What strategies to detect contagious people at the entrance of a bar or an airplane?
Text updated on 2020-12-07
Aside from bioassays and temperature measurements to detect people potentially infected with COVID-19, new approaches are being developed: detection by dogs, computer voice analysis, and analysis of vital parameters via connected watches.
Imagine a restaurant, nightclub, or bar where you are sure not to come across a contagious person. Everyone could then, once inside, relax and live without a mask. But is this imaginable in an area where the virus is actively circulating? This would require rapid and highly sensitive tests, carried out at the entrance to easily detect infected people and only let in those who are not infected. We're not there yet, but researchers are looking at several possibilities.
The first lead concerns biological tests. These tests are carried out using nasopharyngeal swabs, saliva, gargle, and mucus present at the entrance of the nose. The tests are then based on the detection of RNA or certain proteins of the virus. There are more than 2,000 COVID-19 tests available around the world, in use or under development, which are listed on the COVID-19 test database created by Arizona State University.. The quickest tests take a few minutes to give a result but, at the moment, are not sensitive and specific enough for use at the entrance of a bar or restaurant. See the questions What sample to test for COVID-19: nasopharyngeal or buccal? and False positives, false negatives, sensitivity, specificity: what are we talking about?
At the beginning of the epidemic, temperature measurements, for example using a non-contact infrared thermometer, were introduced in some countries at the entrances to airports, nursing homes, and workplaces. Today, it is considered that such temperature screening is of extremely limited effectiveness for several reasons. Firstly, fever can be masked by taking drugs such as paracetamol (acetaminophen). Second, it is estimated that half of all contaminations are due to asymptomatic people (who have not yet developed symptoms or who will never develop them). Finally, about 55% of people with mild to moderate forms of COVID-19 do not have a fever (not including people with a fever who are not infected with coronavirus). For example, since September 2020, temperature checks at many airports have been stopped.
Alternative approaches to biological testing are being developed. They are promising but, for the time being, they have been tested on too few people and their sensitivity and specificity do not yet give satisfactory results to be used at the individual level. It is therefore unclear whether and when these alternative tests will be available for use in the pandemic. See the question False positives, false negatives, sensitivity, specificity: what are we talking about?
Preliminary analyses suggest that people infected with COVID-19 could potentially be detected before the onset of symptoms with a connected watch that measures heartbeat and steps. For example, a two-tiered alert system based on the occurrence of extreme elevations of resting heart rate relative to the individual baseline allowed 86% of the 24 people infected with coronavirus out of a cohort of approximately 5,000 participants to be retrospectively detected. This method is not specific to COVID-19: it detects various cases of respiratory viral infection. It is in development and needs to be developed on a larger number of people.
Voice analysis is another avenue being explored to detect the presence of the SARS-CoV-2 coronavirus in infected people. Voice, breath, cough contain a lot of information and the analysis of vocal clues is currently the subject of much research in the diagnosis of neurodegenerative diseases such as Alzheimer's disease. To improve the detection of COVID-19, the idea is to extract sound clues - known as biomarkers - from COVID-19 in people's coughs. The forced or natural cough of individuals is recorded and then analysed automatically using machine learning techniques. This technique could be useful to speed up large-scale screening, at low cost and without the logistical need for medical analysis.
Another approach is to use trained dogs to detect coronavirus-infected people through their scent. Several dogs have been trained for several months but most of these studies have not yet been published. The researchers working with the dogs met in early November 2020 at an online conference called "International K9 Team" to share their preliminary results and improve the coordination of their research. So far, experiments with dogs are promising but have not been conducted on enough people to know whether the detection of COVID-19 by dogs could have applications during the pandemic.
The study of these new leads to detect COVID-19 is a very active and promising area of research. If you want to help cough research, you can register here:
for people COVID-negative and COVID-positive:
for COVID-positive people only:
On September 9, 2020, the U.S. CDC modified its airport surveillance strategy and prioritized public health measures other than temperature taking to reduce the risk of transmission of COVID-19 related to travel. These measures include: health education for passengers before departure, during the flight, and after arrival; a robust response to the disease at airports; voluntary collection of passenger contact details by electronic means as proposed by some airlines to avoid long queues, congestion, and delays associated with manual data collection; potential tests to reduce the risk of transmission of the virus responsible for COVID-19 and the movement of the virus from one location to another; country-specific risk assessments to help passengers make informed decisions about travel risks; improved training and education of transportation partners and U.S. ports of entry to ensure recognition of the disease and immediate notification to the CDC; and recommendations to passengers after arrival for self-monitoring and precautions to protect others, with enhanced precautions including staying home for 14 days whenever possible for those arriving from high-risk destinations.CDC. Federal Government Adjusts COVID-19 Entry Strategy for International Air Passengers. Media Statement. 9 Sept 2020.
In this modeling study, it is considered that cases of SARS-CoV-2 contamination originate in 46% of cases from presymptomatic (before presenting symptoms) persons, in 38% of cases from symptomatic persons, in 10% of cases from asymptomatic persons (who never present symptoms) and in 6% of cases from indirect transmission via the environment. Estimates for the last two routes are speculative.Ferretti, L., Wymant, C., Kendall, M., Zhao, L., Nurtay, A., Abeler-Dörner, L., ... & Fraser, C. (2020). Quantifying SARS-CoV-2 transmission suggests epidemic control with digital contact tracing. Science, 368(6491).
Only 45% of people with mild to moderate forms of COVID-19 have a fever.Lechien, J. R., Chiesa-Estomba, C. M., Place, S., Van Laethem, Y., Cabaraux, P., Mat, Q., ... & Barillari, M. R. (2020). Clinical and epidemiological characteristics of 1,420 European patients with mild-to-moderate coronavirus disease 2019. Journal of internal medicine.
Analysis of physiological and activity data from 5,262 participants, 32 of whom were infected with COVID-19. 26 of those infected (81%) had alterations in their heart rate, number of daily steps, or sleep time. Using retrospective smartwatch data, the researchers estimated that with their two-level alert system based on the occurrence of extreme elevations of resting heart rate relative to the individual baseline, 62.5% of cases of COVID-19 (15/24) could have been detected before the onset of symptoms in real time. This approach is a general detection method and currently cannot distinguish infections caused by the SARS-CoV-2 coronavirus from those caused by other respiratory viruses.Mishra, T., Wang, M., Metwally, A. A., Bogu, G. K., Brooks, A. W., Bahmani, A., ... & Fay, B. (2020). Pre-symptomatic detection of COVID-19 from smartwatch data. Nature Biomedical Engineering, 1-13.
Digital machine learning tools are being developed to screen through the analysis of vocal cues present in the cough (forced or not) of people infected with SARS-CoV-2. This study focuses on one of the essential points which is to differentiate a characteristic cough from COVID-19 from a cough whose cause is not COVID-19. In this study, the authors developed a smartphone application that records coughs and gives a result in 2 minutes. Depending on the machine learning models used, the sensitivity and specificity results are more or less similar to the RT-PCR tests. This study provides a "proof of concept" with encouraging results that need to be confirmed on larger cohorts of subjects. The advantage is that this type of technique can be used for screening, which is particularly useful for mass testing of large numbers of people on a daily basis at very low cost.Imran, A., Posokhova, I., Qureshi, H. N., Masood, U., Riaz, S., Ali, K., ... & Nabeel, M. (2020). AI4COVID-19: AI enabled preliminary diagnosis for COVID-19 from cough samples via an app. arXiv preprint arXiv:2004.01275.
Study based on cough records from databases in India: 3,621 people including 2,001 who tested positive for COVID-19 and another database of 1,039 people, 376 of whom are positive for COVID-19. The reference test used is RT-PCR. Several learning machine models are tested and the final model allows the detection of an infection at SARS-CoV-2 with a sensitivity of 90% with a specificity of 31%. These results confirm that this type of analysis is promising but needs to be improved for medical use.Bagad, P., Dalmia, A., Doshi, J., Nagrani, A., Bhamare, P., Mahale, A., ... & Panicker, R. (2020). Cough Against COVID: Evidence of COVID-19 Signature in Cough Sounds. arXiv preprint arXiv:2009.08790.
Study based on the analysis of sound cues to identify clinical signs of the neurodegenerative genetic disease, Huntington's disease. Analyses are carried out on 45 patients with Huntington's disease: 16 individuals who carry the mutation but are not sick (asymptomatic) and 24 healthy subjects. According to the sound clues used, the machine learning model correctly identifies 56% of the individuals (sick, asymptomatic, healthy subjects).Riad, R., Titeux, H., Lemoine, L., Montillot, J., Bagnou, J. H., Cao, X. N., ... & Bachoud-Lévi, A. C. (2020). Vocal markers from sustained phonation in Huntington's Disease. arXiv preprint arXiv:2006.05365.
Study based on the analysis of sound cues to diagnose the neurodegenerative disease, Alzheimer's. A learning machine model is trained on 108 people (including 54 Alzheimer patients) and tested on 48 people (including 24 Alzheimer patients). The results show that the model correctly identifies 75% of the individuals.Luz, S., Haider, F., de la Fuente, S., Fromm, D., & MacWhinney, B. (2020). Alzheimer's Dementia Recognition through Spontaneous Speech: The ADReSS Challenge. arXiv preprint arXiv:2004.06833.
Article dated November 2020 summarizing work on dogs that have been trained to detect people infected with SARS-CoV-2 coronavirus. It presents a very promising study at an airport in Lebanon: the dogs examined 1,680 passengers and found 158 cases of COVID-19 that have been confirmed by PCR tests. The animals correctly identified the negative cases with 100% accuracy, and correctly detected 92% of the positive cases, according to unpublished results.Else, H. (2020). Can dogs smell COVID? Here's what the science says. Nature.
In this pilot study, researchers trained 8 dogs for one week from samples taken from the trachea and mouth of seven hospitalized individuals with COVID-19 and seven uninfected people. The dogs were able to identify 83% of the COVID-19 positive cases and 96% of the cases which were negative.Jendrny, P., Schulz, C., Twele, F., Meller, S., von Köckritz-Blickwede, M., Osterhaus, A. D. M. E., ... & Manns, M. P. (2020). Scent dog identification of samples from COVID-19 patients-a pilot study. BMC infectious diseases, 20(1), 1-7.
Researchers trained 8 dogs to detect COVID-19 in 198 sweat samples from various hospitals, about half of which were from people with COVID-19. When these were hidden in a row of negative samples, the dogs identified positive samples 83-100% of the time. The article does not say how well the dogs identified negative test results.Grandjean, D., Sarkis, R., Tourtier, J. P., Julien, C., & Desquilbet, L. (2020). Detection dogs as a help in the detection of COVID-19Can the dog alert on COVID-19 positive persons by sniffing axillary sweat samples? Proof-of-concept study. bioRxiv.
Presentation of the project underway in the United Kingdom to train dogs to detect COVID-19.Jones, R. T., Guest, C., Lindsay, S. W., Kleinschmidt, I., Bradley, J., Dewhirst, S., ... & Logan, J. G. (2020). Could bio-detection dogs be used to limit the spread of COVID-19 by travellers?. Journal of travel medicine.