AI-Driven Blood Test to Revolutionize Diagnosis of Long COVID in Children

Image: The AI blood test offers a new avenue for diagnosing long COVID in the pediatric population (Photo courtesy of 123RF)

Long COVID affects about 0.5% of pediatric patients exposed to SARS-CoV-2. Also known as post-COVID or post-acute sequelae of SARS-CoV-2, this condition is characterized by persistent symptoms that appear after the viral infection and last for at least 8-12 weeks, significantly impacting daily life. Long COVID has been observed across all age groups, with children over the age of 10 being particularly affected in terms of severity. While proteomic studies have highlighted a pro-inflammatory signature and thrombo-inflammation in adults with long COVID, similar findings had not been documented in the pediatric population. Now, a new method can accurately diagnose long COVID in children using a blood sample and artificial intelligence (AI).

In a study led by researchers at Università Cattolica del Sacro Cuore (UCSC, Milan, Italy) and associated institutions, an AI tool was found to effectively detect the molecular signature of long COVID in plasma samples from pediatric patients, with an accuracy rate of 93%. For their research, the team performed a detailed protein analysis on blood plasma from children under 19 years of age, comparing those with long COVID to control groups including children with acute COVID-19, MIS-C, and healthy controls. Children were classified as having long COVID if they experienced persistent symptoms for at least 8 weeks after the initial infection, which negatively impacted their daily activities and could not be attributed to other causes. The study included 112 children, with 34 meeting the clinical criteria for long COVID, 32 with acute SARS-CoV-2 infection, 27 with MIS-C, and 19 healthy controls.

The study, published in the journal Pediatric Research, showed that children with long COVID showed higher expression of a specific set of pro-inflammatory and pro-angiogenic chemokines (CXCL11, CXCL1, CXCL5, CXCL6, CXCL8, TNFSF11, OSM, STAMBP1a) compared to controls. Using a machine learning model based on this proteomic profile, researchers identified long COVID with an accuracy of 93%, a specificity of 86%, and a sensitivity of 97%. The findings indicate that pediatric long COVID is characterized by an increased blood protein signature associated with ongoing general and endothelial inflammation, similar to what has been observed in adults. This study is the first to document the pro-inflammatory profile in children with long COVID, and the results may contribute to the development of future diagnostic tests.