Using tech to treat sepsis in urgent care  

In February, the monthly Medical Data Science Meetup focused on new and innovative ways of treating sepsis. Taking place monthly at Amsterdam UMC, the event brings medical professionals and data scientists together for a night of learning, discussion and pizza while closing the gap between these two interconnected fields.

Fighting a deadly condition

The world’s third leading cause of death and the primary cause of mortality in hospitals, sepsis occurs when the body’s response to infection causes it to damage its own organs and tissues. Annually, it’s estimated that sepsis results in six million deaths worldwide, taking the lives of roughly 260,000 people in the USA and 44,000 in the UK.

However, the best way of treating sepsis is uncertain. Evidence indicates current practices regarding the administration of vasopressors and intravenous fluids, which keep the heart beating and increase blood pressure, can harm a significant number of patients. This session’s guest speaker, Dr Matthieu Komorowski, introduced a machine learning algorithm he had developed which used reinforcement learning to predict the optimal course of action to treat sepsis in clinical practice. Dr Komorowski, who is currently pursuing a PhD at Imperial College and a research fellowship in intensive care at Charing Cross Hospital in London, also emphasised how a close partnership between data scientists and clinicians can lead to the development of tools that can improve patient outcomes while disrupting the current paradigm for creating medical evidence.

An innovative new approach

Dr Komorowski’s model was developed using data from around 96,000 sepsis patients who were treated in over 130 intensive care units across the US during a 15-year period. He and his team used an artificial intelligence (AI) clinician to study the information, which can extract implicit knowledge from an amount of patient data greatly exceeding the lifetime experience of human clinicians. In doing so, Dr Komorowski showed how it could be used to predict the best treatment for individual patients by analysing an array of treatment decisions and their outcomes.

Despite still being in an early stage of development, Dr Komorowski said that tests showed that the treatments chosen by the AI clinician would likely have resulted in better outcomes for the patients that than those selected by doctors. Additionally, patients with the lowest mortality rates were those for whom the doses administered by human clinicians matched those recommended by AI.

Discussing the benefits of working with AI, Dr Komorowski said: “Imagine a big cloud of patients with sepsis. Some of them died, some of them survived, but this is your new patient, so you look at patients who are similar to them, and what treatment was given to them, and what was the outcome. Among all the practice variations, the algorithm is going to tease out what worked and what didn’t.”

Dr Komorowski also believes technology can play an important role in helping physicians overcome the limitations they face every day. “In an ideal world, a problem would be solved because we’d have complete knowledge of everything – all the diseases and all the treatments. That would be simple, but it’s impossible. For the other option, imagine you had access to a large data set in which the patients look very similar to the one in front of you, including what treatment they received throughout their hospital stay and what was the final outcome, and you had a way to model all of this. Maybe that would reduce uncertainty.”

Bringing tech to the bedside

In the future, Dr Komorowski hopes machine learning can be used to help medical professionals determine the best course of action when caring for patients, and he’s not the only one. Event organiser Dr Lucas Fleuren is also working to make sure technological advances make a difference in people’s lives by developing a ‘pipeline’ with his colleagues which can help to reduce the time it takes to implement new data science applications in clinical practice.

“If you look at the literature, there are thousands of models that have been published on machine learning in the intensive care unit, but patients don’t benefit from them yet,” he says. “What we need is this pipeline to see how we can get those models to the bedside. It’s what we’re trying to focus on in this institution…it’s our job to come up with the models and write the proposals for the trials.”

Hopefully, technology will soon be a regular part of helping patients enjoy the best outcomes, whether they’re suffering from a serious condition like sepsis, recovering from injury, or simply dealing with the aches and pains of everyday life.

For more reports from previous Medical Data plus Pizza Meet-ups, click here.

About Amsterdam Medical Data Science

The Amsterdam Medical Data Science Group meetup takes place on the third Tuesday of every month in the Delta Room at the VU University Medical Center Amsterdam’s Intensive Care Unit. The next meeting will take place on 19 february.

To find out more visit:

The Amsterdam Medical Data Science Group meetings are supported by The Right Data Right Now consortium, which includes Amsterdam UMC, OLVG, Vrije Universiteit, Pacmed, and the Amsterdam Economic Board.

21 February 2019

Read more about

Contact us



Want to read more articles like this one?
Follow us daily on LinkedIn, X and sign up for our Board Update.

Read more