ICU nurses are working under enormous pressure, and this situation has to change. Artificial intelligence can help in this process, claims Diederik Gommers, professor of intensive care medicine at the Erasmus Medical Centre (MC).
Tekst: Gert-Jan van den Bemd | Translated by Tony Parr | 12 april 2020
‘The coronavirus crisis suddenly opened our eyes,’ Professor Gommers explains, ‘but the shortage of ICU nurses is in fact a problem that’s afflicted our intensive care units (ICUs) for many years.’ He explains how it’s come about: ‘ICU nurses need to feel that they’re in control. They need to be sure that they have full control over the patients for whom they’re responsible. If that’s the case, they go home with a spring in the step. But if it isn’t, if they lose control, it grinds them down.’ It is this desire – and indeed the need – to have full control over patients that results in such a busy schedule of activities. Every ICU patient is surrounded by a battery of complex equipment that closely monitors their heart, lungs, kidneys and other organs and systems. ICU nurses need to be fully alert at all times to all the signals emanating from these machines. In other words, they have to stay on their toes to remain in full control.
The equipment surrounding an ICU patient produces a wealth of data on which doctors and nurses can base certain conclusions. Has the patient’s blood pressure fallen? Is the patient developing an infection in their bloodstream? What about the kidney function? And how are the lungs doing? Professor Gommers believes that artificial intelligence (AI) can help to streamline and make better use of this information overload – and thus relieving the burden placed on ICU nurses. ‘We need to pool our medical expertise with that of engineers and data scientists in order to develop a dashboard – a panel for each individual patient that can help ICU nurses to take clinical decisions.’
Red, orange or green
Junior intensive care doctor Michel van Genderen is closely involved in the development of the dashboard. Together with Professor Gommers and intensive care doctor Jasper van Bommel, he’s trying to use new techniques, such as big data and data analysis, to raise the standard of care. ‘We’re designing a system that can generate real-time information on a patient’s status, which can tell us things like who needs extra care and whose life is under threat. It’s not an alarm system, mind. The idea is to use it for monitoring and reporting. The system used at the moment use has all sorts of different alarms – one for the heartbeat, another for blood pressure, and so the list goes on. We want to end this situation. A constant stream of false alarms is like the boy who cried wolf. In the end, you don’t take any notice. The dashboard only issues updates when they’re really needed. That’s one of its main benefits.’ Professor Gommers adds: ‘We could have a window on the dashboard that uses colour coding to show whether the patient’s condition is stable or deteriorating: red, orange or green, depending on the situation.’
Professor Gommers: ‘The dashboard uses algorithms to convert the measurements made by the ICU equipment, as well as test results and data from electronic patient records, for example, into useful information. The doctors and nurses on the ICU have a regular routine that involves checking monitors and examining lab results. We can use all this expertise as the basis for an algorithm. And much more than that: the dashboard can detect what the human eye fails to see, or what the human brain can’t recognise or understand. That’s a massive bonus. The fact is that we have a huge amount of data available to us, including data that we don’t actually use. For example, we measure the patient’s heart rate variability – that is, the variability over time of the intervals between individual heartbeats. This information tells us something about the way in which the autonomic nervous system is working. Although we measure the heart rate variability every few milliseconds, this information is not displayed graphically. All we get to see is an ECG (electrocardiogram). A dashboard could combine this underlying data with the results of lab tests to generate extremely valuable information.’
Dr van Genderen reckons that the dashboard could be used more widely than just on an ICU. ‘We could feed information into the dashboard from wards where patients were treated before they were transferred to the ICU. If we attach sensors that record the vital functions, we can then track the changes in the patient’s status over time. That would give us a clear clinical picture, including all the signals associated with any changes. That, in turn, would enable us to plan interventions at an earlier stage and perhaps prevent other patients from having to be transferred to an ICU.’ Professors Gommers sees the system becoming self-learning in the future. ‘The output it generates could be used as input for refining future analyses. It could also produce information on mortality risks and the expected length of stay. This type of information is tremendously useful – not just for the ICU, but for the whole logistic chain in the hospital.’
What’s the current status? Dr van Genderen: ‘We’re currently working on data storage. There’s a massive amount of data coming on stream from the Covid-19 patients on the ICU. What we need to work out is how best to store it all. We’re also developing the algorithm. And we’re trying to design a forecasting model that can let us know, on the basis of our measurements – using a sort of traffic light – that one particular patient may need more care than another.’
Professor Gommers and Dr van Genderen readily acknowledge the value of working closely with engineers. ‘We need to learn each other’s languages if we are to work well together,’ Professor Gommers is keen to stress. ‘Provided that we understand each other, we can set about implementing technical innovations for improving patient care. As a university medical centre, we believe that this can create marvellous opportunities for research. It’s going to be a paradise for PhD students with an interest in medical technology, AI and big data.’