The US had confirmed its first case of a new coronavirus (2019-nCOV) originated from Wuhan, China. According to the Centers for Disease Control and Prevention (CDC), the diagnosed patient is a Washington resident in his 30s. He arrived at the Seattle-Tacoma International Airport on 15 January, before compulsory health screenings were in place at various US airport. He sought medical attention four days later and is now in isolation. The authority has started tracing other individuals who have been in contact with the patient.
Thus far, this new coronavirus had infected 440 people in China and other parts of Asia and had resulted in nine deaths. Initially, it was unclear whether the virus is communicable between people but a Chinese respiratory expert confirmed so on 20 January. The World Health Organization is hosting a meeting today (22 January) to decide if they should declare the outbreak as an international public health emergency.
All along, communicable disease is an adverse challenge to many public healthcare infrastructures around the world. As seen in the cases of Severe Acute Respiratory Syndrome (SARS) in year 2003; Influenza A H1N1 (i.e., swine flu) in year 2009; Ebola and Middle East Respiratory Syndrome (MERS) in year 2014, to Zika virus in year 2016. The World Bank estimated China’s SARS-related losses to be about $14.8 billion even though the US and Europe were largely spared.
Maximizing on outreach effort
Whilst new technology is not a silver bullet, some researchers are tapping onto artificial intelligence (AI) to prevent or reduce the impact of infectious diseases in this densely populated and interconnected era. For example, in 2018, three scientists from the Center for AI in Society at the University of Southern California developed an algorithm which assists public health agencies to achieve the maximum outreach effort by cost-effectively allocating their resources.
The trio believe outreach campaigns encourage high-risk but yet to be diagnosed individuals to seek treatment. As such, their AI model takes into consideration human behavior and disease transmission patterns to give the authorities a clearer picture of which group of people will benefit most from which form of communications and campaigns.
The algorithm was subsequently tested using real-world data on tuberculosis (TB) prevention in India and gonorrhea prevention in the US and it successfully prevented 8000 cases of TB and 20,000 cases of gonorrhea respectively. The researchers also presented their findings at the 32rd Association for the Advancement of AI (AAAI) conference.
A separate group of UK and US investigators made a similar attempt in year 2014, as they use AI to better locate individuals who may be down with HIV but are unaware of their disease status. The model had effectively prevented 5% of new infections, at a time when the UK authority suspected a quarter of British may be living with HIV but are oblivious they have the virus.
Going back to the source
Researchers from Cary Institute of Ecosystem Studies in New York; University of Georgia; University of California Berkeley, and Massey University in New Zealand employed machine learning to predict they types of bats that are more likely to carry filoviruses like Ebola. The team used 57 different types of factors including biographical factors to ecology data to make the predictions and engaged with an independent team to validate the model. Researchers believe while there’s still a lot of work to be done in reducing the spread and effect of Ebola, looking into animal surveillance may forecast the likely outbreak sites and facilitate necessary prevention work.
A Malaysian startup – Artificial Intelligence in Medical Epidemiology (AIME) Healthcare is also interested in predicting where the next outbreak is likely to occur. The initiative developed an AI-driven dengue fever tracking system using past and existing outbreak data, so that public health authorities will have an insight on where they should deploy the most resources to put a stop in time for the next possible outbreak.