Development and evaluation of a machine learning-based in-hospital COvid-19 disease outcome predictor (CODOP): a multicontinental retrospective study
- University of Turku, Finland
- Max Planck Institute of Experimental Medicine, Germany
- University of Malaga, Spain
- General University Hospital of Alicante, Spain
- Gregorio Marañón University Hospital, Spain
- Hospital Universitario 12 De Octubre, Spain
- Hospital Costa del Sol, Spain
- Hospital Británico of Buenos Aires, Argentina
- Hospital Santa Cruz - Caja Petrolera de Salud, Bolivia
- Hospital of San Juan de Dios, Bolivia
- Hospital Honduras Medical Centre, Honduras
- Hospital Velez Sarsfield, Argentina
- Hospital Italiano de Buenos Aires, Argentina
- Clausthal University of Technology, Germany
- Max Planck Institute for Experimental Medicine, Germany
Abstract
New SARS-CoV-2 variants, breakthrough infections, waning immunity, and sub-optimal vaccination rates account for surges of hospitalizations and deaths. There is an urgent need for clinically valuable and generalizable triage tools assisting the allocation of hospital resources, particularly in resource-limited countries. We developed and validate CODOP, a machine learning-based tool for predicting the clinical outcome of hospitalized COVID-19 patients. CODOP was trained, tested and validated with six cohorts encompassing 29223 COVID-19 patients from more than 150 hospitals in Spain, the USA and Latin America during 2020-22. CODOP uses 12 clinical parameters commonly measured at hospital admission for reaching high discriminative ability up to nine days before clinical resolution (AUROC: 0·90-0·96), it is well calibrated, and it enables an effective dynamic risk stratification during hospitalization. Furthermore, CODOP maintains its predictive ability independently of the virus variant and the vaccination status. To reckon with the fluctuating pressure levels in hospitals during the pandemic, we offer two online CODOP calculators, suited for undertriage or overtriage scenarios, validated with a cohort of patients from 42 hospitals in three Latin American countries (78-100% sensitivity and 89-97% specificity). The performance of CODOP in heterogeneous and geographically disperse patient cohorts and the easiness of use strongly suggest its clinical utility, particularly in resource-limited countries.
Data availability
The raw patient data used in this study are not freely available due to legal restrictions of the ethical committees of the different hospitals. However, they can be accessed upon request to the Scientific Committees of these organisms. An exception to this is the patient data from the USA cohort, which has been published elsewhere. All the model's numerical data necessary to generate all figures can be found in the submitted source data tables. Furthermore, all supplementary tables can be found in Supplementary File 1.
Article and author information
Author details
Juan Miguel Antón-Santos
Jose Manuel Ramos-Rincon
David Gómez-Varela
Funding
Max Planck Society (Publication cost)
- David Gómez-Varela
The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.
Reviewing Editor
- Evangelos J Giamarellos-Bourboulis, National and Kapodistrian University of Athens, Medical School, Greece
Version history
- Preprint posted: September 22, 2021 (view preprint)
- Received: November 30, 2021
- Accepted: April 24, 2022
- Accepted Manuscript published: May 17, 2022 (version 1)
- Version of Record published: May 24, 2022 (version 2)
Copyright
© 2022, Klén et al.
This article is distributed under the terms of the Creative Commons Attribution License permitting unrestricted use and redistribution provided that the original author and source are credited.
Metrics
-
- 1,402
- views
-
- 268
- downloads
-
- 5
- citations
Views, downloads and citations are aggregated across all versions of this paper published by eLife.
Download links
A two-part list of links to download the article, or parts of the article, in various formats.
Downloads (link to download the article as PDF)
Open citations (links to open the citations from this article in various online reference manager services)
Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)
Development and evaluation of a machine learning-based in-hospital COvid-19 disease outcome predictor (CODOP): a multicontinental retrospective study
eLife 11:e75985.
https://doi.org/10.7554/eLife.75985
Further reading
-
- Epidemiology and Global Health
- Medicine
- Microbiology and Infectious Disease
eLife has published the following articles on SARS-CoV-2 and COVID-19.
-
- Cell Biology
- Computational and Systems Biology
Circular RNAs represent a class of endogenous RNAs that regulate gene expression and influence cell biological decisions with implications for the pathogenesis of several diseases. Here, we disclose a novel gene-regulatory role of circHIPK3 by combining analyses of large genomics datasets and mechanistic cell biological follow-up experiments. Using time-course depletion of circHIPK3 and specific candidate RNA-binding proteins, we identify several perturbed genes by RNA sequencing analyses. Expression-coupled motif analyses identify an 11-mer motif within circHIPK3, which also becomes enriched in genes that are downregulated upon circHIPK3 depletion. By mining eCLIP datasets and combined with RNA immunoprecipitation assays, we demonstrate that the 11-mer motif constitutes a strong binding site for IGF2BP2 in bladder cancer cell lines. Our results suggest that circHIPK3 can sequester IGF2BP2 as a competing endogenous RNA (ceRNA), leading to target mRNA stabilization. As an example of a circHIPK3-regulated gene, we focus on the STAT3 mRNA as a specific substrate of IGF2BP2 and validate that manipulation of circHIPK3 regulates IGF2BP2-STAT3 mRNA binding and, thereby, STAT3 mRNA levels. Surprisingly, absolute copy number quantifications demonstrate that IGF2BP2 outnumbers circHIPK3 by orders of magnitude, which is inconsistent with a simple 1:1 ceRNA hypothesis. Instead, we show that circHIPK3 can nucleate multiple copies of IGF2BP2, potentially via phase separation, to produce IGF2BP2 condensates. Our results support a model where a few cellular circHIPK3 molecules can induce IGF2BP2 condensation, thereby regulating key factors for cell proliferation.
