Valvular & structural heart disease
Predictors of Length-of-Stay Among Transcatheter Aortic Valve Replacement Patients Using a Supervised Machine Learning Algorithm
Authors' conclusion
ML algorithms may have an important role in identifying novel predictors of short and prolonged length of stay following TAVR. These efforts may facilitate targeted quality improvement programs to decrease length of stay post-TAVR.
Study context
- Population
- Outpatient transcatheter aortic valve replacement (TAVR) procedures, Biome data set, United States, 2017–2021outpatient TAVR procedures from 21 centers
- Participants
- 9,172 procedures
- Comparison
- ML models compared to standard multivariable models in predicting LOS
Methodological details21 centers, 70% development sample, 30% test sample
- 21 centers
- 70% development sample, 30% test sample
Main finding
As reported in the abstract
Important predictors identified
Variables included as important predictors for the short length of stay (SLOS) model
Twenty
Variables included as important predictors for the long length of stay (LLOS) model
22
Further findings
| Outcome | SLOS ML model | LLOS ML model | Standard multivariable model | |
|---|---|---|---|---|
| Predictive power of ML models vs standard multivariable model · ML models more robust than the standard multivariable model | ||||
| Sensitivity | 0.81 | 0.45 | Not reported in the abstract | Standard multivariable model: not reported in the abstract |
| Specificity | 0.70 | 0.94 | Not reported in the abstract | Standard multivariable model: not reported in the abstract |
| AUC | 0.82 | 0.85 | Not reported in the abstract | Standard multivariable model: not reported in the abstract |
| SLOS AUC | Not reported in the abstract | Not reported in the abstract | 0.65 | SLOS ML model, LLOS ML model: not reported in the abstract |
| LLOS AUC | Not reported in the abstract | Not reported in the abstract | 0.65 | SLOS ML model, LLOS ML model: not reported in the abstract |
Previously unreported predictors
- Procedural duration, postprocedure physical therapy, and procedure day of the week
Original publication
Predictors of Length-of-Stay Among Transcatheter Aortic Valve Replacement Patients Using a Supervised Machine Learning Algorithm. JACC Adv. Published 24 June 2025. doi:10.
Summarised from the published abstract; every number is checked against it