Diagnostic Accuracy of Neutrophil Side Scatter Width (NE-WY) for Bloodstream Infection in Critically Ill Patients: A Prospective Observational Study

1. Introduction

Sepsis affects an estimated 49 million people annually and accounts for 11 million deaths worldwide [1,2]. In the ICU, bloodstream infection (BSI) complicates 20–50% of sepsis episodes and independently increases mortality two- to fourfold compared with non-bacteraemic sepsis [3]. Blood culture — the diagnostic reference standard — requires 24–72 hours to yield a result; during this interval, empirical broad-spectrum antimicrobial treatment proceeds without microbiological guidance.

Conventional biomarkers — procalcitonin (PCT) and C-reactive protein (CRP) — improve early infection detection, but lose discriminative power in the ICU, where systemic inflammation from non-infectious causes generates elevations indistinguishable from bacterial BSI [4]. A point-of-care parameter derived from a standard complete blood count would reduce diagnostic delay without added cost.

Cell population data (CPD) are morphological metrics generated automatically during routine CBC analysis on modern hematology analyzers [5,6]. On the Sysmex XN series, NE-WY (neutrophil side scatter width) quantifies granular complexity heterogeneity in the WDF channel; it rises during neutrophil degranulation in response to bacterial pathogens [5,7,8]. NE-SFL (neutrophil fluorescence intensity) captures cytoplasmic RNA and nuclear complexity, reflecting neutrophil immaturity shifts during infection [5,9,10]. Related extended hematology parameters — immature granulocyte fractions, delta neutrophil index, neutrophil-to-lymphocyte ratio, and immature platelet fraction — have each demonstrated diagnostic value across adult and neonatal infection contexts [11,12,13,14,15,16].

Park et al. established NE-WY reference intervals on the Sysmex XN platform and demonstrated elevation in inflammatory states [5]. Subsequent studies confirmed higher NE-WY in bacteraemia [7,8], in liver disease [9], and in bacteraemic versus non-bacteraemic infection [10]. Machine-learning models incorporating Sysmex XN-3000 parameters achieved AUC 0.92 for bacteremia prediction [17], and an extended-parameter panel discriminated early ICU sepsis from non-infectious SIRS [18]. Prospective data specifically pairing NE-WY with confirmed BSI in mixed ICU populations remain limited.

Monocyte Distribution Width (MDW), FDA-cleared and CE-marked for sepsis screening, is measured on the Beckman Coulter DxH 900 [19,20,21,22,23,24,25,26,27,28,29]. Published AUC values in ED and ICU settings range from 0.74 to 0.88 [24], making MDW the current CPD reference benchmark for sepsis biomarker comparison. Head-to-head evaluation of NE-WY against MDW in a BSI-confirmed ICU cohort has not been reported.

This study evaluated the diagnostic accuracy of NE-WY and NE-SFL, measured at ICU admission, for confirmed BSI (positive blood culture by NHSN LCBI criteria) compared with non-sepsis states. Secondary objectives were: to compare NE-WY against MDW, PCT, and CRP; to assess NE-WY as an independent predictor of BSI; and to estimate diagnostic performance when suspected-sepsis patients with non-diagnostic cultures are included as comparators.

2. Methods

2.1. Study Design, Setting, and Reporting

This prospective observational study was conducted in the general ICU of Dr. Soetomo General Academic Hospital, Surabaya, Indonesia — a tertiary referral centre affiliated with Universitas Airlangga. The study is reported per the Standards for Reporting of Diagnostic Accuracy Studies (STARD) 2015; the completed checklist is provided as Supplementary Table S1. Ethics approval: No. 0085/LOE/301.4.2/I/2024, Faculty of Medicine, Universitas Airlangga / Dr. Soetomo General Academic Hospital. The study was conducted in accordance with the Declaration of Helsinki. All patients or their legal guardians provided written informed consent.

2.2. Participants

Adults admitted to the ICU were screened consecutively. Inclusion criteria: age ≥18 years, ICU stay ≥24 hours, and complete blood count obtained within 24 hours of admission. Exclusion criteria: prior antibiotic therapy >48 hours before ICU admission, haematological malignancy, immunosuppressive therapy equivalent to prednisolone >20 mg/day, and incomplete CPD data.

2.3. Patient Classification

Enrolled patients were allocated to three groups at ICU admission. Group 1 (G1, Sepsis+BSI) comprised patients meeting Sepsis-3 criteria (SOFA ≥2) [1,30] with a positive blood culture for a non-contaminant organism by NHSN LCBI definition. Twenty-five patients with suspected sepsis but negative or indeterminate blood cultures were enrolled and appear in the study flow diagram; they were excluded from the primary analysis because this group could not be cleanly classified as bacteraemic or non-bacteraemic, and are included in the pre-specified sensitivity analysis. Group 2 (G2, Non-Sepsis) included patients with SOFA <2 and negative or commensal blood culture. The primary analysis compared G1 versus G2 (n=47).

2.4. Laboratory Measurements and Timing

Two sets of blood cultures (aerobic and anaerobic) were obtained from separate venipunctures at ICU admission per institutional protocol. Blood cultures were incubated in the BACTEC FX400 system (Becton Dickinson, NJ, USA) and monitored for five days. Organisms meeting NHSN LCBI definitions were classified as pathogens; common skin commensals (coagulase-negative staphylococci, Bacillus spp., Corynebacterium spp., Propionibacterium spp., viridans streptococci) in a single culture set were classified as contaminants.

The complete blood count — from which NE-WY and NE-SFL were extracted — was drawn before or within one hour of blood culture collection, as part of the routine ICU admission workup. NE-WY and NE-SFL values were not reported to clinicians during the study period; they were extracted retrospectively from archived Sysmex XN-3000 analyser records, ensuring that clinical management decisions were independent of these CPD parameters.

NE-WY and NE-SFL were measured on the Sysmex XN-3000 (Sysmex Corp., Kobe, Japan) from the WDF channel. The published healthy-adult reference range for NE-WY on the Sysmex XN series is approximately 550–680 arbitrary units (a.u.) [31]. MDW was measured on the Beckman Coulter UniCel DxH 900 (Beckman Coulter Inc., Brea, CA, USA); results were unavailable for 13/72 patients (18.1%) owing to monocyte percentage <4.5%, the technical threshold for MDW reporting on this platform [19]. PCT was measured by electrochemiluminescence immunoassay (ECLIA, Roche Diagnostics); CRP by immunoturbidimetry. All measurements were from ICU admission samples.

2.5. Statistical Analysis

Data are reported as median (interquartile range, IQR) for continuous variables and as count (percentage) for categorical variables. Normality was assessed by Shapiro-Wilk test; NE-WY and NE-SFL in G1 were non-normal (both p<0.001). Mann-Whitney U test compared G1 and G2; effect size was r=Z/sqrt(N). Three-group comparisons used Kruskal-Wallis with Dunn post-hoc Bonferroni correction. ROC curves were constructed and AUC values compared by the DeLong method. The Youden index determined optimal cutoffs. Binary logistic regression assessed independent predictors of BSI; variables with p<0.05 at univariate analysis entered the multivariate model. SOFA caused complete separation in the logistic model and was excluded. Missing MDW data were not imputed; ROC analyses involving MDW used available cases (n=40). Confidence intervals for sensitivity, specificity, PPV, NPV, and accuracy were calculated by the Wilson score method. A post-hoc power calculation was performed using the Hanley-McNeil variance estimator for AUC, testing the observed AUC against a null value of 0.50 at two-sided alpha=0.05. A pre-specified sensitivity analysis included G1 versus G2 combined with the suspected-sepsis group (n=72) to estimate NE-WY performance in a clinically realistic population. All analyses were performed in R version 4.4.2 (R Foundation for Statistical Computing, Vienna, Austria) using the pROC, rstatix, and gtsummary packages.

3. Results

3.1. Patient Characteristics

Seventy-two patients were enrolled (Figure 1). After excluding the 25 patients with suspected sepsis but non-diagnostic cultures, 47 patients entered the primary analysis: G1 (Sepsis+BSI, n=26) and G2 (Non-Sepsis, n=21). Gram-positive organisms predominated in G1 (n=12, 46.2%) over Gram-negative (n=11, 42.3%), with three polymicrobial cases.

Patient characteristics are in Table 1. SOFA score differed between groups (G1: 5.92±2.98 vs G2: 0.33±0.48, p<0.001), consistent with classification criteria. ICU mortality was higher in G1 (73.1%) than G2 (33.3%, p=0.004). The 33.3% mortality in G2 reflects ICU admission for non-infectious reasons — predominantly post-operative monitoring and haemodynamic management — in which severity progressed despite low initial SOFA scores. Serum creatinine was higher in G1 (p=0.023), reflecting greater organ dysfunction.

3.2. CPD Neutrophil Parameters and Conventional Biomarkers

NE-WY was higher in G1 than G2 (median 789.5 [IQR 758.8–867.3] vs 686.0 [650.0–747.0]; Mann-Whitney U p=0.001, r=0.47). NE-SFL did not differ (54.85 [50.92–61.15] vs 52.30 [47.70–55.40]; p=0.164). MDW, PCT, and CRP showed no significant between-group difference (all p>0.05). Results are in Table 2 and Figure 2.

Three-group Kruskal-Wallis analysis (G1, suspected sepsis, G2) confirmed that NE-WY differed across groups (chi-squared=11.40, df=2, p=0.003). Post-hoc Dunn tests showed significant separation between G1 and G2 (p.adj=0.008) and between the suspected-sepsis group and G2 (p.adj=0.010), but not between G1 and the suspected-sepsis group (p.adj=1.0). This absence of discrimination between G1 and the suspected-sepsis group is the basis for the sensitivity analysis interpretation below.

3.3. Diagnostic Performance: Primary Analysis

In the primary ROC analysis (G1 vs G2, n=47), NE-WY yielded AUC=0.775 (95%CI 0.632–0.917). At the Youden-optimal cutoff of 766, NE-WY achieved sensitivity 73.1% (95%CI 53.9–86.3%), specificity 85.7% (95%CI 65.3–95.0%), PPV 86.4% (95%CI 66.7–95.3%), NPV 72.0% (95%CI 52.4–85.7%), and accuracy 78.7% (95%CI 65.1–88.0%). Post-hoc power using the Hanley-McNeil estimator confirmed 98.2% power (n1=26, n2=21, alpha=0.05). NE-WY outperformed MDW (AUC 0.533, DeLong p=0.047). Diagnostic performance for all biomarkers is in Table 3. ROC curves are in Figure 3.

3.4. Diagnostic Performance: Sensitivity Analysis

In the pre-specified sensitivity analysis (G1 vs G2 + suspected-sepsis, n=72), NE-WY yielded AUC=0.620 (95%CI 0.485–0.756) — substantially lower than the primary analysis. This attenuation is consistent with the Kruskal-Wallis finding that G1 and the suspected-sepsis group have similar NE-WY distributions (p.adj=1.0). When the comparator includes patients with suspected sepsis — who show neutrophil activation regardless of blood culture result — NE-WY no longer discriminates adequately. The primary AUC=0.775 represents performance in a population where infection status is maximally separated from non-infection.

3.5. Logistic Regression

At univariate logistic regression, NE-WY (OR 1.008 per unit, 95%CI 1.002–1.016, p=0.020) and serum creatinine (OR 1.339, 95%CI 1.075–1.992, p=0.048) were significant predictors of BSI. Expressed per 10-unit NE-WY increment, OR is 1.083 (95%CI 1.022–1.175); expressed per IQR of G1 (approximately 109 units), OR is 2.38 (95%CI 1.12–5.08), meaning a patient at the 75th percentile NE-WY of the BSI group has 2.38-fold higher odds of BSI than a patient at the 25th percentile. At multivariate analysis (NE-WY + creatinine, n=47), NE-WY showed a marginal trend (OR 1.007, 95%CI 1.001–1.014, p=0.064); creatinine was non-significant (OR 1.213, p=0.134). Hosmer-Lemeshow test: chi-squared=5.37, df=8, p=0.717.

4. Discussion

This study makes four observations. NE-WY discriminated confirmed BSI from non-sepsis in ICU patients with AUC=0.775, sensitivity 73.1%, and specificity 85.7% in the primary analysis. When the comparator included patients with suspected sepsis and non-diagnostic cultures, NE-WY AUC fell to 0.620 — an important caveat for clinical extrapolation. Conventional biomarkers — MDW, PCT, and CRP — provided no meaningful discrimination. NE-WY outperformed MDW (DeLong p=0.047).

The specificity of 85.7% at the optimal cutoff (766 a.u.) supports a 'rule-in' approach: NE-WY >766 provides meaningful probability increase for BSI and could prompt targeted antibiotic escalation while culture results are pending. The cutoff of 766 a.u. exceeds the published healthy-adult upper reference limit of approximately 680 a.u. on the Sysmex XN series [31], and falls within the range of published BSI-associated cutoffs (730–800 a.u.) reported by Miyajima et al. (2023) and Horie et al. (2025) [7,17], supporting biological plausibility. Analytical variation of NE-WY on the Sysmex XN-3000 is approximately 2–4% CV per manufacturer validation data, indicating that the difference between 766 and neighbouring values is analytically meaningful.

The biological basis of NE-WY elevation in BSI is neutrophil degranulation and cytoplasmic reorganisation in response to conserved bacterial structures — lipopolysaccharide, lipoteichoic acid, and peptidoglycan — via pattern recognition receptors [5,6]. Side scatter width reflects the distribution of these morphological changes; a broader distribution indicates greater heterogeneity, consistent with a subset of highly activated cells against a background of less-activated forms [5,6]. We did not formally test for differential NE-WY performance by Gram type. The balanced distribution of Gram-positive (46.2%) and Gram-negative (42.3%) organisms in G1 suggests that the neutrophil activation signal is not pathogen class-specific, consistent with recognition of shared bacterial structural motifs. Formal subgroup analysis by pathogen class would require a substantially larger sample.

Two limitations of the study design require explicit acknowledgement. First, the primary comparator groups differ not only in bacteraemia but in severity of organ dysfunction (G1 SOFA 5.92 vs G2 SOFA 0.33). NE-WY elevation in G1 may reflect the neutrophil response to severe sepsis — driven by cytokines, vasopressor use, and multi-organ stress — rather than bacteraemia specifically. The ideal test of BSI-specificity would compare patients who all meet Sepsis-3 criteria (SOFA ≥2), divided by blood culture result. Our cohort did not achieve this: the Kruskal-Wallis finding that NE-WY did not differ between G1 and the suspected-sepsis group (p.adj=1.0) is consistent with both groups showing similar neutrophil activation in the context of sepsis physiology, regardless of bacteraemia confirmation. Future studies should restrict enrolment to Sepsis-3 patients and compare culture-positive against culture-negative subgroups.

Second, the sensitivity analysis AUC=0.620 is the more clinically realistic performance estimate. Clinicians using NE-WY in practice will encounter patients from the full spectrum, including those with suspected sepsis and pending cultures. The primary AUC=0.775 should be interpreted as an upper-bound estimate obtained in a population where the two groups are maximally separated by disease severity. The sensitivity analysis result should inform clinical decision-making, particularly regarding the likelihood of false reassurance in suspected-sepsis patients with NE-WY below 766.

The absence of NE-SFL significance in our cohort contrasts with the landmark study by Park et al. (2014), where NE-SFL achieved AUC=0.909 in a sepsis-versus-SIRS general population (n>300) [5]. Two factors explain this discrepancy. Our comparator group was non-septic (SOFA <2), so both groups lacked the sustained cytokine milieu that drives RNA-related fluorescence changes in activated immature granulocytes. Additionally, the smaller sample (n=47 vs >300) limits power to detect a modest NE-SFL difference. Urrechaga et al. (2019) reported a similar attenuation of NE-SFL specificity in populations with elevated baseline inflammation [6].

The non-performance of MDW in our ICU cohort merits discussion. MDW is FDA-cleared based on emergency department data, achieving AUC 0.85–0.88 in meta-analysis [24]. In our cohort, both sepsis groups had elevated MDW medians (~25 fl), suggesting that the monocyte response to critical illness itself elevates MDW above the thresholds validated in emergency settings. Morales Indiano et al. (2025) made a similar observation in a broader ICU sepsis population [23]. The 18.1% missing rate for MDW (monocyte percentage below the DxH 900 technical threshold) also reduced statistical power for this comparison.

5. Limitations

Several limitations apply. The SOFA-confounded group comparison is the primary interpretive caveat, discussed above. The primary sample size of n=47 is modest, yielding wide AUC 95%CIs (0.632–0.917). The single-centre design limits generalisability to other ICU case mixes and microbial ecologies. Blood culture turnaround of 24–72 hours means that NE-WY and the reference standard were not contemporaneous for all patients. Corticosteroid use was not systematically recorded; ICU patients with septic shock frequently receive dexamethasone or hydrocortisone, which suppress neutrophil degranulation and could attenuate NE-WY elevation in G1 — biasing toward the null and potentially underestimating true discriminative value. The enrolment criterion of ICU stay ≥24 hours excluded patients who died or were transferred before blood culture, introducing survivor selection bias. NE-WY serial kinetics over the ICU course were not captured; dynamic changes may carry prognostic value independent of admission measurements.

6. Conclusions

NE-WY from routine Sysmex XN-3000 analysis identifies confirmed BSI against non-septic ICU controls with AUC=0.775 and specificity 85.7%, outperforming MDW, PCT, and CRP. Performance against suspected-sepsis comparators falls to AUC=0.620, indicating that severity confounding limits the BSI-specificity claim. Future validation studies should restrict enrolment to Sepsis-3 patients and compare culture-confirmed BSI against culture-negative sepsis to achieve a clinically definitive performance estimate.