Perioperative medicine & outcomes
Development of an automated, general-purpose prediction tool for postoperative respiratory failure using machine learning: A retrospective cohort study
Authors' conclusion
We developed a general-purpose, machine learning powered prediction tool with superior performance for research and quality-based definitions of postoperative respiratory failure.
Study context
- Population
- Anesthetic procedures from 1/2018 to 6/2021
- Participants
- 101,455 anesthetic procedures
- Exposure
- RESPIRE model vs ARISCAT and SPORC-1
- Outcome
- Postoperative respiratory failure (Standardized Endpoints in Perioperative Medicine consensus definition)
Methodological detailsmachine learning (Random Forest), prediction model development and validation
- Machine learning (Random Forest)
- Prediction model development and validation
Main finding
As reported in the abstract
Accuracy for postoperative respiratory failure (validation cohort)
| Group | Area under the receiver operating curve (AUROC) |
|---|---|
| RESPIRE | 0.93 (95% CI, 0.92-0.95) |
| ARISCAT | 0.82 |
| SPORC-1 | 0.82 |
Area under the receiver operating curve (AUROC)
RESPIRE0.93 (95% CI, 0.92-0.95)
ARISCAT0.82
SPORC-10.82
- RESPIRE vs ARISCAT: P for difference < 0.0001
- RESPIRE vs SPORC-1: P for difference < 0.0001
Observational data: associations, not effects.
Secondary outcomes
| Outcome | RESPIRE | ARISCAT | SPORC-1 |
|---|---|---|---|
| PPV and false positive rate at comparable 80-90% sensitivitiesScore cut-offs determined in a separate test cohort | |||
| PPV | 11%, 95% CI: 10-12% | 4% | 4% |
| False positive rate | 12%, 95% CI: 12-13% | 37% | 37% |
Respiratory quality metrics (NSQIP, Society of Thoracic Surgeons, CMS)
- RESPIRE better predicted the established quality metrics for postoperative respiratory failure
Parts of this Brief are withheld: the automatic checks could not confirm them against the abstract.
Original publication
Development of an automated, general-purpose prediction tool for postoperative respiratory failure using machine learning: A retrospective cohort study. J Clin Anesth. Published 7 July 2023. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →