The Epidemiology of Severe Acute Respiratory Syndrome in the 2003 Hong Kong Epidemic: An Analysis of All 1755 Patients

Context Few comprehensive studies describe the 2003 outbreak of severe acute respiratory syndrome (SARS). Contribution This epidemiologic analysis of 1755 cases from Hong Kong found that most cases clustered in hospitals and residential buildings. Close human contact and spread by a sewage system probably explain the clustering. The outbreak lasted about 3 months. The estimated mean incubation period was 4.6 days, and the case-fatality ratio was 17%. Factors associated with increased risk for death included older age and male sex. Implications The observed patterns suggested that SARS had low transmissibility, except in settings of intimate contact or clinically significant environmental contamination. Severe acute respiratory syndrome (SARS) was the first newly emergent communicable disease epidemic of the 21st century. During the first epidemic of this new pathogen, 29 countries were affected. The first human case was identified in Guangdong, China, on 16 November 2002 (1), and the last known case with a symptom onset date of 5 July 2003 was identified in Taiwan. The epidemic reportedly infected 8098 individuals, 774 of whom died (2). Hong Kong bore a large proportion of this morbidity and mortality burden: 1755 cases and 302 deaths occurred from 15 February to 31 May 2003. Hong Kong also provided the link between the cases in China and those in other parts of the world. The resurgence of SARS is distinctly possible given its uncertain origins and the likely existence of an animal reservoir, the palm civet cat (3). Since the end of the first major epidemic in July 2003, 4 new cases were reported from Guangdong province in China in late 2003 and early 2004. An account of the epidemiology of SARS in Hong Kong was undertaken during the outbreak (4) to inform public health policymaking. The data set has since been updated by using information of all 1755 reported cases, allowing for the relaxation of parametric assumptions, necessary in the mid-epidemic analysis, in the analysis of the times from symptoms to admission, admission to death, and admission to discharge. Furthermore, complete case data of the closed cohort allow analysis of predictors of SARS-related death by using logistic regression. We present an epidemiologic analysis of the SARS outbreak in Hong Kong on the basis of all reported cases and deaths classified according to prevailing World Health Organization (WHO) guidelines. In addition, laboratory verification by reverse transcriptase polymerase chain reaction (RT-PCR) test or SARS coronavirus antibody serologic test was obtained for 83.6% of cases. On the basis of the complete data set, we present the following analyses: a detailed description of the temporal and spatial evolution of the epidemic; the estimates of key epidemiologic distributions and their stability over the course of the epidemic; and the characteristics of those who contracted the disease, including factors associated with the likelihood of death from SARS coronavirus infection. Methods Sources of Data We analyzed an integrated database (SARSID), derived from the Hong Kong Hospital Authority eSARS system and the Hong Kong Department of Health's master list, which contained details on all patients reported to have SARS who were hospitalized in Hong Kong throughout the epidemic. The eSARs system is a secure, Web-based data repository that contains mostly real-time clinical data entered on the SARS patient wards. Trained nurses retrospectively collected and confirmed some data fields by a detailed chart review according to a standardized protocol. The Hong Kong Department of Health's master list mostly consisted of information from the questionnaires of case and case contact data. We administered the questionnaires (containing case and case contact information), mostly through telephone interviews, to all patients with SARS confirmed by the Hong Kong Department of Health; in most cases, the questionnaire was administered within 3 days (up to a maximum of 1 week) of initial presentation with SARS. For patients who could not be contacted or who were too ill to be interviewed, we obtained proxy reporting from an immediate family member who was most familiar with the medical and contact history of the patient before infection. Four regional field offices initially administered these questionnaires, and a central interviewing team of nurses later recorded symptoms at presentation to the hospital and identified contacts and events of probable significance to transmission. We collected data on case and contact information for all 1755 patients with SARS, although we did not complete all data elements for all cases. The Appendix provides detailed clinical case definitions for SARS throughout the 2003 epidemic. We based laboratory confirmation of SARS on laboratory techniques that were consistent with the WHO case definition for laboratory confirmed SARS (5): 1) RT-PCR for SARS coronavirus and 2) serologic testing for IgG antibodies against SARS coronavirus. A patient was considered to have laboratory-confirmed SARS if there was a positive RT-PCR result from 2 or more clinical specimens, either from different sites or tested in different laboratories, obtained from patients before or after death or if there was seroconversion by enzyme-linked immunosorbent assay (ELISA), indirect fluorescent antibody test, or neutralization assay. Although these tests were mostly available during the outbreak, not all patients were tested for any or all of them for various reasons, including nonuniform testing protocol (especially in the earlier part of the outbreak), lack of samples due to early case fatality without autopsy examination, and inadequate and missing specimens. Test variables, such as sensitivity and specificity, were unknown because there were no gold standard laboratory or clinicopathologic definitions for the diagnosis of SARS, a new and emerging disease, for comparing diagnostic test performance. The IgG antibodies against SARS coronavirus found on serologic testing seemed to be the best method for confirming SARS in largely seronegative populations (6), where the reported seropositivity rate reached 93% to 99% in Hong Kong (7, 8) and 96.2% in Toronto (9). We collected paired serologic specimens at least 21 to 28 days apart, although anecdotal reports of longitudinal follow-up of patients with SARS in Hong Kong estimate that seroconversion can occur as long as 6 months after acute illness. Tang and colleagues (9) found that the sensitivity of 1 first-generation RT-PCR was 54.1% in their Toronto SARS case series, assuming that all clinically classified patients truly had SARS. These findings were broadly similar to those reported by Peiris and colleagues (8) for the outbreak at Amoy Gardens housing estate in Hong Kong during late March and early April. In both the Hong Kong and Toronto epidemics (8, 9), the peak rate of RT-PCR positivity occurred 9 to 11 days after first symptoms presented, and gastrointestinal specimens gave higher yields than respiratory samples. Poon and colleagues (10) later produced a second-generation RT-PCR assay capable of detecting SARS coronavirus in up to 88% of respiratory tract samples obtained within the first 3 days after illness onset in confirmed SARS cases in the Hong Kong outbreak. This test kit was adopted in the latter part of the Hong Kong epidemic. All SARS coronavirus specimen testing in Hong Kong was performed in 3 designated laboratories (Chinese University of Hong Kong, University of Hong Kong, and Hong Kong Department of Health) where rigorous quality control procedures were established. The World Health Oranization and members of the WHO SARS Reference and Verification Laboratory Network certified all 3 facilities as reference laboratories. Statistical Analysis We constructed the epidemic time series on the basis of all 1755 local cases by date of symptom onset and infection cluster. We classified infection clusters by probable transmission setting (institutional vs. community spread), location (for example, housing estates), occupation (for example, health care workers in public and private sectors), and workplace (for example, hospitals and other buildings). We compared the age and sex distributions of SARS cases with general population estimates derived from the 2001 population census conducted by the Hong Kong Government Census and Statistics Department. To illustrate the geospatial pattern of infection and disease spread, we used a geographic information system (ArcGIS, Environmental Systems Research Institute, Redlands, California) to construct a map of infection clusters in different districts of Hong Kong. We plotted empirical distributions for times from onset to admission, onset to death, and onset to discharge after recovery (from acute care since many patients were transferred to convalescent facilities), and we calculated the mean and variance of these distributions (Appendix). Infection events cannot be observed, but data on patients with short and well-defined periods of 1 exposure to known SARS cases can be used to estimate the distribution of the time from infection to onset of symptoms (the incubation period) by using methods for interval-censored data (4, 11). The database contained 81 patients who had 1 exposure to a confirmed SARS case over a limited time scale (<15 days) with recorded start and end dates, who did not travel, and who were not hospitalized before the onset of symptoms. We estimated the distribution by using maximum likelihood methods, assuming a distribution. We used logistic regression to identify factors significantly associated with case fatality due to SARS. The following variables were tested in the model: age; sex; occupation (health care worker vs. others); symptoms on presentation (typical [patients with SARS whose symptom score as determined by the prediction rule was above the threshold designated as high risk for SARS] vs. atypical [patients wit