International Electronic Health Record-Derived COVID-19 Clinical Course Profile: The 4CE Consortium

Gabriel A Brat; Griffin M Weber; Nils Gehlenborg; Paul Avillach; Nathan P Palmer; Luca Chiovato; James Cimino; Lemuel R Waitman; Gilbert S Omenn; Alberto Malovini; Jason H Moore; Brett K Beaulieu-Jones; Valentina Tibollo; Shawn N Murphy; Sehi L'Yi; Mark S Keller; Riccardo Bellazzi; David A Hanauer; Arnaud Serret-Larmande; Alba Gutierrez-Sacristan; John J Holmes; Douglas S Bell; Kenneth D Mandl; Robert W Follett; Jeffrey G Klann; Douglas A Murad; Luigia Scudeller; Mauro Bucalo; Katie Kirchoff; Jean Craig; Jihad Obeid; Vianney Jouhet; Romain Griffier; Sebastien Cossin; Bertrand Moal; Lav P Patel; Antonio Bellasi; Hans U Prokosch; Detlef Kraska; Piotr Sliz; Amelia LM Tan; Kee Yuan Ngiam; Alberto Zambelli; Danielle L Mowery; Emily Schiver; Batsal Devkota; Robert L Bradford; Mohamad Daniar; Christel Daniel; Vincent Benoit; Romain Bey; Nicolas Paris; Patricia Serre; Nina Orlova; Julien Dubiel; Martin Hilka; Anne Sophie Jannot; Stephane Breant; Judith Leblanc; Nicolas Griffon; Anita Burgun; Melodie Bernaux; Arnaud Sandrin; Elisa Salamanca; Thomas Ganslandt; Tobias Gradinger; Julien Champ; Martin Boeker; Patricia Martel; Alexandre Gramfort; Olivier Grisel; Damien Leprovost; Thomas Moreau; Gael Varoquaux; Jill Jen Vie; Demian Wassermann; Arthur Mensch; Charlotte Caucheteux; Christian Haverkamp; Guillaume Lemaitre; Christian Haverkamp; Tianxi Cai; Isaac S Kohane

There are a multitude of limitations to this study, not least of which is that it is observational and subject to a variety of biases. Perhaps the most severe is that study data is limited to those patients who were seen at or admitted to hospitals, due to severity of illness or other possibly biasing characteristics. Limitations also include heavy right censoring where patient absence can be due to death or discharge, variations in ICD annotations for conditions existing prior to the COVID-19-related admission, delays in updating billing codes or in uploading EHR data to the local analytic data repository. Furthermore, potentially confounding interactions between comorbidities, chronic diseases and their treatments and lifestyle or exposures were not taken into consideration. Again, because of these limitations we were careful to avoid making more than basic and descriptive conclusions. Over the coming weeks, we will work to quantify these biases and adjust for them, if we can. This will include adding data types as well as disaggregating the data to the patient level if and when permitted by IRBs. For the present, with the current limited knowledge of the clinical course of patients suffering from COVID-19, these results add to this small knowledge base. Our paper strikingly shows the power of harmonized data extraction from EHRs to rapidly study pandemics like COVID-19.