When there are rushes in healthcare settings and it gets busier than staff can handle, this can have major consequences – for patients and staff alike. For patients, there is an increased risk of complications, they remain hospitalised for longer and satisfaction levels plummet. For staff, it leads to burnout and turnover.
That’s why a major Danish A&E department gave us this task: can you develop a system that predicts how busy it will get – before we know it?
Using our collective insight into the healthcare system combined with mathematics, we developed a computer system that uses machine learning (AI) to predict busyness over a period of up to 12 hours into the future. It draws on different data sources including how busy it usually is, current major events, the prevailing weather and several other factors and based on this, undertakes a number of complicated mathematical calculations.
The output is a graph that shows how many future patients will come in hour by hour. We present both the exact number that is expected to arrive each hour as well as how certain the system is in its calculation.
This allows staff to decide how much value they want to allocate to the predictions made by the system and thus which actions to take.
Praemostro has been in use at an A&E department since August 2022. Time has shown that staff trust the predictions and use them every day. And it’s easy for them to do that: the predictions hit the mark more than 95% of the time to within +/- one patient an hour. It delivers a sufficient level of accuracy that enables deviations to be absorbed into operations without problems.
Because it’s good to know what the future holds, staff also use the system in other contexts:
When there are staff on sick leave, the system is used to determine whether the staff on duty can handle the expected number of patients for the day or if extra staff will need to be called in.
Healthcare staff are often called in and that’s why it is preferable not to disturb them on their days off unless absolutely necessary.
If the system shows that it will not be busy, the remaining staff will be able to handle the tasks themselves and those who have days off will remain on their days off – blissfully undisturbed.
Although patients often arrive according to a pattern (mostly late afternoon and early evening) there is a lot of fluctuation, which is why staff use the system to plan their workday.
When should they take breaks, when do they have time to clean up and when should they eat?
The system can help to optimise planning for all these areas.
When you get a sense that things are starting to get busy, it’s important to be able to keep up. If there are more tasks than the staff can handle, there is a risk of falling behind. And although healthcare staff are used to being busy, they can become overwhelmed. And if they can’t keep up, that in turn has consequences for the patients. They remained hospitalised for longer and there is an increased risk of complications.
That’s why staff also check our models when things are ramping up. If it’s just a short rush, then it can be handled, and the backlog will be cleared quickly. But if it’s an extended period of busyness, extra help will need to be summoned. People will either have to be called in from home or other staff seconded from the other hospital departments.
