CODOP

(COvid-19 Disease Outcome Predictor)

119

institutions that have already used the calculator

> 12070

number of patients whose data have been analyzed

WHAT IS CODOP?

The emergence of more contagious SARS-CoV-2 virus variants, breakthrough infections, waning immunity, and differential access to vaccines account for the worst yet numbers of hospitalisations and deaths during the COVID-19 pandemic, particularly in resource-limited countries. There is an urgent need for clinically valuable, generalizable and parsimonious triage tools.

CODOP is the result of a multicontinental altruistic collaboration between the team of Dr. David Gómez Varela (Max Planck Institute of Experimental Medicine), Riku Klén (Turku University), the Spanish Society of Internal Medicine (SEMI), The Argentinian Society of Medicine (SAM), and the International Forum of Internal Medicine (FIMI), which the only goal of building a simple, accurate, and clinically useful machine-learning based tool able to predict the risk of mortality in hospitalized COVID-19 patients (manuscript in preparation).

WHY CODOP?

The unique characteristics of CODOP (in comparison to other published predictors) that makes it a clinically useful triage tool are:

    • CODOP has been trained and tested with the most extensive and variable cohorts to date: more than 40,000 samples from more than 21,000 patients hospitalized in more than 110 hospitals of two geographically different locations (Spain and New York) and during three different pandemic waves.
    • CODOP uses only 12 parameters commonly measured at hospital admission (for details download the “Example Table of patient data” located in the “Start Using CODOP” section, below).
    • In order to assist during the changing dynamics of the pandemic, CODOP is offered in two subtypes suited for under- and over-triage situations (see details in the “Start Using CODOP” section).
    • CODOP detected the vast majority of patients in the training cohort that died or survived in the hospital up to nine days before clinical resolution (AUROC: 0.90-0.96; 95% CI 0.88-0.97) and it also enabled dynamic risk stratification during hospitalization.
    • In an independent validation performed with patient data from 3 different Latin American countries, CODOP preserved its performance, despite the many differences of this dataset in comparison to the training and test cohorts.
    • CODOP fulfils all TRIPOD, PROBLAST and MINIMAR standards.
    • CODOP is offered as an open WebCalculator (see the “Start Using CODOP” section, below) that does not store any patient data, that includes an easy uploading method and an imputation method for missing values, and that it has a very fast prediction process (< 2 seconds). This tool fulfils all data protection agreements and the privacy policies of the EU and the USA.

DO YOU WISH TO COLLABORATE TO MAKE CODOP BETTER?

CODOP aims to improve patient care during the COVID-19 pandemic, particularly in resource-limited countries. Participations of more institutions from World areas not represented in this study (e.g. Asia, Africa) will increase its reproducibility, clinically utility, and will enable subgroup-specific predictions (e.g. based on underlying comorbidities or ethnic background). Therefore, we would like to kindly ask you to join us using the patient data available in your institution’s database.

There are two ways you can participate:

  1. Prospective validation: upload a data table, using as scaffold the “Example Table of patient data”, with the values of the 12 parameters for each of your patients into the web calculator. Following, send us an email (gomez@em.mpg.de) with the Excel table that CODOP gives you (accessible by clicking in the “Download Predictions” tag) BUT in which we ask you to add another column with the real patient outcome. 
  2. CODOP 2.0: fill as many patient parameters as possible from the ones included in this table (link). Following, send us the table  to gomez@em.mpg.de. We will use these data for re-training CODOP towards a simpler and better model.

We hope that you consider participating in this project, which is helping many clinicians fighting for the least privileged ones on this pandemic.

Thank you very much in advance.


START USING CODOP

Depending on the availability of resources in your hospital during the COVID-19 pandemic, you can choose between two CODOP subtypes:

  • CODOP-Ovt is suitable for scenarios in which hospital resources are not the limiting factor and over-triage is possible. It maximizes the identification of patients in high risk of in-hospital mortality: 80-98% patients in all training and independent test cohorts that died were correctly predicted up to nine days before death, albeit predicting a 30-40% patients number of patients as high risk that finally survived.
  •  CODOP-Unt might be preferred when hospital resources are limited and under-triage needs to be considered. It maximizes the identification of patients in low risk of in-hospital mortality: 85-95% patients in all training and independent test cohorts that survived were correctly predicted up to nine days before hospital discharged, albeit predicting 40-50% of patients as low risk that finally will died in all training and test cohorts.

CODOP returns a value of 0 for a High probability of Survival and a value of 1 for a High probability of Death