HIV Working Group Seminar

Tanayott (Tony) Thaweethai
Associate Director for Biostatistics Research and Engagement, Massachusetts General Hospital Biostatistics

Development of a symptom-based definition of long COVID using negative-unlabeled data
Abstract: While most people infected with COVID-19 recover after the acute phase of infection, some people continue to experience persistent symptoms months and even years after infection. These symptoms, also known as post-acute sequelae of SARS- CoV-2 infection (PASC), or long COVID, present a significant public health concern. However, long COVID remains poorly understood, as there is no generally agreed upon definition of PASC in the medical community. This has made it difficult to characterize population-level burden of the disease or study pathophysiological mechanisms. Because only those who are infected can develop PASC, we are in the world of negative-unlabeled data, where uninfected individuals are PASC-negative but infected individuals are a mixture of PASC-positive and PASC-negative. In this talk, I will present a novel statistical approach that we used to develop the first data-driven, symptom- based definition of long COVID from RECOVER, the largest observational cohort study of long COVID in adults and children.