Temple University Team Works on Developing Algorithms that Can Predict the COVID-19 based Cytokine Storm

Temple University Team Works on Developing Algorithms that Can Predict the COVID-19 based Cytokine Storm

Temple University’s Lewis Katz School of Medicine recently published the results of a study that risk stratifies COVID-19 patients—identifying those individuals most at risk of experiencing life-threatening reactions such as the cytokine storm. The Philadelphia-based team is sharing their data supporting other study efforts to develop more granular evidence into risks associated with COVID-19. Among other findings, about 20% of those hospitalized due to COVID-19 can fall to more severe disease progression, such as the cytokine storm, with a growing risk of mortality. But adequately and accurately predicting which patients are at most risk is difficult to accomplish.

Led by Doctor Roberto Carricchio, the study is a good start. He emphasized that additional research is required, but the team has contributed to a growing baseline of knowledge.

A Deadly Situation

A cytokine storm associated with COVID-19 leads to horrific and deadly situations. Loved ones can pass on their own with no family nearby. It’s as if in the case in the lungs of an individual drowning for days if not weeks. This event, an immune overreaction, leads to massive inflammation and organs ceasing to work. Dr. Carriccio reports, “It can be the kidney, along with the lungs. It can be the heart, along with the lungs or the heart and liver, so its multi-system organ failure.”

Moreover, once this process commences, it becomes increasingly difficult to turn it around. This led to the doctor and team studying hundreds of COVID-19 patient lab work and scans to further develop criteria used for the development of predictive algorithms.

The Study

The team analyzed 513 hospitalized patients shown to be SARS-CoV-2 positive and ground-glass opacity by chest high-resolution CT. The group studied the patients’ laboratory results during the first seven days of hospitalization in a bid to establish predictive formulas, the implemented logistic regression, and principal component analysis to probe for predictive criteria, also developing a ‘genetic algorithm’ to come up with parameters for each laboratory result. The data was validated via a second patient cohort involving 258 subjects.


The team uncovered that the actual criteria were developed to predict select conditions, such as macrophage activation syndrome and hemophagocytic lymphohistiocytosis, and the HScore didn’t identify the COVID-19 cytokine storm. The team was able to develop new predictive criteria comprising three clusters of laboratory results involving 1) inflammation, 2) cell death and tissue damage, and 3) prerenal electrolyte imbalance. Dr. Carricchio and the team did find that the criteria they established improved the identification of patients with longer hospitalization and a higher probability of mortality. The results point to the potential correlation of COVID-19 and cytokine storm to hyperinflammation and tissue damage.


The study team suggests establishing novel early predictive criteria to better identify the cytokine storm potential in hospitalized patients infected with SARS-CoV-2, the virus behind COVID-19. These criteria would be used in the clinic to improve the provider’s ability to prescribe an earlier intervention in a bid to inhibit the hyperimmune response and reduce mortality.


For instance, a study thereafter revealed that an algorithm could potentially anticipate cytokine storm in up to 85% of cases. Hence in this situation, physicians can move swiftly and more intelligently prescribing steroids or some investigational medications still in clinical trials.

Lead Research/Investigator

Roberto Carricchio, MD, FACR Chief, Section of Rheumatology, Lewis Katz School of Medicine at Temple University, Director, Temple Lupus Program, Temple University Hospital, Professor, Medicine, Lewis Katz School of Medicine at Temple University, Professor, Microbiology and Immunology.