The identification of asymptomatic, non-severe presymptomatic, and severe presymptomatic coronavirus disease (COVID-19) in patients may help optimize risk-stratified clinical management and improve prognosis. This single-center case series from Wuhan Huoshenshan Hospital, China, included 2,980 patients with COVID-19 who were hospitalized between February 4, 2020 and April 10, 2020. Patients were diagnosed as asymptomatic (n=39), presymptomatic (n=34), and symptomatic (n=2,907) upon admission. This study provided an overview of asymptomatic, presymptomatic, and symptomatic COVID-19 patients, including detection, demographics, clinical characteristics, and outcomes. Upon admission, there was no significant difference in clinical symptoms and CT image between asymptomatic and presymptomatic patients for diagnosis reference. The mean area under the receiver operating characteristic curve (AUC) of the differential diagnosis model to discriminate presymptomatic patients from asymptomatic patients was 0.89 (95% CI, 0.81-0.98). The severe and non-severe presymptomatic patients can be further stratified (AUC = 0.82). The two-step risk-stratification model based on 10 laboratory indicators on admission can facilitate early identification of asymptomatic patients with COVID-19 and predict illness severity for appropriate clinical management. Moreover, single-cell data analyses revealed that the CD8+T cell exhaustion contributed to the progression of COVID-19.