Computer models could help reduce surgery cancellations
Each day, a hospital’s Intensive Care Unit (ICU) needs to make important decisions around staffing and resources to prepare for the next day’s incoming patients and scheduled surgeries. Unfortunately, ICUs are highly variable environments, known for unexpected changes in urgency for its services.
June Lau, a Doctoral Candidate in the Department of Statistics at the University of Auckland, is using NeSI supercomputers help ICUs better predict and respond to the day-to-day demands for staffing, beds, and equipment.
“It’s about ensuring they always have sufficient capacity for every patient that presents themselves to the ICU,” says Lau.
Her work to date has focused on cardiovascular intensive care. Generally, a cardiac ICU needs to account for three types of patients each day. Nearly 75% of patients are admitted for planned surgeries, such as those who require heart surgery but who have been on a waiting list. Alongside the planned surgeries, between seven to 10 elective surgeries are also scheduled each day.
The remaining capacity is held for emergency patients, such as those who have been in an accident, suffered a heart attack, or who are unexpectedly transferred in from a different hospital. These patients are the hardest to plan for because it is difficult to predict how many will arrive, when they will arrive, how long they will stay, and what they will require for care.
Each ICU has on-call nurses and other resources it can tap into as spare capacity if needed. However, an unexpected rush of acute trauma patients can use up space capacity without warning. When demand is at its height, one of the ways to free up staff and beds is to postpone the elective surgeries.
“That puts the [elective surgery] patient and their family in obviously a pretty anxious state,” says Lau. “Also, if the patient and their family have come from rural parts of NZ to support the patient and are told that their surgery is going to be postponed, that is not a great patient outcome.”
Lau’s work aims to change that. She’s developing a computer model using discrete event simulation methods to help ICUs better predict and plan for short- and long-term capacity in the ward. Ultimately, she hopes, it will minimise the chances of elective surgeries being cancelled.
Her research, which is supported by an award from The Green Lane Research and Educational Fund Board, is part of a larger programme of work being undertaken by researchers at the University of Auckland in collaboration with District Health Boards (DHB).
“It’s a difficult problem because there are really short stays in the ICU unit but also really long stays,” she says. “As a result, there is a lot of variability involved in trying to plan for and predict future occupancy of the unit. In addition, the unit itself is quite small, so any small changes or inaccuracies in a model really stand out.”
Lau has been working closely with the cardiac ICU at the Auckland Hospital to build an anonymised replica of its historical data for her computer model. To account for the variability of emergency patients, the model runs thousands of years worth of simulations. It’s possible to do this on a desktop computer, Lau says, but the enormity of the task means it would take at least two to three days to complete.
“Whereas on NeSI, in the parallel computing performance that we have access to, it means we can run thousands of years of simulations in just over two hours,” she says. “It’s been amazing to be able to use NeSI to do that because it means we can change different parameters in our simulation model and through that we can do wider scenario analysis.”
The better analysis they can do, the better metrics ICUs will have for predicting the number of incoming patients. In particular, knowing the chances of receiving acute trauma patients will hopefully mean fewer ‘surprises’ and so fewer cancelled elective surgeries.
Without access to a supercomputer, her project would have been especially challenging to complete, she says.
“I think it would have been very frustrating for me to do my research without NeSI,” she said. “I may have had to purchase another computer and just dedicate that one to just run models the whole time.”
Wrapping up this project in a timely manner was a priority for Lau, as it is her PhD project. She’s hoping to complete her thesis within the next 12 months and they are currently in the final stages of a paper on the outputs of simulation for submission to a clinician-based journal.
Alongside this project, she is also working with colleagues on another proposal to optimise and model treatment pathways in elective surgeries. It’s another simulation that requires them to run a large number of iterations of the same model, but using different data and performing scenario analysis around it.
“If we get funding for that research program, then it would definitely be something that is very suited to using NeSI,” she said.
- - - - - -
Do you have an example of how NeSI platforms have supported your work? We’re always looking for projects to feature as a case study. Get in touch by emailing email@example.com.