Perioperative medicine & outcomes
Towards reliable prediction of intraoperative hypotension: a cross-center evaluation of deep learning-based and MAP-derived methods
Study context
- Population
- Surgical patients, Karolinska University Hospital (Sweden) and VitalDB (Korea) datasets
- Exposure
- Deep learning (AI) model vs MAP threshold method for predicting intraoperative hypotension
- Outcome
- Prediction of intraoperative hypotension
Methodological detailsmulticentre, cross-validation, cross-age validation
- Multicentre
- Cross-validation
- Cross-age validation
Main finding
As reported in the abstract
- AI model performs better than MAP threshold in more ambiguous (Gray Zone, MAP 65-75 mmHg) cases
- With clearly separated MAP values between classes, both methods performed similarly well
Observational data: associations, not effects.
Further findings
Cross-dataset generalizability
- Models trained on datasets with more Gray Zone cases generalized better to datasets with clearer class separation; the reverse struggled
- Standardizing MAP difference between positive and negative samples virtually eliminated class separation; inflated performance attributable to selection bias
Cross-age and ASA validation
- Models trained on older patients generalized better to younger cohorts
- Differences in ASA classification had minimal effect
Parts of this Brief are withheld: the automatic checks could not confirm them against the abstract.
Original publication
Towards reliable prediction of intraoperative hypotension: a cross-center evaluation of deep learning-based and MAP-derived methods. J Clin Monit Comput. Published 12 September 2025. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →