Diagnostic potential of extended inflammation parameters for sepsis identification: a retrospective case-control study

Background: An early and accurate diagnosis of sepsis is critical for improving patient outcomes. Extended inflammation parameters (EIPs), derived from routine complete blood count (CBC) analysis, have emerged as promising biomarkers for inflammatory response. This study aimed to explore the diagnostic potential of a model combining several EIPs for identifying sepsis in a case-control setting. Participants and methods: A retrospective, single-center, case-control study was conducted at Tawam Hospital, AlAin, United Arab Emirates involving 157 participants; 53 patients with confirmed sepsis per Sepsis-3 criteria admitted to the Intensive Care Unit (ICU) and 104 control participants from outpatient clinics with no obvious evidence of infection. EIPs, including immature granulocyte count (IG#), neutrophil reactivity intensity (NEUT-RI), and reactive lymphocyte percentage per lymphocyte (RE-LYMP%/L), were retrieved from initial CBCs performed on a Sysmex XN-1000 analyzer. A three-parameter logarithmic model was developed, and its performance was assessed using receiver operating characteristic (ROC) curve analysis. Internal validation was performed using 1,000 bootstrap iterations to estimate bias-corrected performance. Results: The logarithmic model, i.e., log(IG# + 1) + log(NEUT-RI/100 + 1) + log(RE-LYMP%/L/50 + 1), combining IG#, NEUT-RI, and RE-LYMP%/L demonstrated high apparent discrimination for identifying sepsis, with an Area Under the Curve (AUC) of 0.941 (95% CI: 0.902–0.980), a sensitivity of 88.5% (95% CI: 77.0–95.8%), and a specificity of 91.3% (95% CI: 84.2–96.0%). Bootstrap internal validation yielded an optimism-corrected AUC of 0.923 (95% CI: 0.874–0.966), with minimal optimism (0.018), suggesting model stability within this dataset. Conclusion: A prediction model combining three different EIPs demonstrated high discrimination in a case-control setting, however this design of comparing ICU sepsis patients to healthy outpatient controls introduces severe spectrum bias characteristic of two-gate studies, which can inflate discrimination metrics significantly when compared with single-gate Emergency Department populations where diagnostic uncertainty is genuine. These results should be considered preliminary exploratory findings only. The extreme spectrum bias inherent to our case-control design means reported performance reflects statistical discrimination in an artificial scenario rather than real-world diagnostic accuracy, with expected ED performance substantially lower (estimated AUC 0.70–0.79). Rigorous prospective validation in consecutive ED patients with suspected infection, including head-to-head comparison with established biomarkers procalcitonin and C-reactive protein, is essential before any clinical consideration.