Contrastive learning improves critical event prediction in COVID-19 patients

Tingyi Wanyan; Hossein Honarvar; Suraj K Jaladanki; Chengxi Zang; Nidhi Naik; Sulaiman Somani; Jessica K De Freitas; Ishan Paranjpe; Akhil Vaid; Jing Zhang; Riccardo Miotto; Zhangyang Wang; Girish N Nadkarni; Marinka Zitnik; Ariful Azad; Fei Wang; Ying Ding; Benjamin S Glicksberg

doi:10.1016/j.patter.2021.100389

Contrastive learning improves critical event prediction in COVID-19 patients

Patterns (N Y). 2021 Dec 10;2(12):100389. doi: 10.1016/j.patter.2021.100389. Epub 2021 Oct 25.

Authors

Tingyi Wanyan^{1

2}, Hossein Honarvar¹, Suraj K Jaladanki¹, Chengxi Zang³, Nidhi Naik¹, Sulaiman Somani¹, Jessica K De Freitas^{1

4}, Ishan Paranjpe¹, Akhil Vaid¹, Jing Zhang⁵, Riccardo Miotto¹, Zhangyang Wang⁶, Girish N Nadkarni^{1

7

8}, Marinka Zitnik⁹, Ariful Azad², Fei Wang³, Ying Ding^{10

11}, Benjamin S Glicksberg^{1

4}

Affiliations

¹ Hasso Plattner Institute for Digital Health at Mount Sinai, Icahn School of Medicine at Mount Sinai, New York, NY, USA.
² School of Informatics, Computing, and Engineering, Indiana University, Bloomington, IN, USA.
³ Department of Population Health Sciences, Weill Cornell Medicine, New York, NY, USA.
⁴ Department of Genetics and Genomic Sciences, Icahn School of Medicine at Mount Sinai, New York, NY, USA.
⁵ Renmin University of China, Beijing, China.
⁶ Department of Electrical and Computer Engineering, University of Texas at Austin, Austin, TX, USA.
⁷ Division of Nephrology, Department of Medicine, Icahn School of Medicine at Mount Sinai, New York, NY, USA.
⁸ The Charles Bronfman Institute for Personalized Medicine, Icahn School of Medicine at Mount Sinai, New York, NY, USA.
⁹ Department of Biomedical Informatics, Harvard University, USA.
¹⁰ Dell Medical School, University of Texas at Austin, Austin, TX, USA.
¹¹ School of Informatics, University of Texas at Austin, Austin, TX, USA.

Abstract

Deep learning (DL) models typically require large-scale, balanced training data to be robust, generalizable, and effective in the context of healthcare. This has been a major issue for developing DL models for the coronavirus disease 2019 (COVID-19) pandemic, where data are highly class imbalanced. Conventional approaches in DL use cross-entropy loss (CEL), which often suffers from poor margin classification. We show that contrastive loss (CL) improves the performance of CEL, especially in imbalanced electronic health records (EHR) data for COVID-19 analyses. We use a diverse EHR dataset to predict three outcomes: mortality, intubation, and intensive care unit (ICU) transfer in hospitalized COVID-19 patients over multiple time windows. To compare the performance of CEL and CL, models are tested on the full dataset and a restricted dataset. CL models consistently outperform CEL models, with differences ranging from 0.04 to 0.15 for area under the precision and recall curve (AUPRC) and 0.05 to 0.1 for area under the receiver-operating characteristic curve (AUROC).

Keywords: COVID-19; contrastive loss; deep learning; electronic health records; machine learning; recurrent neural network.

Grants and funding

UL1 TR001433/TR/NCATS NIH HHS/United States