Identifying COVID-19-infected healthcare workers using an electronic ‘nose’

Abstract

Introduction: The coronavirus disease 2019 (COVID-19) pandemic has put tremendous pressure on the healthcare system, highlighting the need for a fast, highly sensitive and non-invasive test. The large number of infected healthcare workers results in a shortage of personnel and can become a source of infection for patients. In this study volatile molecules in exhaled air were analyzed in personnel using an electronic nose (Aeonose®).

Objective: To rapidly differentiate COVID-19 infected and non-infected healthcare workers with mild symptoms.

Methods: 724 healthcare workers from three large teaching hospitals were tested for COVID-19 according to routine hospital guidelines. Signed informed consent and data on sex, age and symptoms were collected before the COVID-19 test results were available. Follow-up ensured correct classification of infection status. Breath profiles were obtained with the Aeonose and analyzed using a neural network. Logistic regression analyses were performed in SPSS.

Results: Of 724 participants, 107 (15%) tested positive for COVID-19. In the training set, a model based on symptoms (coughing, fever, headache and loss/change of smell) yielded an AUC of the ROC (AUROC) of 0.79. Aeonose breath profile analysis yielded an AUROC of 0.77. The integration of both models resulted in an AUROC of 0.90. This combined model distinguished between COVID-19 positive and negative personnel in 65% of the cases, with a negative predictive value of 98% (431/440) and a positive predictive value of 98% (41/42). Independent validation is ongoing.

Conclusions: Combining breath analysis with symptoms is a rapid novel diagnostic tool to identify COVID-19 infection in healthcare workers with mild symptoms.