Background: Whether subclinical or atypical presentations of severe acute respiratory syndrome (SARS) occur and whether clinical judgment is accurate in detecting SARS are unknown.

Objectives: To describe the spectrum of SARS coronavirus infection in a large outbreak and to compare diagnoses based on clinical judgment with the SARS coronavirus test.

Design: Secondary analysis of prospectively collected clinical data and archived serum.

Setting: A SARS screening clinic of a university hospital in the New Territories of Hong Kong.

Patients: 1221 patients attending the clinic between 12 March 2003 and 12 May 2003.

Measurements: SARS coronavirus serology.

Results: 145 of 553 (26%) patients had serologic evidence of SARS coronavirus infection. Of 910 patients who were managed without hospitalization, only 6 had serologic evidence of SARS. Five of the six patients had normal chest radiographs, and four had symptoms such as myalgia, chills, coughing, and feeling feverish. With the SARS coronavirus serologic test as the gold standard, the clinical diagnosis of probable SARS at hospitalization had a sensitivity of 0.96 (95% CI, 0.91 to 0.98) and a specificity of 0.96 (CI, 0.92 to 0.97).

Limitations: Follow-up serologic samples were not obtained from almost half of the patients because they declined further testing. Some people living in the community who were infected but who had minor or no symptoms might not have visited the clinic.

Conclusions: There is little evidence of widespread subclinical or mild forms of SARS coronavirus infection. Clinical diagnoses during the outbreak were reasonable and resulted in appropriate triaging.