The smell of sepsis: Electronic nose measurements improve early recognition of sepsis in the ED

Objective: Early recognition of sepsis is essential for timely initiation of adequate care. However, this is challenging as signs and symptoms may be absent or nonspecific. The cascade of events leading to organ failure in sepsis is characterized by immune-metabolic alterations. Volatile organic compounds (VOCs) are metabolic byproducts released in expired air. We hypothesize that measuring the VOC profile using electronic nose technology (eNose) could improve early recognition of sepsis. Material and methods: In this cohort study, bedside eNose measurements were collected prospectively from ED patients with suspected infections. Sepsis diagnosis was retrospectively defined based on Sepsis-3 criteria. eNose sensor data were used in a discriminant analysis to evaluate the predictive performance for early sepsis recognition. The dataset was randomly split into training (67 %) and validation (33 %) subsets. The derived discriminant function from the training subset was then applied to classify new observations in the validation subset. Model performance was evaluated using receiver operating characteristic (ROC) curves and predictive values. Results: We analyzed a total of 160 eNose measurements. The eNose measurements had an area under the ROC (AUROC) of 0.78 (95 % CI: 0.69–0.87) for diagnosing sepsis, with a sensitivity of 72 %, specificity of 73 %, and an overall accuracy of 73 %. The validation model showed an AUC of 0.83 (95 % CI: 0.71–0.94), sensitivity of 71 %, specificity of 83 %, and an accuracy of 80 %. Conclusion: eNose measurements can identify sepsis among patients with a suspected infection at the ED. Clinical trial registration: The study is embedded in the Acutelines data-biobank (www.acutelines.nl), registered in Clinicaltrials.gov (NCT04615065).