Abstract:
Elective surgeries are one of the services provided at tertiary healthcare institutions in
New Zealand. We present a case study of the scheduling of ear, nose, and throat (ENT)
elective surgeries at the Waitemata District Health Board (DHB). The durations of surgeries,
necessary for the planning of surgical lists, are estimated by surgeons and may have
high variability. As such, the booking clerks have a challenging task of planning surgical
lists that both maximise theatre utilisation levels and minimise the chances of list overruns.
In surgical list planning, the following are required: (1) accurate predictions of the
surgery durations and (2) an appropriate performance measure that evaluates surgical
lists.
For (1), we compare the predictions from different models with the surgeons’ estimates.
As no single model predicts the entire data equally well, we develop a hybrid
model that allows each subset of the data to be allocated to the best-performing prediction
model. The hybrid model is subsequently extended to allow all prediction models
to contribute to the final estimate for each surgery duration by using a weighted sum of
predictions. As we are proposing new prediction techniques, the performances of these
techniques are established using synthetic datasets before we evaluate them on real-life
datasets. Finally, we compare several prediction techniques under an operational setting
and recommend a prediction technique for use at the Waitemata DHB.
For (2), we introduce the operating room (OR) scheduling metric which assigns a
score to each surgical list. As compared to probability-based and/or expectation-based
measures, scores from the OR scheduling metric are better reflections of the risks of
undesirable theatre outcomes associated with a surgical list. Using historical surgical
lists, we demonstrate an association between scores from the OR scheduling metric and
theatre utilisation rates. Therefore, the OR scheduling metric can be used as an evaluation
tool for surgical lists. As the scores are compatible across surgical lists of different session
durations, we perform a simulation study showing that the OR scheduling metric can
also be used as a scheduling tool for surgical lists.