Multitask prediction of organ dysfunction in the intensive care unit using sequential subnetwork routing. (21st June 2021)

• Record Type: Journal Article
• Authors: Roy, Subhrajit; Mincu, Diana; Loreaux, Eric; Mottram, Anne; Protsyuk, Ivan; Harris, Natalie; Xue, Yuan; Schrouff, Jessica; Montgomery, Hugh; Connell, Alistair; Tomasev, Nenad; Karthikesalingam, Alan; Seneviratne, Martin
• Is Part Of: Journal of the American Medical Informatics Association. Volume 28:Number 9(2021)
• Page Start: 1936
• Page End: 1946
• Publication Date: 2021-06-21
• Subjects: Machine Learning -- Deep Learning -- Multitask Learning -- Electronic Health Records -- Intensive Care; Medical informatics -- Periodicals; Information Services -- Periodicals; Medical Informatics -- Periodicals; Médecine -- Informatique -- Périodiques; Informatica; Geneeskunde; Informatique médicale; Computer network resources; Electronic journals; 610.285
• DOI: 10.1093/jamia/ocab101
• Languages: English
• ISSNs: 1067-5027

Abstract

Abstract: Objective: Multitask learning (MTL) using electronic health records allows concurrent prediction of multiple endpoints. MTL has shown promise in improving model performance and training efficiency; however, it often suffers from negative transfer – impaired learning if tasks are not appropriately selected. We introduce a sequential subnetwork routing (SeqSNR) architecture that uses soft parameter sharing to find related tasks and encourage cross-learning between them. Materials and Methods: Using the MIMIC-III (Medical Information Mart for Intensive Care-III) dataset, we train deep neural network models to predict the onset of 6 endpoints including specific organ dysfunctions and general clinical outcomes: acute kidney injury, continuous renal replacement therapy, mechanical ventilation, vasoactive medications, mortality, and length of stay. We compare single-task (ST) models with naive multitask and SeqSNR in terms of discriminative performance and label efficiency. Results: SeqSNR showed a modest yet statistically significant performance boost across 4 of 6 tasks compared with ST and naive multitasking. When the size of the training dataset was reduced for a given task (label efficiency), SeqSNR outperformed ST for all cases showing an average area under the precision-recall curve boost of 2.1%, 2.9%, and 2.1% for tasks using 1%, 5%, and 10% of labels, respectively. Conclusions: The SeqSNR architecture shows superior label efficiency compared with ST and naive multitasking, suggesting utility in scenarios in which endpoint labels are difficult to ascertain.