During the early, pre-vaccination phase of the COVID-19 pandemic, rapid, scalable triage tools were essential in emergency departments (EDs). Complex clinical risk scores can be difficult to apply during high patient volumes. Simple indices derived from the complete blood count (CBC) — notably the systemic immune-inflammation index (SII), systemic inflammatory response index (SIRI), and pan-immune-inflammation value (PIV) — have been proposed as inexpensive, universally available biomarkers to aid risk stratification. This study evaluated whether these CBC-derived inflammatory indices retained prognostic accuracy for 30-day all-cause mortality and ED-to-ICU admission across age groups in ED patients with RT‑PCR–confirmed SARS‑CoV‑2 infection.
This retrospective observational cohort included adults (≥18 years) presenting to a tertiary academic ED in Istanbul, Turkey, between March 1 and May 31, 2020. The hospital functioned as a designated pandemic center and operated with centralized COVID-19 triage pathways. Inclusion required retrospective confirmation of SARS‑CoV‑2 infection by RT‑PCR. Patients with hematologic malignancy, active chemotherapy, or missing CBC data were excluded. After exclusions, 2,778 patients were analyzed and stratified into younger (<65 years, n = 2,269) and elderly (≥65 years, n = 509) groups.
Demographic information, comorbidities, vital signs, and laboratory parameters recorded at ED admission were extracted from electronic health records. CBC-derived indices were calculated from admission values as defined in the source: SII = platelets × neutrophils / lymphocytes; SIRI = neutrophils × monocytes / lymphocytes; PIV = platelets × neutrophils × monocytes / lymphocytes. Primary outcome was 30-day all-cause mortality (ascertained from hospital records and the national death registry). Secondary (exploratory) outcome was direct ED-to-ICU admission at initial presentation.
Continuous variables were tested for normality and compared using t-tests or Mann–Whitney U tests as appropriate; categorical variables used χ² tests. Receiver operating characteristic (ROC) analyses evaluated discrimination with area under the curve (AUC) reported. Optimal cut-offs were identified using the Youden index; sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were calculated. DeLong’s test compared AUCs between age groups and between indices. Univariate logistic regression screened predictors (p < 0.10 threshold), followed by sequential multivariate modelling: Model 1 (demographics), Model 2 (+ comorbidities), Model 3 (+ inflammatory indices categorized into quartiles). Age‑stratified multivariate analyses were performed separately for the two age bands.
The cohort (n = 2,778) had a mean age of 47.8 ± 16.2 years and 58.7% male. Overall 30-day mortality was 6.2% (171 deaths). Mortality differed sharply by age: 2.8% among patients <65 years versus 21.0% among those ≥65 years (p < 0.001). Direct ED-to-ICU admission occurred in 2.9% of patients (80/2,778). Comorbidity prevalence and burden were substantially higher in the elderly: 80.2% of elderly patients had at least one comorbidity versus 39.6% in younger patients. Laboratory differences by age included higher neutrophil counts and lower lymphocyte and platelet counts in elderly patients; median index values were likewise higher in the elderly (SII, SIRI, PIV all elevated, p < 0.001).
In the full cohort all three indices provided modest discrimination for 30-day mortality: SII AUC 0.663 (95% CI: 0.610–0.717), SIRI AUC 0.659 (95% CI: 0.604–0.715), and PIV AUC 0.649 (95% CI: 0.594–0.704), all statistically significant. For ED-to-ICU admission the indices also showed modest performance with AUCs around 0.63.
Age-stratified ROC analyses revealed marked differences. Among younger patients (<65 years), discrimination improved: SII AUC 0.727 (95% CI: 0.639–0.815), PIV AUC 0.677, and SIRI AUC 0.676 (all p < 0.001). In contrast, among elderly patients (≥65 years) discrimination was reduced: SII AUC 0.570 (95% CI: 0.501–0.639, p = 0.032), SIRI AUC 0.604, and PIV AUC 0.588. Direct comparisons using DeLong’s test showed a statistically significant age-related attenuation for SII (ΔAUC = 0.159; P = 0.0055). Differences for SIRI and PIV trended in the same direction but were not reported as statistically significant in the source.
Direct ED-to-ICU admission was uncommon (2.9%) and represented an exploratory secondary endpoint reflecting a pandemic-era admission pathway in which most PCR-positive patients were admitted to wards with subsequent ICU escalation if they deteriorated. Index performance for predicting direct ICU admission mirrored mortality findings: modest discrimination overall and variable results across age strata, with some indices showing moderate AUCs in younger patients.
Sequential multivariate logistic regression models adjusted for demographics and comorbidities and then added inflammatory indices (by quartile). Age‑stratified multivariate analyses were performed to determine independent associations with 30-day mortality. The source reports that the indices had better prognostic utility in younger adults, including high negative predictive values, whereas predictive power was attenuated in older adults, where comorbidity burden and age-related immunologic changes likely reduced biomarker signal.
CBC-derived inflammatory indices — SII, SIRI, and PIV — showed modest overall prognostic discrimination for 30-day mortality in ED patients with PCR-confirmed COVID-19 during the pre-vaccination era. Their prognostic performance was age-dependent: discrimination and negative predictive value were substantially better in patients younger than 65 years (SII showing the strongest performance), but markedly reduced in patients ≥65 years. The authors suggest these readily available, low-cost indices could help rule out higher short-term mortality risk among younger ED patients during surges, while in older adults clinicians should prioritize comprehensive clinical assessment and comorbidity profiling over sole reliance on these inflammatory markers.
Note: The source article contains detailed tables, ROC curves, and specific multivariate adjusted odds ratios not reproduced here; any precise numeric estimates beyond those reported above should be referenced directly from the full publication.