An Algorithm for Systemic Inflammatory Response Syndrome Criteria-Based Prediction of Sepsis in a Polytrauma Cohort.

Objectives: Lifesaving early distinction of infectious systemic inflammatory response syndrome, known as "sepsis," from noninfectious systemic inflammatory response syndrome is challenging in the ICU because of high systemic inflammatory response syndrome prevalence and lack of specific biomarkers. The purpose of this study was to use an automatic algorithm to detect systemic inflammatory response syndrome criteria (tachycardia, tachypnea, leukocytosis, and fever) in surgical ICU patients for ICU-wide systemic inflammatory response syndrome prevalence determination and evaluation of algorithm-derived systemic inflammatory response syndrome descriptors for sepsis prediction and diagnosis in a polytrauma cohort.

Design: Cross-sectional descriptive study and retrospective cohort study.

Setting: Electronic medical records of a tertiary care center's surgical ICU, 2006-2011.

Patients: All ICU admissions and consecutive polytrauma admissions.

Interventions: None.

Measurements And Main Results: Average prevalence of conventional systemic inflammatory response syndrome (≥ 2 criteria met concomitantly) from cross-sectional application of the algorithm to all ICU patients and each minute of the study period was 43.3%. Of 256 validated polytrauma patients, 85 developed sepsis (33.2%). Three systemic inflammatory response syndrome descriptors summarized the 24 hours after admission and before therapy initiation: 1) systemic inflammatory response syndrome criteria average for systemic inflammatory response syndrome quantification over time, 2) first-to-last minute difference for trend detection, and 3) change count reflecting systemic inflammatory response syndrome criteria fluctuation. Conventional systemic inflammatory response syndrome for greater than or equal to 1 minute had 91% sensitivity and 19% specificity, whereas a systemic inflammatory response syndrome criteria average cutoff value of 1.72 had 51% sensitivity and 77% specificity for sepsis prediction. For sepsis diagnosis, systemic inflammatory response syndrome criteria average and first-to-last minute difference combined yielded 82% sensitivity and 71% specificity compared with 99% sensitivity and only 31% specificity of conventional systemic inflammatory response syndrome from a nested case-control analysis.

Conclusions: Dynamic systemic inflammatory response syndrome descriptors improved specificity of sepsis prediction and particularly diagnosis, rivaling established biomarkers, in a polytrauma cohort. They may enhance electronic sepsis surveillance once evaluated in other patient populations.