Mohamadlou, H. | bioRxiv Science

bioRxiv2018

Multicenter validation of a machine learning algorithm for 48 hour all-cause mortality prediction

PurposeThis study evaluates a machine-learning-based mortality prediction tool.\n\nMaterials and MethodsWe conducted a retrospective study with data drawn from three academic health centers. Inpatients of at least 18 years of age and with at least one observation of each vital sign were included. Predictions were made at 12, 24, and 48 hours before death. Models fit to training data from each institution were evaluated on hold-out test data from the same institution and data from the remaining institutions. Predictions were compared to those of qSOFA and MEWS using area under the receiver operating characteristic curve (AUROC).\n\nResultsFor training and testing on data from a single institution, machine learning predictions averaged AUROCs of 0.97, 0.96, and 0.95 across institutional test sets for 12-, 24-, and 48-hour predictions, respectively. When trained and tested on data from different hospitals, the algorithm achieved AUROC up to 0.95, 0.93, and 0.91, for 12-, 24-, and 48-hour predictions, respectively. MEWS and qSOFA had average 48-hour AUROCs of 0.86 and 0.82, respectively.\n\nConclusionThis algorithm may help identify patients in need of increased levels of clinical care.

bioinformatics↗

bioRxiv2017

Prediction of Acute Kidney Injury with a Machine Learning Algorithm using Electronic Health Record Data

BackgroundA major problem in treating acute kidney injury (AKI) is that clinical criteria for recognition are markers of established kidney damage or impaired function; treatment before such damage manifests is desirable. Clinicians could intervene during what may be a crucial stage for preventing permanent kidney injury if patients with incipient AKI and those at high risk of developing AKI could be identified.\n\nMethodsWe used a machine learning technique, boosted ensembles of decision trees, to train an AKI prediction tool on retrospective data from inpatients at Stanford Medical Center and intensive care unit patients at Beth Israel Deaconess Medical Center. We tested the algorithms ability to detect AKI at onset, and to predict AKI 12, 24, 48, and 72 hours before onset, and compared its 3-fold cross-validation performance to the SOFA score for AKI identification in terms of Area Under the Receiver Operating Characteristic (AUROC).\n\nResultsThe prediction algorithm achieves AUROC of 0.872 (95% CI 0.867, 0.878) for AKI onset detection, superior to the SOFA score AUROC of 0.815 (P < 0.01). At 72 hours before onset, the algorithm achieves AUROC of 0.728 (95% CI 0.719, 0.737), compared to the SOFA score AUROC of 0.720 (P < 0.01).\n\nConclusionsThe results of these experiments suggest that a machine-learning-based AKI prediction tool may offer important prognostic capabilities for determining which patients are likely to suffer AKI, potentially allowing clinicians to intervene before kidney damage manifests.

bioinformatics↗