Objective Thymus and activation-regulated chemokine (TARC) can predict severe disease in patients with coronavirus disease 2019 (COVID-19). However, no reports have addressed the predictive value of TARC with the widespread use of vaccines and medications for COVID-19 during the Omicron variant period of the pandemic. Methods This single-center prospective cohort study enrolled COVID-19 patients admitted to our institution between December 1, 2021, and August 15, 2022. Patients with respiratory failure due to diseases other than COVID-19 were also excluded. We measured the serum TARC levels of patients at admission. Results We enrolled 157 patients, with 89 in the severe group and 68 in the non-severe group. The severe group was more likely than the non-severe group to include older patients, those with no or one dose of vaccine, and those with interstitial lung disease. The cutoff level of TARC derived from a receiver operator characteristic curve analysis to predict severe disease was 174.0 pg/mL. The sensitivity, specificity, positive predictive value, and negative predictive value were 72.1%, 69.7%, 64.5%, and 76.5%, respectively. The area under the curve was 0.722 (95% confidence interval: 0.635-0.809). A multivariate analysis showed that 2 vaccination doses were associated with non-severe disease, and TARC ≤174 pg/mL was associated with severe disease. Conclusion TARC was a predictive factor for severe disease, but its cutoff value was higher and its predictive accuracy lower than those in previous reports. We surmised that during the Omicron variant period of the pandemic, the widespread use of vaccines and medications for COVID-19 decreased the predictive accuracy of TARC.