissue, SGH derivation 93 (7.6%), SPH validation 51 (6.3%). Row 19: Suspected infection source, n (%), Urinary, SGH derivation 81 (6.6%), SPH validation 43 (5.3%). Row 20: Suspected infection source, n (%), Other, SGH derivation 2 (0.2%), SPH validation 5 (0.6%).

* CRP missing in 339 (27.7%) of SGH cohort.

† WBC missing in 81 (6.6%) of SGH cohort.

‡ Lactate missing in 887 (72.5%) of SGH cohort).

IQR, interquartile range; NEWS, National Early Warning Score; qSOFA, quick Sequential Organ Failure Assessment.

SIRS, Systemic Inflammatory Response Syndrome; NLR, neutrophil-lymphocyte ratio; CRP, C-reactive protein; WBC, white blood cell count.

The validation cohort had higher baseline severity scores (median NEWS 4 vs 3, SIRS 2 vs 1) and inflammatory markers (median NLR 8.6 vs 7.4, CRP 69 vs 50 mg/L) compared with the derivation cohort. Initial temperature was higher in SPH (median 38.2°C vs 37.6°C). Suspected infection sources differed between sites, with SGH having more respiratory (29.7% vs 18.9%) and fewer unclear/occult (47.9% vs 62.3%) sources.

Model derivation

Univariate analysis of the derivation cohort found that blood-culture-positive patients had higher illness severity by NEWS (median 4 vs 3; P < .001) and modestly higher SIRS (excluding WBC) and qSOFA scores (P = .005 and P = .012, respectively), while age was slightly higher in the positive group (median 75 vs 71 years; P = .038). Among the lab variables, NLR showed the strongest association with blood culture positivity (median 15.76 vs 7.02; P < .001) compared to the other tests (Table 2).

Table 2.

Univariate analysis of candidate predictors in derivation cohort (N = 1,224)

A table comparing clinical scores and laboratory variables between BC positive and BC negative groups.
Table 2. Long description

A table comparing clinical scores and laboratory variables between BC positive (n = 117) and BC negative (n = 1,107) groups. The table has 12 rows and 5 columns. Column headers are Variable, BC positive (n = 117) Median (IQR), BC negative (n = 1,107) Median (IQR), Effect size (r), and P-value. Row labels include clinical scores such as NEWS score, SIRS (excluding WBC), qSOFA score, and Age, as well as laboratory variables like NLR, Lymphocytes, Neutrophils, CRP, Lactate, and WBC. Each row provides median values with interquartile ranges for both BC positive and BC negative groups, along with effect sizes and P-values. Notable trends include higher NEWS scores, NLR, Neutrophils, CRP, and Lactate levels in the BC positive group compared to the BC negative group, with corresponding lower P-values indicating statistical significance.

* CRP available in 895 patients (82 positive, 813 negative).

† Lactate available in 340 patients (45 positive, 295 negative).

‡ WBC available in 1,151 patients (107 positive, 1,044 negative).

Effect size r = rank-biserial correlation; positive values indicate higher values in blood culture-positive group.

NEWS, National Early Warning Score; SIRS, Systemic Inflammatory Response Syndrome; qSOFA, quick Sequential Organ Failure Assessment; NLR, neutrophil-lymphocyte ratio; CRP, C-reactive protein; WBC, white blood cell count; IQR, interquartile range.

The grid search identified a total of 7 candidate rules that fulfilled the prespecified criteria (Table 3). The rule that resulted in the largest reduction in blood cultures used an age cutoff of 75 years, with Step 1 culturing patients with NEWS ≥5 if age < 75 or NEWS ≥ 3 if age ≥ 75, and Step 2 culturing remaining patients with NLR ≥ 5 (Figure 1).

Figure 1.

The SCANNED (Stepwise Culture Assessment with NEWS and NLR in ED) rule.

A textbox outlining a rule for culturing based on age and NEWS thresholds.

Table 3.

Grid search results: candidate rules meeting prespecified criteria (<5% false-negative rate)

A table comparing candidate rules for blood culture reduction based on age and health scores.
Table 3. Long description

The table presents a comparison of candidate rules for reducing blood cultures based on age cutoffs, NEWS (National Early Warning Score) cutoffs, NLR (Neutrophil to Lymphocyte Ratio) cutoffs, and the resulting reduction in cultures, missed positives, and the percentage of cultures in Step 1 and Step 2. The table has 8 rows and 10 columns. Column headers are Age cutoff (years), NEWS cutoff (Age < Cutoff), NEWS cutoff (Age ≥ Cutoff), NLR cutoff, Cultures obtained n (%), Reduction (%), Missed positives n (%), Step 1 cultures (%), and Step 2 cultures (%). Row 1: Age cutoff 75 years, NEWS cutoff 5 (Age < Cutoff), NEWS cutoff 3 (Age ≥ Cutoff), NLR cutoff 5, Cultures obtained 994 (81.2 percent), Reduction 18.8 percent, Missed positives 5 (4.3 percent), Step 1 cultures 56.1 percent, Step 2 cultures 43.9 percent. Row 2: Age cutoff 70 years, NEWS cutoff 5 (Age < Cutoff), NEWS cutoff 3 (Age ≥ Cutoff), NLR cutoff 5, Cultures obtained 1,003 (81.9 percent), Reduction 18.1 percent, Missed positives 5 (4.3 percent), Step 1 cultures 59.5 percent, Step 2 cultures 40.5 percent. Row 3: Age cutoff 80 years, NEWS cutoff 4 (Age < Cutoff), NEWS cutoff 3 (Age ≥ Cutoff), NLR cutoff 5, Cultures obtained 1,005 (82.1 percent), Reduction 17.9 percent, Missed positives 5 (4.3 percent), Step 1 cultures 61.4 percent, Step 2 cultures 38.6 percent. Row 4: Age cutoff 65 years, NEWS cutoff 5 (Age < Cutoff), NEWS cutoff 3 (Age ≥ Cutoff), NLR cutoff 5, Cultures obtained 1,013 (82.8 percent), Reduction 17.2 percent, Missed positives 5 (4.3 percent), Step 1 cultures 63.1 percent, Step 2 cultures 36.9 percent. Row 5: Age cutoff 75 years, NEWS cutoff 4 (Age < Cutoff), NEWS cutoff 3 (Age ≥ Cutoff), NLR cutoff 5, Cultures obtained 1,019 (83.3 percent), Reduction 16.7 percent, Missed positives 5 (4.3 percent), Step 1 cultures 63.6 percent, Step 2 cultures 36.4 percent. Row 6: Age cutoff 70 years, NEWS cutoff 4 (Age < Cutoff), NEWS cutoff 3 (Age ≥ Cutoff), NLR cutoff 5, Cultures obtained 1,025 (83.7 percent), Reduction 16.3 percent, Missed positives 5 (4.3 percent), Step 1 cultures 65.5 percent, Step 2 cultures 34.5 percent. Row 7: Age cutoff 65 years, NEWS cutoff 4 (Age < Cutoff), NEWS cutoff 3 (Age ≥ Cutoff), NLR cutoff 5, Cultures obtained 1,030 (84.2 percent), Reduction 15.8 percent, Missed positives 5 (4.3 percent), Step 1 cultures 67.2 percent, Step 2 cultures 32.8 percent.

Highlighted row = selected final rule (maximum reduction with <5% false-negative rate).

NEWS, National Early Warning Score; NLR, neutrophil-lymphocyte ratio.

Applied to the derivation cohort, the SCANNED rule would have reduced blood cultures by 18.8% (994/1,224 cultured, 95% CI 16.6 to 20.9) while missing 4.3% of positives (5/117 positives, 95% CI 0.9 to 8.3). Among cultures obtained under the rule, 56.1% (95% CI 53.0–59.3) of all cultures were triggered by Step 1 (Table 4, Supplement 2).

Table 4.

Performance of 2 step blood culture rule in derivation and validation cohort

A table comparing the performance of a two-step blood culture rule in derivation and validation cohorts.
Table 4. Long description

The table compares the performance of a two-step blood culture rule in derivation and validation cohorts. It has four rows and three columns. The columns are labeled Metric, SGH derivation (N = 1,224), and SPH validation (N = 808). The rows are grouped under Primary Performance, Culture workflow, and Culture yield. Row 1: Culture reduction, percent (95 percent CI), 18.79 percent (16.58, 20.92), 15.10 percent (12.75, 17.57). Row 2: Missed positives, n / total positives, 5/117, 5/112. Row 3: False negative rate, percent (95 percent CI), 4.27 percent (0.88, 8.33), 4.46 percent (0.93, 8.52). Row 4: Step 1 share of all cultures, percent (95 percent CI), 56.14 percent (53.04, 59.30), 59.04 percent (55.40, 62.70). Row 5: Step 1 cultures, n, 558, 405. Row 6: Step 2 cultures, n, 436, 281. Row 7: Step 1 positivity, percent (95 percent CI) [pos/step1], 12.90 percent (10.18, 15.90) [72/558], 15.80 percent (12.27, 19.65) [64/405]. Row 8: Overall cultured positivity, percent (95 percent CI) [pos/overall], 11.27 percent (9.47, 13.37) [112/994], 15.60 percent (12.96, 18.45) [107/686].

External validation

External validation in the independent SPH cohort (N = 808; 112 culture-positive) demonstrated similar performance. The rule would have reduced cultures by 15.1% (686/808 cultured, 95% CI 12.8–17.6) while missing 4.5% (5/112 positives, 95% CI 0.9–8.5), corresponding to 95.5% positivity capture. In SPH, 59.0% (95% CI 55.4–62.7%) of cultures were triggered by Step 1, supporting generalizability (Table 4, Supplement 2).

Subgroup analysis

Subgroup analysis across infection sources showed consistent rule performance; in patients with unclear infection source - the hardest group to stratify - culture reduction was 17.5% with a false-negative rate of 3.7% (Supplement 1).

Sensitivity analysis

Excluding 27 neutropenic patients (ANC <1.5 x 10 ^ 9/L) from the derivation cohort (remaining n = 1,197), the rule reduced cultures by 18.0% (982/1,197 cultured, 95% CI 15.9–20.2) with a false-negative rate of 4.6% (5/108 positives, 95% CI 1.9–9.9), consistent with the primary analysis.

Missed bacteremia cases

The organisms among the 10 missed bacteremia cases (5 per cohort) are presented in Supplement 3. All but 1 case from SPH were admitted. One SPH case had a neutropenia (ANC 0.7) resulting in an NLR of 0.47- a known limitation of the NLR. None of the patients died in the ED; subsequent clinical outcomes for missed cases were not available for retrospective review.

Discussion

This study derived and externally validated the SCANNED rule, a simple two-step clinical decision rule for blood culture stewardship in the ED. Applying the SCANNED rule at two independent sites across two countries, the rule reduced blood cultures by 15–19% while missing fewer than 5% of positive cultures. Notably, rule performance held up despite substantial differences in baseline severity, demographics and clinical settings between the two cohorts.

From a diagnostic stewardship perspective, the rule’s primary value lies not in merely reducing test volume, but concentrating cultures where bacteremia is most probable. Cultures done in the ED carry disproportionately high contamination rates compared to inpatient draws, contributing to antimicrobial escalation and prolonged admissions for workup of false-positive results.Reference Sacchetti, Travis, Steed and Webb25 By withholding cultures from the lowest-risk tier – where the pretest probability of bacteremia is already low - the rule reduces exposure to this contamination cascade without sacrificing diagnostic yield for true bacteremia. Furthermore, the NO-BACT cohort study demonstrated that 40% of patients with negative blood cultures received inappropriate antimicrobial treatment at 48 hours, with insufficient coverage independently predicting 30-day mortality - highlighting that negative cultures do not reliably prevent antimicrobial misuse, and that not drawing cultures from genuinely low-risk patients may avoid antibiotic inertia rather than perpetuate it.Reference Girón-Ortega JA, Fernández-Guerrero and De Oca Arjona4

The rule’s performance compares favorably with existing blood culture decision tools while offering practical advantages. In its original derivation and validation study, the Shapiro Score can potentially reduce cultures by up to 26.7%, but requires blood test results and a subjective judgement on clinical suspicion of infective endocarditis.Reference Shapiro, Wolfe, Wright, Moore and Bates9 The BAROC score does not require blood test results and can potentially reduce cultures by 10%,Reference Pehlivan, Douillet and Jérémie10 but requires 12 items, a web-calculator and clinical judgement on whether “bacterial infection is the most likely diagnosis” and “suspected endocarditis”. Comparatively, the SCANNED rule can potentially reduce blood cultures by 15% with practical advantages: it requires only three variables (age, NEWS, NLR), needs no electronic health record integration (unlike machine-learning approaches);Reference Boerman, Schinkel and Meijerink26 and leverages scores already used for sepsis screening.Reference Inada-Kim23 As a three-variable rule requiring no additional infrastructure, the SCANNED rule is readily integrated into institutional diagnostic stewardship programmes alongside existing antimicrobial stewardship committee oversight.

The two-step design concentrates cultures where bacteremia risk is highest. Step 1 uses NEWS thresholds already integrated into sepsis screening pathways; Step 2 applies NLR – readily calculated from routine complete blood counts – as a safety net after initial assessment. 56 to 59% of cultures are triggered at Step 1 alone, capturing most bacteremic patients before laboratory results are required.

Prior blood culture decision-making frameworks have shown variable real-world impact,Reference Fabre, Sharara, Salinas, Carroll, Desai and Cosgrove27,Reference Theophanous, Ramos and Calland28 with Fabre et al.’s guidance- based on suspected types of infection- yielding culture reductions ranging from 0.8–25% across ED implementation.Reference Theophanous, Ramos and Calland28,Reference Ramos, Theophanous and Gettler29 The SCANNED rule complements this and operates upstream, providing a validated risk-stratification tool prior to provisional diagnosis- a distinction that might be valuable where infection source is initially unclear. Subgroup analysis in patients with unclear infection source - arguably the hardest group to stratify - demonstrated a 17.5% culture reduction with a false-negative rate below 3.7%, supporting the rule’s utility precisely in this setting.

The use of NLR as a step two criteria, while supported by our univariate analysis, is also a practical choice- this value is likely to be widely available regardless of resource setting. At the same time, NLR has its drawbacks in neutropenic patients (falsely low), or malignancy/ chronic inflammatory diseases/ steroid use (falsely high).Reference Russell, Parajuli and Gale30,Reference Templeton and McNamara31 Institutions with varying populations may wish to evaluate differing NLR thresholds or other step 2 biomarkers.

Limitations

Several limitations warrant consideration. First, derivation and validation cohorts were substantially different- The SGH cohort took all consecutive blood cultures, and included immunocompromised patients, while the SPH cohort took a sicker cohort and excluded certain populations at higher risk of bacteremia. In centers with large oncology populations or high prevalence of injection drug use, the applicable proportion of patients might be lower as such information might not be available at triage.Reference Choi, Chia and Koh32 Second, our study is a retrospective study on patients for whom blood cultures were already ordered. Real-world implementation might paradoxically increase culture rates if applied to all patients with suspected infection, rather than as a rule-out tool for patients being considered for cultures (the design intent). Third, antibiotic administration timing relative to culture draw was not captured and may affect culture positivity rates. Fourth, CRP, lactate and/or procalcitonin was not available for a large proportion of the derivation cohort, limiting biomarker comparisons. Procalcitonin may achieve superior performance to NLRReference Kaal, Meziyerh and Van Burgel33 and should be evaluated as a step 2 marker in further studies. Fifth, with only 5 missed positive cultures in each cohort, the false negative rate estimates have wide confidence intervals (0.9–8.5%). This imprecision is inherent to studies with low event rates, and further prospective studies are required. Sixth, different approaches were used for adjudicating contaminants- while both are accepted approaches, borderline cases might be classified differently, potentially affecting false-negative estimates in either direction. The consistent performance across both definitions is reassuring but does not eliminate this concern. Seventh, the derivation cohort only spanned 3 months and might miss seasonal variation in infection epidemiology. However, the validation cohort spanned 3 years – consistent rule performance is partially reassuring. Eighth, downstream clinical outcomes were not assessed, and the rule’s impact on mortality, ICU admission or antibiotic decisions remain unknown and should be evaluated prospectively.

Conclusion

The SCANNED rule offers a pragmatic approach to blood culture stewardship, leveraging clinical variables already embedded in acute care sepsis pathways. Its consistent performance across two geographically and clinically distinct cohorts is encouraging. As a three-variable tool deployable without electronic health record integration, it is readily incorporated into institutional diagnostic stewardship programmes alongside existing antimicrobial stewardship committee oversight. Prospective evaluation is the necessary next step - to determine not only whether culture reduction is achieved in real-world implementation, but whether this translates to measurable reductions in contamination events, inappropriate antimicrobial use, and associated patient outcomes.

Supplementary material

The supplementary material for this article can be found at https://doi.org/10.1017/ice.2026.10507.

Data availability statement

Deidentified data and statistical code will be made available upon reasonable request to the corresponding author following publication.

Acknowledgments

The authors thank the Office of Value Based Care, Singapore General Hospital for extracting deidentified data from SGH.

Author contribution

Study concept and design: ZL, RS. Acquisition of data: ZL (SGH site supervision), RS (SPH site supervision), MH. Analysis and interpretation of data: ZL, MH. Drafting of the manuscript: ZL. Critical revision of the manuscript for important intellectual content: RS, MFR, JJLY, XYT, SYT, FS. Statistical expertise: ZL. Acquisition of funding: ZL, RS, FS. All authors attest to meeting the four ICMJE.org authorship criteria: (1) substantial contributions to conception and design, acquisition of data, or analysis and interpretation of data; (2) drafting the article or revising it critically for important intellectual content; (3) final approval of the version to be published; and (4) agreement to be accountable for all aspects of the work.

Financial support

ZL is supported under the Nurturing Clinician Researcher Scheme (2025/PG/000020/C2) from the SingHealth Duke-NUS Academic Medical Centre. RS is supported by St. Paul’s Hospital Foundation Emergency Department Support Fund. RS is the Innovation Program Lead for Sepsis and Infections, Emergency Care BC.FS is supported by the St. Paul’s Hospital Foundation Emergency Department Support Fund, and University of British Columbia Department of Emergency Medicine.The funders had no role in study design, data collection, analysis, interpretation of data, writing of the manuscript, or decision to submit for publication.

Competing interests

ZL reports no conflict of interest. MFR reports no conflict of interest. JJLY reports no conflict of interest. XYT reports no conflict of interest. MH reports no conflict of interest. SYT reports no conflict of interest. FS reports no conflict of interest. RS serves as an unpaid Medical Director for ASEP Medical Holdings. The company had no role in the design, conduct, analysis, or reporting of this study.

Footnotes

Prior Presentations: This work has not been previously presented.

References

Goto, M, Al-Hasan, MN. Overall burden of bloodstream infection and nosocomial bloodstream infection in North America and Europe. Clin Microbiol Infect 2013;19:501–509. https://linkinghub.elsevier.com/retrieve/pii/S1198743X1461507X.10.1111/1469-0691.12195CrossRefGoogle Scholar
Schinkel, M, Boerman, AW, Bennis, FC, et al. Diagnostic stewardship for blood cultures in the emergency department: a multicenter validation and prospective evaluation of a machine learning prediction tool. eBioMedicine 2022;8:104176. https://linkinghub.elsevier.com/retrieve/pii/S2352396422003577.10.1016/j.ebiom.2022.104176CrossRefGoogle Scholar
Mountain, D, Bailey, PM, O’Brien, D, Jelinek, GA. Blood cultures ordered in the adult emergency department are rarely useful: Eur J Emerg Med 2006;13:76–79. https://journals.lww.com/00063110-200604000-00004.10.1097/01.mej.0000188231.45109.ecCrossRefGoogle ScholarPubMed
Girón-Ortega JA, JA, Fernández-Guerrero, R, De Oca Arjona, MM, et al. Antibiotic use and outcome in patients with negative blood cultures, a new target population for antimicrobial stewardship interventions: a prospective multicentre cohort (NO-BACT). J Infect 2024; 88:95–102. https://linkinghub.elsevier.com/retrieve/pii/S0163445323005856.10.1016/j.jinf.2023.11.013CrossRefGoogle Scholar
Fabre, V, Klein, E, Salinas, AB, et al. A diagnostic stewardship intervention to improve blood culture use among adult nonneutropenic inpatients: the DISTRIBUTE study. J Clin Microbiol 2020; 58:e01053-20. https://journals.asm.org/doi/10.1128/JCM.01053-20.CrossRefGoogle ScholarPubMed
Dempsey, C, Skoglund, E, Muldrew, KL, Garey, KW. Economic health care costs of blood culture contamination: a systematic review. Am J Infect Control 2019; 47:963–967. https://linkinghub.elsevier.com/retrieve/pii/S0196655318311830.10.1016/j.ajic.2018.12.020CrossRefGoogle ScholarPubMed
Bates, DW, Goldman, L, Lee, TH. Contaminant blood cultures and resource utilization. The true consequences of false-positive results. JAMA 1991;265:365–369.10.1001/jama.1991.03460030071031CrossRefGoogle ScholarPubMed
Quintero Montealegre, S, Flórez Monroy, AF, Garzón Herazo, JR, et al. External validation of ID-BactER and Shapiro scores for predicting bacteraemia in the emergency department. Ther Adv Infect Dis 2024;11:20499361241304508. https://journals.sagepub.com/doi/10.1177/20499361241304508.Google ScholarPubMed
Shapiro, NI, Wolfe, RE, Wright, SB, Moore, R, Bates, DW. Who needs a blood culture? A prospectively derived and validated prediction rule. J Emerg Med 2008; 35:255–264. https://linkinghub.elsevier.com/retrieve/pii/S0736467908004447.10.1016/j.jemermed.2008.04.001CrossRefGoogle Scholar
Pehlivan, J, Douillet, D, Jérémie, R et al. A clinical decision rule to rule out bloodstream infection in the emergency department: retrospective multicentric observational cohort study. Emerg Med J 2024; 41:20–26. https://emj.bmj.com/lookup/doi/10.1136/emermed-2022-212987.CrossRefGoogle Scholar
Takeshima, T, Yamamoto, Y, Noguchi, Y, et al. Identifying patients with bacteremia in community-hospital emergency rooms: a retrospective cohort study. PLOS ONE 2016;11:e0148078. https://dx.plos.org/10.1371/journal.pone.0148078.10.1371/journal.pone.0148078CrossRefGoogle ScholarPubMed
Lautenbach, E, Localio, R, Nachamkin, I. Clinicians required very high sensitivity of a bacteremia prediction rule. J Clin Epidemiol 2004;57:1104–1106.10.1016/j.jclinepi.2004.03.003CrossRefGoogle ScholarPubMed
Eliakim-Raz, N, Bates, DW, Leibovici, L. Predicting bacteraemia in validated models—a systematic review. Clin Microbiol Infect 2015;21:295–301. https://linkinghub.elsevier.com/retrieve/pii/S1198743X15002499.10.1016/j.cmi.2015.01.023CrossRefGoogle Scholar
Usman, OA, Usman, AA, Ward, MA. Comparison of SIRS, qSOFA, and NEWS for the early identification of sepsis in the emergency department. Am J Emerg Med 2019;37:1490–1497. https://linkinghub.elsevier.com/retrieve/pii/S0735675718308891.10.1016/j.ajem.2018.10.058CrossRefGoogle ScholarPubMed
MacRedmond, R, Hollohan, K, Stenstrom, R, Nebre, R, Jaswal, D, Dodek, P. Introduction of a comprehensive management protocol for severe sepsis is associated with sustained improvements in timeliness of care and survival. BMJ Qual Saf 2010; 19:e46. https://qualitysafety.bmj.com/lookup/doi/10.1136/qshc.2009.033407.CrossRefGoogle ScholarPubMed
Collins, GS, Reitsma, JB, Altman, DG, Moons, KGM. Transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD): the TRIPOD statement. Circulation 2015; 131:211–219. https://www.ahajournals.org/doi/10.1161/CIRCULATIONAHA.114.014508.CrossRefGoogle ScholarPubMed
Bates, DW. Rapid classification of positive blood cultures: prospective validation of a multivariate algorithm. JAMA 1992;267:1962. https://jama.jamanetwork.com/article.aspx?doi=10.1001/jama.1992.03480140088039.CrossRefGoogle ScholarPubMed
Sharma, D, Chouhan, Y, Yadav, S, Shrimal, P. To compare blood culture positivity with various sepsis scores in sepsis patients presenting to emergency department: an annotated bibliographic review. Santosh Univ J Health Sci 2024; 10:308–311. https://journals.lww.com/10.4103/sujhs.sujhs_81_24.10.4103/sujhs.sujhs_81_24CrossRefGoogle Scholar
Jeppesen, KN, Dalsgaard, ML, Ovesen, SH, et al. Bacteremia prediction with prognostic scores and a Causal probabilistic network - a cohort study of emergency department patients. J Emerg Med 2022; 63:738–746. https://linkinghub.elsevier.com/retrieve/pii/S0736467922005315.10.1016/j.jemermed.2022.09.009CrossRefGoogle Scholar
Jiang, J, Liu, R, Yu, X, et al. The neutrophil-lymphocyte count ratio as a diagnostic marker for bacteraemia: a systematic review and meta-analysis. Am J Emerg Med 2019; 37:1482–1499. https://linkinghub.elsevier.com/retrieve/pii/S073567571830888X.10.1016/j.ajem.2018.10.057CrossRefGoogle ScholarPubMed
Tupper, P, Redfern, O, Harrison, CH, et al. Is age associated with different vital signs in adults presenting to hospital with bacterial infection? A systematic review and meta-analysis. Age Ageing 2025;54:afaf194. https://academic.oup.com/ageing/article/doi/10.1093/ageing/afaf194/8210816.CrossRefGoogle ScholarPubMed
Clifford, KM, Dy-Boarman, EA, Haase, KK, Maxvill, K, Pass, SE, Alvarez, CA. Challenges with diagnosing and managing sepsis in older adults. Expert Rev Anti Infect Ther 2016; 14:231–241. https://www.tandfonline.com/doi/full/10.1586/14787210.2016.1135052.CrossRefGoogle ScholarPubMed
Inada-Kim, M. NEWS2 and improving outcomes from sepsis. Clin Med 2022; 22:514–517. https://linkinghub.elsevier.com/retrieve/pii/S1470211824026265.10.7861/clinmed.2022-0450CrossRefGoogle ScholarPubMed
Gurol, G, Ciftci, IH, Terzi, HA, Atasoy, AR, Ozbek, A, Koroglu, M. Are there standardized cutoff values for neutrophil-lymphocyte ratios in bacteremia or sepsis? J Microbiol Biotechnol 2015; 25:521–525. https://www.jmb.or.kr/journal/view.html?doi=10.4014/jmb.1408.08060.CrossRefGoogle ScholarPubMed
Sacchetti, B, Travis, J, Steed, LL, Webb, G. Identification of the main contributors to blood culture contamination at a tertiary care academic medical center. Infect Prev Pract 2022; 4:100219. https://linkinghub.elsevier.com/retrieve/pii/S2590088922000208.10.1016/j.infpip.2022.100219CrossRefGoogle Scholar
Boerman, AW, Schinkel, M, Meijerink, L, et al. Using machine learning to predict blood culture outcomes in the emergency department: a single-centre, retrospective, observational study. BMJ Open 2022;12:e053332. https://bmjopen.bmj.com/lookup/doi/10.1136/bmjopen-2021-053332.CrossRefGoogle ScholarPubMed
Fabre, V, Sharara, SL, Salinas, AB, Carroll, KC, Desai, S, Cosgrove, SE. Does this patient need blood cultures? A scoping review of indications for blood cultures in adult nonneutropenic inpatients. Clin Infect Dis 2020; 71:1339–1347. https://academic.oup.com/cid/article/71/5/1339/5703622.10.1093/cid/ciaa039CrossRefGoogle ScholarPubMed
Theophanous, R, Ramos, J, Calland, AR, et al. Blood culture algorithm implementation in emergency department patients as a diagnostic stewardship intervention. Am J Infect Control 2024;52:985–991. https://linkinghub.elsevier.com/retrieve/pii/S0196655324004723.10.1016/j.ajic.2024.04.198CrossRefGoogle ScholarPubMed
Ramos, J, Theophanous, R, Gettler, E, et al. A comparison of blood culture diagnostic stewardship across three emergency departments in a healthcare network. Am J Emerg Med 2025;93:135–139. https://linkinghub.elsevier.com/retrieve/pii/S0735675725002396.10.1016/j.ajem.2025.04.006CrossRefGoogle Scholar
Russell, CD, Parajuli, A, Gale, HJ, et al. The utility of peripheral blood leucocyte ratios as biomarkers in infectious diseases: a systematic review and meta-analysis. J Infect 2019;78:339–348. https://linkinghub.elsevier.com/retrieve/pii/S0163445319300556.10.1016/j.jinf.2019.02.006CrossRefGoogle ScholarPubMed
Templeton, AJ, McNamara, MG, Šeruga B, et al. Prognostic role of neutrophil-to-lymphocyte ratio in solid tumors: a systematic review and meta-analysis. JNCI J Natl Cancer Inst 2014;10:dju124. https://academic.oup.com/jnci/article-lookup/doi/10.1093/jnci/dju124.Google Scholar
Choi, EC-E, Chia, YH, Koh, YQ, et al. Appropriateness of blood culture: a comparison of practices between the emergency department and general wards. Infect Dis Health 2019; 24;49–55. https://linkinghub.elsevier.com/retrieve/pii/S2468045118301457.10.1016/j.idh.2018.10.003CrossRefGoogle ScholarPubMed
Kaal, AG, Meziyerh, S, Van Burgel, N, et al. Procalcitonin for safe reduction of unnecessary blood cultures in the emergency department: development and validation of a prediction model. J Infect 2024; 89:106251. https://linkinghub.elsevier.com/retrieve/pii/S0163445324001853.10.1016/j.jinf.2024.106251CrossRefGoogle ScholarPubMed
Figure 0

Table 1.Baseline characteristics of derivation and validation cohortsTable 1 long description.

Figure 1

Table 2.Univariate analysis of candidate predictors in derivation cohort (N = 1,224)Table 2 long description.

Figure 2

Figure 1.The SCANNED (Stepwise Culture Assessment with NEWS and NLR in ED) rule.

Figure 3

Table 3.Grid search results: candidate rules meeting prespecified criteria (<5% false-negative rate)Table 3 long description.

Figure 4

Table 4.Performance of 2 step blood culture rule in derivation and validation cohortTable 4 long description.

Supplementary material: File

Liu et al. supplementary material

Liu et al. supplementary material
Download Liu et al. supplementary material(File)
File 176.9 KB