BloodCOVID Algorithm

Last update: Nov 17, 2020.

Routine Blood Tests for Diagnosis of Suspected COVID-19

 
 

Conference results presentation

 
 

Presentation of the Decision Trees at the conference entitled “Diagnostic algorithms and blood biomarkers for COVID-19” by José Diego Santotoribio, MD, PhD, organized by the Institute of Biomedicine of Seville (IBIS) and Virgen del Rocío University Hospital (HUVR), Wednesday, November 18th at 8:30 a.m. (GMT/UTC+1). View the Press Release.

 
 

Santotoribio MD PhD Interview

 
 
 
 

Objectives

 
 
The outbreak of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) to date, the epidemic has gradually spread to 190 countries worldwide with more than 45 million infected people and 1,180,000 deaths.

Amplification of viral RNA by RT-PCR serves as the gold standard for confirmation of infection, yet it needs a long turnaround time (3–4 h to generate results) and shows false-negative rates as large as 15%–20%. In addition, massive testing policies together with the need of certified laboratories, expensive equipment and trained personnel, delay delivery of results up to 7-10 days.

However, since mid-September, a new COVID-19 antigen test with faster results is growing in popularity. The antigen test uses the same technology as the rapid flu or strep test. The advantage of the antigen test is being able to rapidly diagnose someone who is COVID-19 positive and advise them on isolation precautions so that they are not in the grocery store behind you. Nevertheless, the CDC still calls the molecular test the “gold standard” because the antigen has seen more false negatives. In this way, symptomatic people who test negative are advised to get a new test.

In summary, while most Molecular or PCR tests can take days, a rapid antigen test takes minutes.

Despite all this, neither PCRs nor rapid antigen tests have been able to prevent the second wave that is hitting most countries right now. On the one hand, PCRs are still awfully expensive, and rapid antigen tests (much cheaper than PCRs) cannot be produced in enough quantity that current world demand requires.

Thus, there is a need for alternative, less expensive and more accessible tests.
 
 

Methods

 
 
This study is based on previous one Evaluation of Routine Blood Tests for Diagnosis of Suspected Coronavirus Disease 2019 performed by Jose D. Santotoribio (Department of Clinical Biochemistry, Virgen del Rocío University Hospital, Seville, Spain), David Nuñez-Jurado (Department of Clinical Biochemistry, Virgen del Rocío University Hospital, Seville, Spain), Esperanza Lepe-Balsalobre (Department of Clinical Biochemistry, Virgen del Rocío University Hospital, Seville, Spain, and JL Castaño Foundation, Spanish Society of Laboratory Medicine, Barcelona, Spain), which it sought to evaluate the routine blood tests for diagnosis of COVID-19, determining the diagnostic accuracy of blood biomarkers to differentiate between patients with and without COVID-19.

This study has four main objectives: the first one, to analyze the database of 203 patients that Dr. Santotoribio used in his previous publication to see if it was possible to improve the diagnostic performance achieved by said doctor ―that is, 91% Sensitivity and 47% Specificity by using MedCalc 13.0 (MedCalc Software, Ostend, Belgium)―, by using our own Artificial Intelligence (AI) and Deep Learning (DL) algorithms; the second one, to analyze whether the results can be represented by a Decision Tree (DT) ―as the best tool to share results to be widely and easily used in clinical practice―, without losing the diagnostic performance achieved previously; the third one, to analyze whether with this database another Decision Tree could be obtained that would also allow discriminating those patients with a worse prognosis of the disease to prioritize their treatment, in order to reduce the mortality rate; and the fouth one, to perform a Health Economics Outcome Research (HEOR) study ―mainly a cost/opportunity one―, to analyze the benefits of using a simple routine blood test as a mass screening method and leaving both PCRs and rapid antigen tests to confirm those resulted positive with our algorithm.
 
 

Results

 
 
Statistically enough significant differences were observed for Ferritin, WBC, GPT and GOT (for males), and Ferritin, WBC, LDH and CRP (for females). Both Decision Trees for each gender allowed the identification of 100% of either COVID-19 positive or negative patients on the basis of routine blood test results, that is, 100.00% Sensitivity and 100.00% Specificity for whole database.
 
 

Conclusions

 
 
According previous results, by combining appropriate cutting-edge technologies, such as own Artificial Intelligence (AI) and Deep Learning (DL) algorithms, together with Decision Tree (DT) models based on certain hematological parameters, could help in identifying COVID-19 positive patients in those countries which suffer from a large shortage of RT-PCR reagents, specialized laboratory and/or shortage of rapid antigen tests, with a similar accuracy to both RT-PCR as well as rapid antigen tests.

That said, the study presents three main limitations: the first, and more obvious one, regards the relatively low number of cases considered (n = 203); the second one, is related to the potential vendor dependence ―plasma cell count were determined in a Sysmex XN-2000 analyzer (Sysmex, Kobe, Japan), plasma concentration of D-dimer were determined in a Sysmex CS-5100 analyzer (Sysmex, Kobe, Japan), and biochemical parameters were determined in an Hitachi Cobas c 702 modular analyzer (Roche Diagnostics, Rotkreuz, Switzerland)―, so the use of other vendors could also affect accuracy; the third one may be less obvious, as it regards the reliability of the ground truth itself. Although this was built by means of the current gold standard for COVID-19 detection, i.e., the RT-PCR test, recent studies observed that the accuracy of this test may be highly affected by problems like inadequate procedures for collection, handling, transport and storage of the swabs, sample contamination, and presence of interfering substances, among the others. As a result, some recent studies have reported up to 20% false-negative results for the RT-PCR test, and a recent systematic review reported an average sensitivity of 92% and cautioned that “up to 29% of patients could have an initial RT-PCR false-negative result”.

Thus, contrary to common belief, the accuracy of this test could be less than optimal ―mainly in other medical centers that use vendors other than those used by the Virgen del Rocío University Hospital (Seville, Spain)―. However, besides being a limitation, this is also a further motivation to pursue alternative ways to perform the diagnosis of SARS-CoV-2 infection, such as our method is.
 
 

Decision Tree Models

 
DISCLAIMER: BELOW DECISION TREES (DIAGRAMS) WERE CREATED FOR RESEARCH AND TESTING PURPOSES. MEDICAL DECISIONS MUST NOT BE BASED ON THE RESULTS OF THESE DIAGRAMS WHICH CANNOT BE CONSIDERED DIAGNOSTIC TOOLS. ALTHOUGH THESE DIAGRAMS HAVE BEEN TESTED THOROUGHLY, THE ACCURACY OF THE INFORMATION CANNOT BE GUARANTEED AND THE AUTHORS SHALL NOT BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY.
 

Decision Tree for COVID-19 Diagnosis (Male)

 

Diagnostic performance (n = 102): 100% Sensitivity and 100% Specificity.


 

Decision Tree for COVID-19 Diagnosis (Female)

 

Diagnostic performance (n = 101): 100% Sensitivity and 100% Specificity.
 

 

Decision Tree for COVID-19 Positive Prognosis (with D-Dimer)

 

Diagnostic performance (n = 203): 100% Sensitivity and 100% Specificity.
 

 

Decision Tree for COVID-19 Positive Prognosis (without D-Dimer)

 

Diagnostic performance (n = 203): 100% Sensitivity and 100% Specificity.
 

 

License

 
 
The PDF documents above (Decision Tree for Diagnosis of Suspected COVID-19 based on Routine Blood Tests for Male and Female), are published under an Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0) license.
 
 


 
 
This is a human-readable summary of (and not a substitute for) the license. Disclaimer.
 
 

You are free to:

 
 
Share — copy and redistribute the material in any medium or format (the licensor cannot revoke these freedoms as long as you follow the license terms).
 
 

Under the following terms:

 
 
Attribution — You must give appropriate credit, provide a link to the license, and indicate if changes were made. You may do so in any reasonable manner, but not in any way that suggests the licensor endorses you or your use.

NonCommercial — You may not use the material for commercial purposes.

NoDerivatives — If you remix, transform, or build upon the material, you may not distribute the modified material.

No additional restrictions — You may not apply legal terms or technological measures that legally restrict others from doing anything the license permits.
 
 

Acknowledgments