From March 13 to May 5, 2003, we identified all epidemiologically linked patients whose disease met the Centers for Disease Control and Prevention’s case definition of SARS issued on April 29, 2003 (2). Initial investigations included a complete blood count (with a differential count), serum biochemical measurements (including electrolytes, renal and liver function values, creatine kinase, and lactate dehydrogenase), and a chest x-ray. Since the cause of the virus was not known at the onset of the outbreak, routine microbiologic cultures of sputum, urine, and blood were done to rule out common bacterial causes of pneumonia. In addition, mycoplasma serology and urine Legionella antigen testing were carried out. When reverse transcriptase–polymerase chain reaction (RT-PCR) kits for coronavirus detection became available, later patients also provided samples for RT-PCR.

Probable SARS was diagnosed in 14 patients and healthcare workers at National University Hospital. The median age of the patients (five men and nine women) was 58 years (range 21–84). Detailed patient characteristics, including background, medical histories, symptoms, and signs, are shown in Tables 1 and 2.

On March 18, 2003, our emergency department began screening all febrile patients with respiratory complaints for possible SARS (2). All suspected SARS patients were admitted to a negative-pressure isolation room for monitoring. Probable SARS case-patients were transferred to Tan Tock Seng Hospital for further management to keep the hospital free of the SARS virus.

The varied clinical signs and symptoms of SARS have limited the success of such a triage system (4–8). We have observed that current diagnostic guidelines may not be sufficiently sensitive for assessing patients before admission to hospital. For example, although a temperature of >38°C is part of the diagnostic criteria (2), this symptom was notably absent in two patients. The lack of a fever in case-patient 4 may have been the result of long-term steroid use. Case-patient 6 never had a documented temperature >37.7°C and was hypothermic throughout his hospitalization. In addition, a subset of patients may have fever and myalgias without respiratory complaints or chest x-ray changes until later (cases 3 and 11). Booth et al., in a retrospective analysis of 144 patients with SARS, reported that 11% of patients had no respiratory symptoms and would not meet current criteria for SARS, despite having fever, contact history, and chest x-ray infiltrates (9). Rainer et al., in a study of 515 patients attending a SARS screening clinic, found that current World Health Organization guidelines, which emphasize respiratory tract symptoms, had a low sensitivity of 26% (10). We expanded our isolation criteria to include all patients with undifferentiated febrile illness and chest x-ray infiltrates with or without fever or respiratory symptoms until an alternative diagnosis is made, or until they defervesce and serial chest x-ray findings are normal.

The nonspecific symptoms of this illness often require establishing a history of contact with a SARS patient as a critical clue to the diagnosis (2). However, a contact history may not be forthcoming at the initial interview. In Singapore, SARS is an imported infection in which the epidemiology remains well defined with clear lines of secondary spread in the hospitals and community (3). Emphasis is placed on using local epidemiologic clues to elicit a history of visits or previous admissions to Tan Tock Seng Hospital or other healthcare facilities and to inquire whether any family members are ill or hospitalized. Patients without a contact history are still isolated if the clinical suspicion is high. A national computerized database of SARS patients and their contacts has been established to assist with this process. An epidemiology team is also used to perform more exhaustive contact tracing and liaisons with other healthcare facilities. This team also monitors and investigates any clusters of pneumonias within our hospital. Suspected case-patients are then immediately transferred to the isolation area for further investigation.

Our initial infection-control policy first required staff working in the emergency department, ICU, and isolation rooms to wear full personal protection equipment (PPE), which included an N95 mask, disposable gloves, and long-sleeve gowns. Virus transmission is likely due to droplet infection, but fecal-oral transmission has also been reported (11). The role of fomites has yet to be defined; preliminary data suggest that the virus can remain viable at room temperature for at least 2 days. Recognizing that patients with unsuspected cases of SARS could slip through our triage system onto the general wards (see case 6), all hospital staff with direct patient contact were also required to wear full PPE beginning April 8. In addition, all patients are now examined with the use of dedicated equipment.

Despite the use of PPE, two of our physicians were infected with the virus. The first physician performed an invasive procedure (bronchoscopy) on case-patient 2. The outlet port of the patient’s ventilator was later discovered to have malfunctioned during the procedure, exposing the physician to a large jet of exhaled air. Reports of protected healthcare workers becoming infected during intubation of SARS patients have also emerged from Canada (12). The recommendations on the use of PPE are likely insufficient for procedures that may promote aerosolization of respiratory secretions (13). Therefore, these procedures are now performed in a negative-pressure room (with an anteroom) using a positive air–purifying respirator suit.

The second physician (case-patient 11) may have become infected because the N95 mask was poorly fitting. Alternatively, transmission may have occurred through the conjunctival mucosa. All healthcare workers are now required to use eye protection when examining patients. Fit-testing of N95 masks is also mandatory, in addition to training on the correct use and disposal of PPE.

SARS has demonstrated remarkably efficient transmission in the hospital environment (76% of Singapore cases were nosocomially acquired) (3). At our hospital, nosocomial transmission occurred in nine persons. Our hospital structure of open wards with large numbers of beds separated by curtains may have been a contributory factor to the spread of the virus. These wards are subdivided into cubicles of four to eight beds and have open windows and ceiling fans with no controlled airflow patterns. Another likely factor was the failure to implement a policy of universal PPE use early in the outbreak. Where such measures were implemented (i.e., isolation rooms), no nosocomial transmission occurred.

In addition, because hospital staff are a recognized source of secondary transmission (3,4,6), all healthcare workers are now required to monitor their temperatures three times a day. Anyone with a respiratory illness or a temperature >37.5ºC is removed from duty, pending further evaluation. This policy has successfully prevented the secondary transmission of SARS from affected healthcare workers.

The lessons gathered from our hospital outbreak have resulted in dramatic changes to our triage and infection-control policies. All patients with undifferentiated febrile illness, respiratory complaints, or chest x-ray infiltrates are isolated, screened for SARS contacts, and nursed with full PPE. Despite continued community transmission of SARS in Singapore (the last community case was identified on May 10, 2003), the measures implemented since April 8, 2003, enabled us to identify and contain eight additional SARS cases and prevent the nosocomial transmission of the virus.