Algorithm makes it easier to detect sepsis
28 October 2021
Each year, nearly 8,000 Swedes die of sepsis, commonly known as blood poisoning. Inger Persson and her team have developed the NAVOY Sepsis algorithm to aid healthcare professionals by enabling the early detection and treatment of sepsis, a disease that is difficult to diagnose and often progresses very quickly. The goal: to save more lives.
The idea originally came from one of Persson’s former colleagues in the pharmaceutical industry.
“He contacted me and asked if we could put our heads together and see if we could come up with something that enabled us to use the vast amounts of data available in healthcare,” she says.
They started by bouncing around ideas about which diseases could be predicted and where this would be most useful. They settled on sepsis early on, as it is a condition that is difficult to diagnose, has a high mortality rate, and lacks tools for seeing which patients are at risk of developing the disease.
“We saw that there was an opportunity to do something here, based entirely on data,” says Inger Persson, who has been researching and teaching at the Department of Statistics at Uppsala University for the past ten years.
Collaborating with sepsis researchers
That is how the idea for NAVOY Sepsis was born, where a combination of AI and existing patient data helps healthcare professionals detect sepsis at an early stage. To train the algorithm, they used a US dataset as an initial starting point. Early on, they could see that the algorithm detected patterns in the data that could not be detected by a human.
“We developed a prototype and at the same time began collaborating with one of Sweden’s leading sepsis researchers at Skåne University Hospital in Malmö,” explains Inger Persson.
The same hospital is currently in the final stages of a clinical study for which it has received a Vinnova grant of SEK three million. Proof that this innovation developed by Inger Persson and her team has great potential. The fact that the algorithm works on retrospective data has already been demonstrated in a recently published study.
Combining research with business development
Testing and developing the idea commercially did not initially come naturally to Inger Persson. Her partners saw it as a matter of course: if this is to be done, we must be able to make a living from it. They focused on the business concept, and Inger Persson on the science.
“I had no doubts about it, but it didn’t come naturally to me at first. That came later,” she says.
Inger Persson now combines her position at the University with work at the research company AlgoDx, which she co-founded.
“The intense first research phase is over, so now I can hand over a lot of my work. I have more of an advisory role in the company at the moment,” says Inger Persson.
Decision support for healthcare professionals
So how does the machine learning algorithm that will help healthcare professionals detect sepsis early work?
“The simplest way to look at it is as a mathematical formula,” explains Inger Persson.
Twenty or so parameters, coming from electronic health record systems, are fed into the platform that contains the algorithm. For example, it could be the patient’s body temperature or blood test results. The platform summarises data at a set time interval and feeds it into the algorithm. The algorithm looks at both new data and older data to capture changes over time, and then calculates a figure that shows how high the risk is that the patient will develop sepsis.
“If the patient is at risk of developing sepsis, the healthcare professionals are notified and then can decide what to do,” says Inger Persson.
At the moment, the focus is on patients in intensive care units. Patients with COVID-19 are being investigated separately, as there is a proven link between this disease and sepsis.
Interest in the solution is high
The next step is to conduct a study with an electronic health record system at Södersjukhuset in Stockholm. Interest in AlgoDx’s innovative solution is high, both in the healthcare sector and among financers. Today, the company is owned by Inger Persson and her two co-founders plus three investors.
“The thoughts and feelings out in the healthcare sector were mainly communicated to us through the sepsis researchers and healthcare professionals we work with. It is clear that this need exists,” says Inger Persson.
Because the need is great: 30 percent of those admitted to the ICU have or develop sepsis.
“The team is everything”
The challenges ahead lie in finding the best way to get the product to market. Defining the intended buyer, the healthcare sector or companies that make health record systems, and how to price the product are some of the challenges that Inger Persson sees going forward. There are also regulations that need to be followed in the development and commercialisation of the idea. She has the rest of her team on hand to help her resolve future issues.
“The team is everything. The three of us who founded AlgoDx have been a fantastic team, and we have now hired several other people,” says Inger Persson.
Next year, NAVOY Sepsis can begin to be used in hospitals, just four years after the idea was born.
What is sepsis?
- Each year, sepsis affects 49 million people worldwide, 11 million of whom die. Sepsis is an infection that affects the whole body and causes important organs to stop functioning properly.
- Sepsis starts with an infection somewhere in the body that activates the immune system. This activation can sometimes become unbalanced, leading to problems such as the lungs not being able to oxygenate the blood and the kidneys not working properly. One of the most common causes of sepsis is pneumonia, and it is more common in the elderly or those with a previous illness.
- Sepsis cannot be prevented in individual cases, but smoking, high alcohol consumption and being overweight increase the risk. It is very rare to become seriously ill with sepsis if you are under the age of 50 and otherwise healthy.
Subscribe to the Uppsala University newsletter
Search all Uppsala University news