News, Voice

HTN Voice: Data and AI must play a frontline role in tackling the ongoing COVID-19 problem

By Orlando Agrippa, CEO of Draper & Dash Predictive Healthcare Analytics

There is a dawning realisation across the UK that the COVID-19 pandemic is not just going to be a one-off event. Unlike Asia and the Middle East, an outbreak of this scale is not something we have seen so close to home in decades, if not a century, and while few imagined it would reach its current level of impact when the first cases began to arise in China, it now looks like a problem set to impact our lives for years to come.

My organisation has been tracking the epidemic since its early days in Wuhan, examining the global picture on an ongoing basis in order to apply our knowledge and data analysis technology within the UK healthcare system and support NHS Trusts in their efforts to respond to the crisis.

The truth is, for most healthcare practitioners, data and AI are not subjects they think about in their daily lives. And yet, as Trusts have thrown themselves headfirst into preventing our health service from buckling under the strain of COVID-19, what we’ve seen is that robust data analysis and scenario modelling can make a huge difference to the effectiveness of each individual healthcare organisation’s response.

How NHS trusts managed the early stages of the pandemic

It’s been a frantic time for NHS Trust leadership teams. They’ve had to keep close to staff at every stage of the pandemic to understand the situation on the ground and work in partnership with clinicians to try and establish what might be needed in terms of additional ICU capacity.

The early days of the crisis were sobering to say the least, with constant reports of Italy’s struggles compounded by Imperial College London’s prediction that there could be half a million deaths across the UK. Within individual regions, Trusts acted hastily to shift around resources and attempt to expand ICU wards in the right locations, preventing the influx of COVID-19 patients from exceeding critical care bed capacity.

One of the key issues faced throughout this period was access to data. Trusts in the North of England were left to rely on anecdotal evidence from colleagues in London to try and build an understanding of what was to come. Many within the NHS thought that it would be a sprint and acted accordingly; now the growing realisation is that responding to COVID-19 will be a marathon, with the emphasis on trying to better control the severity of future waves.

Why better data analysis makes such a difference

It’s easily possible to accurately analysis and model data on an organisation by organisation basis in areas such as Emergency Department (ED) performance, workforce availability, equipment and resources, and bed capacity – all fundamental components in the running of any healthcare system. What we’ve seen throughout the pandemic is that the organisations with the best handle on their own data – combined with the ability to compare it to wider data sources – have invariably been more agile and effective in their response.

Perhaps the best example we’ve observed has been in the UAE. Abu Dhabi Health Services (SEHA) used data and AI technology to model all possible COVID-19 scenarios and optimise the way in which healthcare demand and capacity was tracked, managed and predicted across each of the organisation’s facilities. It used ongoing data analysis to deliver real-time insights in areas such as global growth curves, SIR modelling, testing strategies, health system capacities, patient profiles. It compared its own data against that of other countries at different stages of the outbreak and used AI to help better determine the progression of the virus in its own region.

In fact, using AI it is possible for any country to interrogate its own healthcare ‘profile’ against that of other countries in terms of population demographics, demand curves, case growth, mortality rates, and COVID-19 testing regiments. Taking this type of forensic approach to data can enable a country to accurately establish its closest nation match worldwide, and harness this analysis to produce stronger predictions about what will happen next.

Applying these lessons in the UK

SEHA’s data-driven approach serves as a blueprint for how AI can support the UK’s health services in the ongoing fight against COVID-19. It’s now perfectly possible to overlay publicly available country comparison data against the data sets of any local NHS Trust – for example, R-rate, length of stay, patient profile, and patient destination – to create a precise demand and allow each Trust to plan and model an appropriate response strategy.

At SEHA, tracking the data in this manner has enabled the organisation to equip hospitals with the right number of ventilators at any point, effectively doubling ICU bed capacity in a short period of time. And the applications of the technology do not end there. Unique population profiles within any city – for example, the blue collar versus white collar population split – can be accurately modelled to better predict the likely COVID-19 curve in a given area. Consider the implications of this in relation to the UK Government’s strategy of enforcing local lockdowns. The value-add is clear, all as a consequence of using existing data sources and deploying AI to quickly model scenarios in a way that would take human analysts months to prepare.

Recovery planning and tackling the next COVID-19 wave

Trusts across the UK have propelled straight from the first wave of COVID-19 into the recovery phase, and are now actively looking at predictions for this period to establish how best to handle those patients who have been avoiding or unable to use healthcare services over the past weeks.

One issue that is already emerging is the problem of comparing and sharing data across regions. Any efforts to optimise the patient pathway and cope with future upsurges in demand rely heavily on the availability and willingness to share data between healthcare organisations. This not just a UK challenge – globally, everyone is grappling with the issue of data sharing, not to mention disseminating what is learnt in one country to others. For the foreseeable future, we will likely need to rely on local and publicly available data in driving our algorithms and technology.

However, we already have a huge amount of data within these systems. The trick lies in how we handle and interpret it. Technology needs to be combined with human experience and good judgement. By working this into AI algorithms we will finally be able to break through the current rigid data systems to match the adaptability and agility of the virus itself.

Ultimately, we don’t yet know precisely what the winter will bring. Alongside the usual seasonal concerns, we may be dealing with a virus that has further adapted or mutated. Understanding what is happening within our own local systems is therefore key to our preparations. The UK’s health system is certainly in a better position to fight another wave of the virus than it was at the start of the year. However, we still need stronger predictions and better models of what the future will look like. A pro-active, data-driven approach to planning is something that has been proven in other countries to make a tangible, pronounced impact in the fight against COVID-19; it is something we must strive towards now while we still have the time to prepare.