Model based on artificial intelligence helps COVID-19 doctors to make better decisions

>> Read this in Dutch

One question I always ask myself is: can we provide more help for our public-sector clients through better insights derived from available data? This question was even more pressing this spring, when doctors swiftly had to learn how to deal with an unknown virus and understand a disease progression that was not in the textbooks. During this period, I was in contact with IC doctors Michiel van Genderen and Diederik Gommers at Erasmus MC. We had already discussed data solutions several times in the past, and I wondered if we might be able to use our expertise to provide support for the hospital in this area during the crisis. They had a few ideas, and not long after, my colleagues Martijn Ludwig and Maren Diether started working on this project. As I was curious about their experiences, I decided to sit down with them. 

COVID-19 patients spending less time in IC

‘The doctors there were faced with an acute challenge,’ says Maren, looking back to the start of the project. ‘Because very little was known about the disease, they had to manage on the basis of their own observations. One important question was why some patients developed certain complications and others didn’t. While some doctors thought they might know, they lacked the data to support their theory.’ 

Maren and Martijn used data from 108 COVID-19 patients who had been in IC in the spring in order to train a model. To this end, they drew on the knowledge of SAS regarding data use in healthcare. Maren and Martijn consciously chose a relatively simple model: a decision tree. Martijn explains, ‘In a decision tree, you can easily grasp each step, even as a layman. We felt that this link to practice was important, because it greatly increases the chance that the model will actually be used.’ 

Within just two weeks, the doctors were provided with a workable model that supported their hypothesis. The model is now supporting the doctors, identifying complications that, if detected early, might reduce someone’s stay at the ICU.

Better care thanks to new data insights

‘And that’s exactly what we want to achieve through use of data,’ says Martijn. ‘Analyse the situation together with the professionals, find out precisely what their requirements are and then select a useful model that we can subsequently train with data.’ Maren adds, ‘In this way, we ensure that specialists and other healthcare professionals can do their work better, supported by facts.’

In my opinion, this is exactly what the role of technology and AI models should be. Data solutions are possible in every phase of the healthcare process, in prevention, in predicting the course of an illness, in personalised healthcare, etc. When interacting with clients, we always aim to find out the best way to support their services: we use data solutions to assist professionals so that they can provide better healthcare or improve other aspects of their work. These solutions only work if they are developed jointly. We can only have true impact if we combine data expertise and healthcare expertise. 

Scientific publication

The results of the project are described in a scientific article published recently in The Journal of Clinical and Translational Research. Hence, the research not only helps the doctors at Erasmus MC in their day-to-day work but also contributes to the scientific understanding of COVID-19.  . Zo helpt het onderzoek niet alleen de artsen in het Erasmus MC in de praktijk, maar draagt het ook bij aan de verdere ontwikkeling van wetenschappelijke kennis over COVID-19. 

The health(care) future of the Netherlands

Deloitte wants to play an active role in the health ecosystem and to strengthen and accelerate the improvements described in this article. We will do this through multiple events: hackathons, roundtable discussions and C-level dialogues. The aim is to bring together the various parties in the new health ecosystem and explore which alliances can create value.

Read more about the health(care) future of the Netherlands.