Using Logistic Regression to Predict Long COVID Conditions in Chronic Patients

Abstract

Chronic diseases pose significant challenges to patients and healthcare systems, and the COVID-19 pandemic has further deteriorated that situation. This paper presents a method for predicting selected long COVID conditions in chronic and multimorbidity patients. It produces a logistic regression model for each long COVID condition by examining electronic medical records (EMRs) of COVID-19 patients and taking their chronic conditions as predictors. The models were developed and tested using the Jumpstart EMR database, provided in the COVID-19 Research Environment of Hopkins University, containing about 250,000 EMRs of the outpatient and ambulatory COVID-19 patients across the US. They are illustrated by predictions of 20 prevalent acute and chronic long-COVID conditions in patients diagnosed with frequent pre-COVID chronic diseases. These models can aid in investigating long COVID impacts on various chronic patients, finding their underlying pathophysiology, and establishing guidelines for their treatment and prevention.