Scent dog identification of samples from COVID-19 patients - a pilot study

Abstract

Background: As the COVID-19 pandemic continues to spread, early, ideally real-time, identification of SARS-CoV-2 infected individuals is pivotal in interrupting infection chains. Volatile organic compounds produced during respiratory infections can cause specific scent imprints, which can be detected by trained dogs with a high rate of precision.

Methods: Eight detection dogs were trained for 1 week to detect saliva or tracheobronchial secretions of SARS-CoV-2 infected patients in a randomised, double-blinded and controlled study.

Results: The dogs were able to discriminate between samples of infected (positive) and non-infected (negative) individuals with average diagnostic sensitivity of 82.63% (95% confidence interval [CI]: 82.02-83.24%) and specificity of 96.35% (95% CI: 96.31-96.39%). During the presentation of 1012 randomised samples, the dogs achieved an overall average detection rate of 94% (±3.4%) with 157 correct indications of positive, 792 correct rejections of negative, 33 incorrect indications of negative or incorrect rejections of 30 positive sample presentations.

Conclusions: These preliminary findings indicate that trained detection dogs can identify respiratory secretion samples from hospitalised and clinically diseased SARS-CoV-2 infected individuals by discriminating between samples from SARS-CoV-2 infected patients and negative controls. This data may form the basis for the reliable screening method of SARS-CoV-2 infected people.

Keywords: COVID-19; Olfactory detection; SARS-CoV-2; Saliva; Scent detection dogs; Volatile organic compounds.

Fig. 1

Diagnostic specificity and sensitivity by dog and for all dogs together. Whiskers show 95% confidence intervals