Hemodynamics, fluids & transfusion
American society of anesthesiologists physical status classification significantly affects the performances of machine learning models in intraoperative hypotension inference
Authors' conclusion
Different ASA physical status classes present different data distributions, and thus calls for distinct machine learning models to improve prediction accuracy and reduce predictive uncertainty.
Study context
- Population
- Hysterectomy surgeries in patients ≥18 years, UofL health hospital, June 2020–July 2021
- Participants
- 562 hysterectomy surgeries
- Exposure
- Machine learning models (Bayesian generalized linear model (BGLM), Bayesian neural network (BNN), BNNMR, Gaussian Process (GP)) trained per American Society of Anesthesiologists (ASA) physical status class vs without controlling ASA class
- Outcome
- Sensitivity for binary hypotensive event, by ASA class
Methodological detailssingle-centre, machine learning prediction modelling
- Single-centre
- Machine learning prediction modelling
Main finding
As reported in the abstract
- ASA 2: BNN highest sensitivity
- 0.429
- ASA 3: BNN highest sensitivity
- 0.415
- ASA 4: BNNMR and GP tied highest sensitivity
- 0.857
- Not divided by ASA class: BNNMR sensitivity
- 0.031
- Not divided by ASA class: BNN sensitivity
- 0.429
- Not divided by ASA class: BGLM sensitivity
- 0.165
- Not divided by ASA class: GP sensitivity
- 0.305
Observational data: associations, not effects.
Further findings
SBP prediction root mean squared errors (RMSE), GP regression
- ASA 1
- 2.072
- ASA 2
- 7.539
- ASA 3
- 9.214
- ASA 4
- 0.295
- Trained without controlling ASA class
- 126.894
- Lowest RMSE of all models
SBP prediction RMSE, BGLM
- ASA 1
- 2.175
- ASA 2
- 13.861
- ASA 3
- 17.560
- ASA 4
- 22.426
DBP prediction RMSE, GP regression
- ASA 1
- 2.152
- ASA 2
- 6.573
- ASA 3
- 5.371
- ASA 4
- 0.831
- Trained without controlling ASA class
- 8.084
- Lowest RMSEs; other models far higher
Width of 95% CI of mean SBP and DBP prediction
- GP regression gives narrower CI with much smaller margin of error across all four ASA classes
Parts of this Brief are withheld: the automatic checks could not confirm them against the abstract.
Original publication
American society of anesthesiologists physical status classification significantly affects the performances of machine learning models in intraoperative hypotension inference. J Clin Anesth. Published 2 November 2023. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →