Data sits at the very core of the global response to the coronavirus, from governments and healthcare bodies, to education leaders and transport operators. The more data that’s captured about COVID-19’s geographic spread, rate of infection and post-infection outcomes, the better able public bodies and policy makers will be to plan for what comes next, while containing what’s happening now.
This isn’t new: Public health experts, academia and health data professionals have been using data models for years in the analysis of epidemics, such as the annual influenza outbreaks, and to predict their likely path. This modeling is typically based on broad and generic data sets that cover characteristics such as age and risk factors in a population. Now, however, the data models being used to look at the dynamics of the COVID-19 pandemic are changing.
Accurate insights derived from data models, real-time data and AI
The world is learning more every day about transmission of this novel coronavirus and its effects. We can see that the virus spreads differently from country to country, region to region, and across cities, towns and villages. The generic data models used in the past do not have the necessary granularity to tell a particular hospital or local authority what impact COVID-19 might have on them. They are also not dynamic enough to enable an effective response to the daily-changing pandemic landscape.
But these differential equation-based models are still proving pivotal in the fight against COVID-19.
At Capgemini Invent, we are working with universities on how to adapt their data modeling to the evolving situation that public health bodies and authorities are responding to.
And we’re collaborating with health data professionals who understand the structure of health data, what data is kept (and where), and how to access it from within the data ecosystem.
The use of mathematical modeling, data science, and statistics, informed by medical knowledge, is also helping to reveal what the risk factors are for different populations at a far more granular level than with previous data modeling – not just age, but gender, ethnicity, and even places where there were large gatherings before the lockdown that have since had a high infection rate.
Predicting the path and impact of COVID-19
The value of this lies in helping organisations better anticipate future scenarios as the pandemic evolves. They can make more reliable projections of the impact of different lockdown or public health policies on their own unique situations.
For example, using artificial intelligence (AI) and other technology tools, we can take real-time data captured on the ground to model a hospital’s response in terms of capacity planning.
So, earlier in the pandemic, the focus might have been on computing how many beds were needed on average per day across all the different wards to cope with COVID-19 admissions; and what non-COVID related treatment could be postponed to create capacity. Once the peak has passed, the analysis may turn to what resources can be safely diverted back to other areas, for example when it might be possible to re-start non-COVID-19 surgical procedures.
Our data scientists and mathematical modeling experts are working with medical and other healthcare professionals in France and the United Kingdom to define stochastic models, whereby the probable outcomes are used to predict what’s next. For example, if the modeling suggests an imminent spike in demand, the hospital might liaise with another local center to manage capacity. We can model a patient’s pathway through the hospital, forecasting which department they will come into contact with, and over what timeline. This helps to predict the likely volume of infection within the hospital environment on a day-to-day basis.
Tapping into the power of AI
We see a productive interaction between AI, data and the models being generated. Firstly, it’s important to remember that the standard approaches used to model epidemics are traditionally calibrated once at the outset of the disease. With the current pandemic, they need to be updated and refined to reflect the reality shown by the data to give more accurate projections. This is where AI can help: it enables us to incorporate real-time and diverse data to improve the forecast accuracy of these models. Further, where testing is limited, AI can be used to pose questions via questionnaires, chatbots, etc., on a large scale and gather input to train classification algorithms that evaluate the likelihood of someone having COVID-19.
The pandemic has created an immediate need for real-time modeling to inform policy making and management of precious healthcare resources. It will now be interesting to see the impact on broader transformation programs utilising AI and data.
Organisations in all sectors have been developing their capabilities in data science for years, but the pandemic has been a catalyst to accelerating that transformation. Will this continue after the current crisis?
Certainly, the speed at which COVID-19 spread across the world and the daily changing lockdown measures provide food for thought.
Hospitals have never previously needed such granular data modeling, nor experienced such a dynamically changing modeling cycle. With AI making it possible to push data models to the limit, it is more than likely that hospitals, other public sector organisations, transport operators and more will continue to embrace data-driven transformation to better prepare for future scenarios, while managing current operations more effectively.
In the private sector, the pandemic has also highlighted many areas where more accurate forecasting and planning could help meet customer needs more effectively. For consumer products and retail companies, the shift to more online retailing, demand for specific products that have been in short supply, and the need to anticipate bottlenecks or delays in the supply chain are clear examples. For the travel and hospitality sectors, accurately anticipating customer demand as lockdown restrictions are eased will be key to ramping up the availability of services. As long as the pandemic disrupts business as usual, our assumptions about what comes next will depend on our data.