Last month, when the US confirmed its first case of the novel coronavirus (2019-nCOV) originated from Wuhan, China, AIMed talked about how artificial intelligence (AI) may be helpful in stopping an epidemic by maximizing on outreach effort and underpinning possible sources.

About a week later, BlueDot, a global health monitoring platform based in Canada said its AI-driven “epidemiologist” was able to give company’s clients advance warning of a possible 2019-nCOV outbreak as early as 31 December. While the US Centers for Disease Control and Prevention (CDC) and World Health Organization (WHO) released words of caution to the public on 6 and 9 January respectively.

According to the Founder and Chief Executive Officer (CEO) of BlueDot, Kamran Khan, who used to by a hospital infectious disease specialist, the company’s algorithm does not rely on social media postings because of the overwhelmingly messy data. Rather, they tracked global airline ticketing data which forecast where and when residents in affected areas will be heading next. That’s how they accurately pointed out Bangkok, Tokyo, and Seoul to be the next few places with initial appearances.

Indeed, the role of AI is no longer dismissive. In China, the ongoing novel coronavirus emergency had gradually surfaced some of its mass surveillance network. Public security cameras with facial recognition capabilities would identify individuals who do not oblige to the 14-day compulsory quarantine and send warnings to both the authorities and their companies. Recognition accuracies are not at all compromised even if the person is wearing a face mask.

At the same time, local telecom companies are also following the movements of users, informing them if a flight or train they are about to take, will have suspected coronavirus carrier. Although there has been some discomfort expressed anonymously over the internet, most Chinese citizens had generally accepted it as a way for additional precaution.

The role of conversational AI

On the other hand, some US healthcare startups are also stepping up to update their algorithms and ensure their medical chatbots would screen users and advise them on actions to take when cough or fever emerged. For example, Seattle-based 98point6 told Stat it had presented two suspected cases of 2019-nCOV to CDC but thus far, no patient was isolated or referred for further testing as a result.

Nevertheless, most technology-driven healthcare companies see themselves as the first line of defence: providing users with preliminary triage to keep them away from the long-hour of waiting at the already overwhelmed hospitals or clinics. Shall a patient indicate the presence of flu-like symptoms like shortness of breath, cough, and fever, the chatbots will proceed to ask if they have travelled to China or if they have been in close contact with anyone who has.

98point6 made its coronavirus screening algorithm live on 24 January, a few days after the first US confirmed case. Mainly, users are offered remote primary care as its AI chats with them before referring them to a human physician shall the need arise.

Another company, Bright.md pushed out its coronavirus screening feature on 29 January. Similarly, AI was deployed for distant consultations with any suspected cases automatically referred to human physicians. Again, Bright.md hopes to enhance the screening process and make sure patients experiencing any symptoms are quickly and appropriately treated.

The question of necessity

However, because symptoms are also present in flu, it can be challenging for algorithms to rightly distinguish if a person is affected by coronavirus or other medical conditions. On 2 February, the CDC had issued a guide on how to evaluate and report patients who may be suspected of an infection.

The question thus become, anyone may be able to access themselves based on these guidelines, so, is the assistance provided by a medical chatbot necessary in this case? As of now, all 2019-nCOV cases present in the US were confirmed through the US Food and Drug Administration cleared laboratory testing conducted at the CDC, whereby patients’ nasal and throat specimens and blood serum are shipped to Atlanta and tested.

As such, some companies chose to target misconception or wrong information instead. Like Buoy Health, they realized fear may actually travel faster than the virus itself, so its application will ask patients if they are concerned of any specific diagnosis and follow up questions will be directed towards what patients have indicated. For example, if the result turns out to be coronavirus, patients will be asked about their symptoms and risk factors before they are being categorized into high or low risk group. Those in the former category will be asked to call and seek help from the nearest emergency room.

The company believes by seeing patients in every three seconds, they can facilitate the identification of an outbreak before someone reaches the hospital. At the moment, Buoy Health is partnering with HealthMap, a disease detection project initiated by researchers from Boston Children’s Hospital to map potential clusters of new illness. If they are successful in locating a particular disease occurring within an area, they will be able to call upon and prepare local health officials and medical facilities before an outbreak.

*

Author Bio

Hazel Tang A science writer with data background and an interest in the current affair, culture, and arts; a no-med from an (almost) all-med family. Follow on Twitter.