Anti-SARS IgG can be detected 1–2 weeks after the onset of symptoms. Almost all SARS patients in the convalescent stage had anti-SARS IgG in their serum samples (11–14). In our study, all healthcare workers with SARS were in the convalescent stage, and SARS-CoV infected most while they were caring for the same index patient who was subsequently identified as a superspreader (15). The finding of a 100% seroprevalence of SARS IgG among 17 SARS-infected healthcare workers in the Third Affiliated Hospital was identical to the results by Li et al. (12), who tested SARS IgG at different stages among the same group of SARS-infected healthcare workers by using the same ELISA. By contrast, 63 (86.3%) of 73 healthcare workers with SARS in the Second Affiliated Hospital were seropositive for SARS IgG. Some of these healthcare workers might have been misdiagnosed, as the clinical diagnosis of SARS was not specific (16). Even allowing for this possibility, the overall high seropositivity rate of 88.9% among SARS patients is similar to findings by Wang et al. in Beijing, who used the same test (11); Chow et al. in Singapore, who used a different EIA (17); and Chan et al. in Hong Kong, who used an immunofluorescence assay (18). All of these studies indicate that serum IgG antibodies to SARS-CoV at the convalescent stage of the illness can be useful in confirming the disease.

The low seroprevalence o SARS IgG (0.3%–2.8%) in healthy healthcare workers with different levels of exposure to SARS patients is similar to that reported by Wang et al. (11). However, a similar study by Chow et al. in Singapore did not find any serologic evidence of subclinical infection among a population with a high likelihood of exposure to the virus. Our ELISA was 99% specific (11). This specificity could have produced a few false-positive results, which accounts for a low seropositive rate of 0.4% (3/709) among healthcare workers in the reference group, who had no exposure to SARS in their hospitals. One healthcare worker in Zhuhai, where no SARS occurred, was seropositive, which could be a false-positive result. However, we could not exclude the possibility of inapparent infection among healthcare workers in the 4 hospitals in Guangzhou. Another possibility is cross-reaction with other human coronaviruses. A more specific test, such as the indirect immunofluorescence test, should clarify this uncertainty (18).

The low seroprevalence of SARS IgG, at 1.4% (15/1,057) among apparently healthy frontline healthcare workers in all 3 SARS hospitals, suggests that inapparent infection is relatively uncommon. We did not, however, ascertain whether the healthcare workers with a positive antibody test result were carriers of SARS-CoV. Overall, the low seropositivity among healthy healthcare workers suggests that the level of immunity to SARS in the general population in Guangzhou was too low to constitute an effective immune barrier against the spread of SARS. Should the disease recur there, every effort should be made to protect healthcare workers and the general public from being infected by SARS patients.

The First Affiliated Hospital only admitted SARS patients after outbreaks had occurred among healthcare workers in the Second and Third Affiliated Hospitals. After these outbreaks, a series of protective measures were adopted in all 3 hospitals. Sufficient preparation, such as personal protection and designated SARS wards, is important to avert hospital outbreaks. The low seroprevalence of SARS IgG among healthcare workers working in the First Affiliated Hospital indicated the effectiveness of these measures. This finding is consistent with the study by Chow et al. (17).

The seroprevalence rates were significantly different among the healthcare workers who cared for SARS patients when classified by their age, sex, educational level, hospital, department, and job title. These differences could be due to the probability of exposure to the SARS index case. On January 30, 2003, the index patient was admitted to the Department of Respiratory Diseases of the Second Affiliated Hospital. On February 1, he was transferred to the Third Affiliated Hospital because of worsening dyspnea. During his stay in these 2 hospitals, where protective measures were lacking, he directly and indirectly infected 90 healthcare workers and 22 healthcare workers in the Second and Third Affiliated Hospitals, respectively. This finding accounts for the much higher seroprevalence of SARS IgG among healthcare workers in the departments of respiratory diseases and infectious diseases. Healthcare workers in SARS wards and fever clinics were fully equipped with personal protective measures (caps, gowns, multilayered cotton face masks, eye shields, gloves, and foot covers), which might explain their much lower seroprevalence. None of our laboratory healthcare workers, who performed serologic tests but not live viral tests, were seropositive, which suggests that the probability of SARS infection by handling serum samples of SARS patients was low. Of all occupations, healthcare attendants had the highest seropositive rate, which might be related to their general lower level of education and a lack of training in infection control measures. Future efforts to improve SARS containment should also address this problem among nonprofessional staff.

This study has several limitations. Even though the response rate among healthcare workers who cared for SARS patients was high (82.3%), some selection bias is inevitable. Moreover, the study was limited to 5 university hospitals, so we caution against the extrapolation of our findings to healthcare workers in other hospitals and to the general population. We have not tested serum samples from our healthcare workers against other pathogens, e.g., Mycoplasma pneumoniae and influenza virus, which limits our ability to exclude the nonspecific and atypical pneumonia caused by these agents (16).

In conclusion, this study shows that a high proportion of healthcare workers who have contracted SARS have IgG against SARS-CoV in their serum samples after they have fully recovered. Inapparent infection with SARS is uncommon. The low seropositivity against SARS among healthcare workers who have not been exposed to SARS patients suggests a lack of immunity in this group and in the general population, where the number of SARS cases is comparatively small.