Body Temperature Trajectory Associated with Venous Thromboembolism in COVID-19 Patients

Abstract

Rationale: COVID-19 is associated with significant morbidity and presents unique challenges, including an increased risk of venous thromboembolism (VTE). In a single-center study early in the pandemic, we identified four distinct COVID-19 subphenotypes using longitudinal body temperature (i.e., temperature trajectory subphenotypes). Importantly, these subphenotypes had significant differences in hematological labs such as platelets and d-dimer, suggesting a relationship between temperature and coagulopathy. In this study, we aim to validate the temperature trajectory subphenotypes in a multi-center cohort of COVID-19 patients and evaluate whether temperature trajectory can identify patients at higher risk for VTEs. Methods: We included all patients hospitalized with laboratory-confirmed diagnosis of COVID-19 across four hospitals in the greater Atlanta area. For the trajectory analyses, we included patients' temperature measurements from the first 72 hours of hospitalization. We compared the temperature measurements from the study patients to each of the four trajectories from the published model to calculate the “trajectory distance” (i.e., the distance the patient is away from each trajectory). The patients were classified into the trajectory subphenotype from which they were the smallest distance away. We used ICD-10 codes at discharge to identify patients who had documented diagnoses of acute VTEs and evaluated the association between VTEs and trajectory subphenotype. Then, we used logistic regression to evaluate whether trajectory distance could predict VTE when controlling for demographics and ddimer levels. Results: The 2,107 hospitalized patients who met study criteria had a median age of 59 years (IQR 47-71 years), were 51% female, 65% Black, 21% White, and 10% Hispanic. The incidence of VTE was 12% and the inpatient mortality rate was 11.6%. By temperature trajectory subphenotype: 12% were Group 1, 31% Group 2, 48% Group 3, and 8.1% Group 4 (“hypothermic”). Temperature trajectory had significant association with mortality (p<0.001), with Groups 1 and 4 having the highest mortality rates (17 and 18%, respectively). Temperature trajectory subphenotype was significantly associated with VTE (p=0.004), with “hypothermic” patients having twice the incidence of other subphenotypes. On logistic regression, trajectory distance was significantly associated with VTEs even controlling for d-dimer (Figure). Conclusions: We validated our temperature trajectory subphenotypes in a multi-center cohort of hospitalized patients with COVID-19. We found that temperature trajectory could have utility in identifying patients at higher risk for VTEs who may require more aggressive anticoagulation. (Table Presented).