Neuroanesthesia & neurocritical care
Early prediction of ventricular peritoneal shunt dependency in aneurysmal subarachnoid haemorrhage patients by recurrent neural network-based machine learning using routine intensive care unit data
Authors' conclusion
RNN-based machine learning has the potential to predict VP shunt dependency on Day 4 after ictus in aSAH patients using routine data collected in the ICU.
Study context
- Population
- Patients with aneurysmal subarachnoid haemorrhage (aSAH) treated in the intensive care unit (ICU), November 2010 to May 2020
- Participants
- 602 patients
- Outcome
- Prediction of ventriculoperitoneal (VP) shunt dependency
Methodological detailssingle-centre, machine learning prediction model (recurrent neural network (RNN)), More than 120 parameters analysed (routine neurocritical care data, vital signs, blood gas analyses); RNNs and gradient boosting machines evaluated
- Single-centre
- Machine learning prediction model (RNN)
- More than 120 parameters analysed (routine neurocritical care data, vital signs, blood gas analyses); RNNs and gradient boosting machines evaluated
Main finding
As reported in the abstract
Prediction of VP shunt dependency by RNN
- AUC-ROC after 1 day of ICU stay
- 0.77 (CI: 0.75-0.79)
- AUC-ROC on Day 4
- 0.81 (CI: 0.79-0.84)
- Accuracy on Day 4
- 76% (CI: 75.98-83.09%)
- Sensitivity on Day 4
- 85% (CI: 83-88%)
- Specificity on Day 4
- 74% (CI: 71-78%)
Observational data: associations, not effects.
Original publication
Early prediction of ventricular peritoneal shunt dependency in aneurysmal subarachnoid haemorrhage patients by recurrent neural network-based machine learning using routine intensive care unit data. J Clin Monit Comput. Published 21 March 2024. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →