Covid-19 and Beyond: Role of AI in Health Crises
Faisal Farooq, PhD
Head of the Digital Health Research Group
Qatar Computing Research Institute
There has been a lot of hype around Artificial Intelligence (AI) recently - some myth and some facts. However, the undeniable reality is we are currently in the biggest health crisis of our lifetime – the COVID-19 pandemic. So, to ground it in reality, the objective of this session is to understand and outline some of the current applications of AI in the fight against thepandemicas well as to discuss how AI can impact healthcare crises in the future. The US Federal Emergency Management Agency (FEMA) broadly divides disaster management into four phases : 1) Preparedeness(before Covid-19 started); 2) Response (initial reaction to the pandemic); 3) Recovery(returning to normal); and 4) Mitigation(preventing further spread and surges). AI has found its application in all four stages of this pandemic.
Considering there are no known therapeutic interventions for Covid-19, most of the emphasis for the initial responsehas been on non-pharmaceutical interventions such as social distancing, wearing masks and lockdown, with the sole aim of slowing down and even stopping the spread. AI-based models that utilize significantly more information (mobility data, population characteristics, traffic patterns, etc.) to augment the conventional compartmental epidemiological models have informed better public policy guidelines. In addition, there are various symptom-tracking tools, self-assessment and acuity triage bots as well as wearable device-based digital monitoring systems driven by AI technology that have enabled home care for less sick patients, thereby making more beds available for acutely sick patients.
As we head into the recovery phase, AI is being heavily used in parts of the world to drive decisions of when and how to institute restrictions such that there is a balance between minimizing the economic and social impact versus maximizing the public health benefit . Multiple studies utilizing AI models are being conducted to understand the diverse clinical pathways and patient outcomes. AI is also being used to distill and answer specific high-level questions from large amounts of knowledge and data such as the CORD-19  dataset.
As we look forward and try to mitigate the impact of Covid-19 or similar health crises, AI-based ‘surveillance’ systems that continuously mine global public health data or provide possible transmission information for contagious diseases are being proposed. On one hand, they can be very promising, on the other, they raise important ethical and privacy questions that the policy makers need to be proactive about.
Lastly, we need to be preparedfor such emergencies. The reason I write about this at the end despite being the first stage of management is because Covid-19 caught us woefully unprepared. While health data platforms and exchanges have often been on “roadmaps” and “strategies” of most, they have not received the attention they should. Early identification systems, syndromic surveillance, predicting pathogenesis using phylogenetic trees as well as rapid optimal stockpiling and deployment of resources is all possible and achievable using AI techniques, if and only if we have sufficient good quality data being collected and analyzed. So, AI holds much more promise but for that we need to invest in preparedness, and for AI the currency is data - it’s an insurance policy that everyone hates paying for, until one needs it.
 National Governors’ Association. Comprehensive Emergency Management – A Governor’s Guide. Washington DC: Defense Civil Preparedness Agency, May, 1979