Development of a machine learning model to predict intensive care unit bed demand for adult elective surgical patients at a large United Kingdom National Health Service Trust
Authors' conclusion
We demonstrate that aggregating individual-level ICU admission predictions for elective surgeries provides a bed demand estimate that improves on the current electronic bed demand indicator 1 day before surgery at two out of three sites conducting the majority of inpatient elective surgery at our large UK NHS Trust.
Study context
- Population
- Adult elective inpatient surgeries at three sites of a large UK National Health Service (NHS) trust, 1 May 2019 to 31 December 2023
- Participants
- 38,656 elective inpatient surgeries
- Comparison
- CoreML vs current electronic ICU bed demand indicator; CoreML vs FullML
- Outcome
- Prediction of ICU admission after elective surgery, aggregated to forecast ICU bed demand
- Follow-up
- 1 day before surgery
Methodological detailsretrospective dataset, prospective validation, multicentre (three sites), tree-based machine learning
- Retrospective dataset
- Prospective validation
- Multicentre (three sites)
- Tree-based machine learning
Performance
As reported in the abstract
Root mean square error at site A, CoreML vs current electronic indicator
1.28 vs 1.64
Root mean square error at site C, CoreML vs current electronic indicator
0.76 vs 1.16
Further findings
- Individual-level ICU admission prediction, CoreML, prospective evaluation · 1 day before surgery
- 0.88 · AUC
Elective surgical cases admitted to ICU with an ICU bed electronically requested, prospective evaluation
- Cases admitted to ICU after surgery that had an ICU bed electronically requested
- 71.6%
Aggregate prediction, CoreML vs FullML
- CoreML outperformed FullML at all sites in prospective evaluation; converse in retrospective evaluation
Original publication
Composed by AI from the published abstract and checked against it. How Briefs are made →