Leveraging clinical data across healthcare institutions for continual learning of predictive risk models

• Type of Material: Article
• Authors: Fatemeh Amrollahi, University of California San Diego; Supreeth P Shashikumar, University of California San Diego; Andre Holder, Emory University; Shamim Nemati, Emory University
• Language: English
• Date: 2022-05-19
• Publisher: NATURE PORTFOLIO
• Publication Version: Final Published Version
• Copyright Statement: © The Author(s) 2022
• License: Creative Commons Attribution 4.0 International
• Final Published Version (URL): http://dx.doi.org/10.1038/s41598-022-12497-7
• Title of Journal or Parent Work: SCIENTIFIC REPORTS
• Volume: 12
• Issue: 1
• Start Page: 8380
• End Page: 8380
• Grant/Funding Information: Computational resources for the reported experiments were made possible via a generous cloud credit grant from Amazon, as a part of an AWS Research Award to Dr. Shashikumar.; Dr. Nemati is funded by the National Institutes of Health (#R01LM013998 and #R35GM143121), and the Gordon and Betty Moore Foundation (#GBMF9052).; Dr. Holder is supported by the National Institute of General Medical Sciences of the National Institutes of Health (#K23GM37182), and from Baxter International.
• Keywords: Science & Technology; SEPSIS; Science & Technology - Other Topics; Multidisciplinary Sciences; ARTIFICIAL-INTELLIGENCE; CANCER
• Research Categories: Health Sciences, Medicine and Surgery

Abstract

The inherent flexibility of machine learning-based clinical predictive models to learn from episodes of patient care at a new institution (site-specific training) comes at the cost of performance degradation when applied to external patient cohorts. To exploit the full potential of cross-institutional clinical big data, machine learning systems must gain the ability to transfer their knowledge across institutional boundaries and learn from new episodes of patient care without forgetting previously learned patterns. In this work, we developed a privacy-preserving learning algorithm named WUPERR (Weight Uncertainty Propagation and Episodic Representation Replay) and validated the algorithm in the context of early prediction of sepsis using data from over 104,000 patients across four distinct healthcare systems. We tested the hypothesis, that the proposed continual learning algorithm can maintain higher predictive performance than competing methods on previous cohorts once it has been trained on a new patient cohort. In the sepsis prediction task, after incremental training of a deep learning model across four hospital systems (namely hospitals H-A, H-B, H-C, and H-D), WUPERR maintained the highest positive predictive value across the first three hospitals compared to a baseline transfer learning approach (H-A: 39.27% vs. 31.27%, H-B: 25.34% vs. 22.34%, H-C: 30.33% vs. 28.33%). The proposed approach has the potential to construct more generalizable models that can learn from cross-institutional clinical big data in a privacy-preserving manner.