Diakonessenhuis deploys an AI model to more accurately predict surgery scheduling. The model predicts the duration of operations, with the aim of optimising the utilisation of operating theatres (ORs). In this way, Diakonessenhuis aims to be able to help more patients.
The model was developed by PhD student Hayo Bos, affiliated with the University of Twente. Diakonessenhuis speaks of impressive results. The AI model predicts the consequences of planning for the surgery complex. For example, it maps how likely it is that an operation will be delayed, and how likely it is that space will be created to help another patient.
Making predictions more accurate
The AI model supports the existing workflow. In it, staff from the Admission Planning department schedule operations of patients in the OR schedule. In doing so, the electronic patient record HIX automatically gives an average time per surgery and operator. However, this prediction does not always match reality, reports Diakonessenhuis.
"After filling our schedules completely, we use the OR predictor to see if the predicted OR times match the HIX-scheduled times. If, according to the predictor, there is a chance of walk-in or walk-out, we contact the doctor and decide whether to keep or adjust the schedule. This allowed us to schedule additional patients more often," explains Admissions Planning Team Manager Karim El Ghoulbzouri.
'Realistic OR planning is essential'
Anne Krajnc, head of OKC/CSA: "For us, realistic OK planning is essential. The OR predictor helps us to make optimal use of OR hours, without OR colleagues often having to work overtime. This is increasingly important."
The model is part of the integral capacity management portal (ICM portal). Every 15 minutes, the AI model provides planning feedback, supporting planners.
Predict occupancy on Acute Admissions Department
The AI model is unique in the Netherlands, according to the Diakonessenhuis. However, various other AI models are in use in Dutch hospitals to optimise care. One example is an AI model used by Leiden University Medical Centre (LUMC) to predict the number of admissions to the Acute Admissions Department. The tool provides immediate insight into the number of beds required.
"Most patients in the Acute Admissions Department come from the Emergency Department," says Britt de Boer, nursing admissions coordinator at the Acute Admissions Department (AOA). "With us are patients who can go home within 48 hours or go to a specialist department. Before the admission predictor was in place, we never knew exactly how many patients to expect when. As a result, we could not always free up beds in time. Sometimes it seemed quiet, but we would suddenly get a call from the Emergency Department: 'It's busy, we are expecting five admissions for you.'"
The tool was developed by CAIRELab, LUMC's AI expertise centre. Using AI, the tool predicts how many patients go from the Emergency Department to the Acute Admissions Department. It takes all kinds of variables from the patient record into account. These include measurements of vital body functions and requests for blood tests and radiological examinations.
Improve planning in intensive care units
Another example is an AI model developed by Gelre Hospitals in collaboration with SAS and KPMG. The model predicts the influx of patients into the intensive care unit (ICU), with the aim of improving staffing levels. This is important because, due to factors such as increasing staff shortages and absenteeism caused by high workloads, staffing levels can pose a challenge.
Intensivist Nicolas Schroten: "A better prediction of the number of patients coming to the ICU can help improve flexible scheduling, allowing you to give staff rest when you can and deploy extra staff when needed. To this end, it is important to have a good understanding of all patients being treated in the hospital and at risk of deteriorating. A lot of important information is recorded in the electronic patient record (EHR), but it is impossible for healthcare professionals to continuously monitor this for all patients. This is where artificial intelligence (AI) can help us. By providing timely alerts and filtering the essential data needed to plan ahead. This allows us to provide better care, for more patients and prevent staff from dropping out due to overwork."
Author: Wouter Hoeffnagel
