Researchers from three institutes develop algorithm for better contact-tracing
Researchers from three premier scientific institutes in India have developed an algorithm for ‘improved’ Covid-19 contact-tracing via mobile phones that can identify indirect contacts who may be at risk of contracting the virus.
Currently digital contract-tracing algorithms – including the government’s Aarogya Setu Covid-19 app – are undocumented but seem to be largely restricted to identifying and following-up on individuals who may have come in direct contact with a Covid-19 positive person.
The newly created algorithm can identify individuals who are at risk of contracting the disease from high-risk contacts of a Covid positive person who are not themselves known to be infected, said researchers at the Bengaluru-based Indian Institute of Science (IISc) and National Centre for Biological Sciences (NCBS), and Institute of Mathematical Sciences (IMSc) at Chennai.
“For example, if you met A, and I later met you, and A is found to be infected, not just you but I too could be at risk. Our methodology calculates this probabilistically,” said Rahul Siddharthan, professor of computational biology at IMSc. “In our simulations, we found our method to be more effective than direct contact tracing methods; that is, for a given ‘false positive rate’ (fraction of negative cases that we wrongly call positive), we get a higher ‘true positive rate’ (fraction of positive cases that we correctly call positive).”
On Friday, their paper ‘Risk assessment via layered mobile contact tracing for epidemiological intervention’ was uploaded on medRxiv, the preprint server for health sciences operated by US-based Cold Spring Harbor Laboratory.
With concerns about surveillance and privacy surrounding Aarogya Setu and similar apps, the new algorithm stores most data locally on the individual’s phone and requires minimal communication with a server; thus it can be secured to ensure privacy.
For instance, only when a person is tested and found to be infected, his/her contact traces need to be informed/identified, possibly via a centralised system. The algorithm maintains privacy in terms of contact history and location history, and shares only the risk factor to the government.
“Current apps are opaque on what exactly they do with the collected data and how they process it,” said Vishwesha Guttal, associate professor at IISc. “By making our algorithm public, we bring the issues of privacy concerns to the forefront. For apps that do not publish the algorithm, there is no way to know how they process the data, and therefore, serious concerns of privacy will exist.”
“Current apps do not present their methodology of how they assess risk of an individual having contracted the disease,” said professor Sandeep Krishna, member of the Simons Centre for the Study of Living Machines, NCBS. “We hope that our work will encourage app developers (including that of Aarogya Setu) to make their algorithm public. One can then compare all the algorithms and adopt the one that’s most efficient.”
The three-member team has shared their work with K VijayRaghavan, principal scientific advisor to the government of India.