Perioperative medicine & outcomes
Development, external validation and integration into clinical workflow of machine learning models to support pre‐operative assessment in the UK
Authors' conclusion
We conclude that conducting modelling together with systems development can yield accurate prediction models that may be implemented directly into electronic health records.
Study context
- Population
- UK surgical patients undergoing pre-operative assessment, two NHS Trusts using different electronic record systems110,732 development
- Comparison
- ASA physical status prediction model vs Mortality prediction model
- Outcome
- Categorisation into lower- and higher-risk cohorts by predicted ASA physical status (1 or 2 vs. 3-5) and 30-day postoperative mortality risk
Methodological detailsmachine learning, two NHS Trusts, Certified electronic linkages between Smart PreOp and NHS England GP Connect to retrieve GP record data into pre-operative questionnaires
- Machine learning
- Two NHS Trusts
- Certified electronic linkages between Smart PreOp and NHS England GP Connect to retrieve GP record data into pre-operative questionnaires
Performance
As reported in the abstract
Identification of lower-risk (ASA physical status 1 or 2) patients, external validation
| Test or model | Recall | Precision |
|---|---|---|
| ASA physical status prediction model | 0.69 | 0.95 |
ASA physical status prediction model
Recall0.69
Precision0.95
Further findings
30-day postoperative mortality prediction, external validation
- Mortality model discriminated well but was poorly calibrated
Original publication
Development, external validation and integration into clinical workflow of machine learning models to support pre‐operative assessment in the UK. Anaesthesia. Published 14 September 2025. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →