The National Taiwan University Hospital is a 2,400-bed teaching hospital that provides both primary and tertiary care. National Taiwan University Hospital is located in downtown Taipei not far from hospital A. The ER is organized into several sections, including triage, examination, observation, critical care, and a clean area reserved for staff activities. A SARS screening unit was established on March 17, 2003, to interview and triage febrile patients with suspected cases of SARS. The patients were questioned about the presence of fever, myalgia, gastrointestinal or respiratory symptoms, whether they had close contact with a SARS patient, and recent travel. N-95 respirators were placed on patients suspected to have SARS early during the triage process, and they were immediately placed in private rooms (SARS area) to avoid contact with others in the ER. A daily record was maintained of all patients screened in the SARS screening unit. SARS cases were defined according to the World Health Organization criteria (10), modified to expand the definition of contact to include any healthcare setting with nosocomial transmission.

Since March 14, 2003, infection-control measures required that all healthcare workers who had contact with patients with SARS use personal protective equipments, including gown, gloves, N-95 respirators, disposable cap, and shoe covers. Later, a face shield was included for healthcare workers with close contact to SARS patients.

Healthcare workers who had any contact with SARS patients or their environment were placed under surveillance for 14 days after the last exposure. Those who had unprotected exposure, or those who were protected but had high-risk exposures to SARS patients were excluded from new duty assignments and were restricted from direct patient care and contact with other healthcare workers. Any healthcare worker in whom fever developed was placed in specially designated isolation wards.

Infection-control measures in the non-SARS area were upgraded stepwise in response to possible healthcare-associated transmission and the increasing possibility of community spread of SARS. After the outbreak in hospital A, healthcare workers wore N-95 respirators for all patient care in the ER. In the ER, SARS areas were cleaned at least three times; non-SARS areas were cleaned once a day.

We obtained source and contact information for all patients identified at our hospital as having suspected or probable SARS. After the outbreak in hospital A, health insurance records were used to trace prior visits to other hospitals. In addition, we obtained information about social, hospital, and occupational contacts and other members of the household who had been exposed to suspected SARS patients within 10 days before the onset of their symptoms. All close contacts exposed to SARS patients during the period from 2 days before the onset of fever to the time of isolation were traced to identify the need for quarantine. Inpatients who had close contact with SARS patients were quarantined in private rooms, and contact and droplet precautions were implemented. On April 30, because of the occurrence of two closely spaced cases of SARS in the ER observation unit (a non-SARS area), we immediately identified a potential outbreak. Accordingly, we screened all inpatients that had been admitted through the ER and telephoned all the patients who stayed in the observation unit from April 23 through April 29.

Soon after the cluster of SARS was identified among healthcare workers on May 12, we devised a questionnaire to identify the source and the factors contributing to infection. Data about daily exposures were collected from April 30 through May 12, 2003. These data included contact with SARS patients; work areas; day, time, and characteristics of duty; exposure to high-risk aerosol-generating procedures; use of personal protective equipment; hand hygiene practices; and contact with other healthcare workers who did not use N-95 respirators.

On May 15, surfaces of environment and equipment were sampled with moistened sterile cotton swabs. The swabs were spread immediately onto 3 mL of viral transport medium. Samples were collected from various objects in different areas of ER.

Air samples were taken with both high-volume and low-volume samplers in 10 locations in the ER. We used a high-volume air sampler (XMX Virtual Impactor, Dycor Technologies Ltd., Alberta, Canada) to draw air at calibrated sampling rates of 400 L per min for 5 min into a collector with 5 mL of phosphate-buffered saline (PBS). We also used a low-volume pump (Sidekick, SKC Inc., Eighty Four, PA) to draw air at a calibrated sampling rate of 2.0 L per min for 10 h onto a 37-mm diameter, 0.3-μm pore size polytetrafluoroethylene membrane filters. The collected samples were then frozen at –70°C before RNA extraction. Once environmental contamination was identified, cleaning was performed. Follow-up surveillance for contaminated objects was conducted on May 25.

Swab samples were suspended in 5 mL of PBS or 3 mL of viral transport medium. Total RNA from 140 μL of the sample was extracted by using a QIAamp Virus RNA Mini Kit (Qiagen, Hilden, Germany) and eluted in 60 μL of buffer. A volume of 5 μL of RNA solution was analyzed. RealArt HPA-Coronavirus LC RT-PCR Reagents (Roche, Penzberg, Germany) were used for one-step real-time reverse transcription–polymerase chain reaction (RT-PCR) in the Roche LightCycler Instrument (Roche, Mannheim, Germany). This ready-to-use system is designed for specific amplification of the 80-bp region of the SARS-CoV genome and for directly detecting the specific amplicon in fluorimeter channel F1 of the LightCycler Instrument. In addition, these reagents contain a second heterologous amplification system to identify possible PCR inhibition. Internal controls in each run of the experiment included two negative controls (one for RNA extraction and one for RT-PCR) and four quantification standards (1 x 10¹ copies/μL, 1 x 10² copies/μL, 1 x 10³ copies/μL, and 1 x 10⁴ copies/μL). We used the following formula to convert the values determined by using the standard curve into copies per milliliter of sample material: results (copies/mL) = result (copies/μL) x elution volume (μL)/sample volume (mL). Data were presented as number of copies per sample.

From March 15 through April 22, a median of 6 patients per day (range 0–29) were screened at the ER for febrile illnesses (Figure 1). After the outbreak in hospital A, a median of 36 patients per day (range 21–67) were screened. Thus, the ER was used to screen a large portion of persons during this rapidly progressing epidemic. Of 754 patients screened at the ER from April 23 through May 12, a total of 63 patients were identified as SARS cases and were admitted to National Taiwan University Hospital, 68 SARS patients were transferred to another hospital, and 155 received care in a temporarily designated ER area because of shortages of isolation rooms and staff. On May 7, up to 18 SARS patients stayed in the ER overnight. Of 232 SARS patients admitted to the National Taiwan University Hospital from March 14 through June 19, 31 (13.4%) did not have a history of travel, exposure to SARS patients, or a hospital visit within 10 days before illness, and the only contact history was a stay at the National Taiwan University Hospital ER.

We identified three distinct clusters by plotting the dates of onset of fever for each case (Figure 2) and allocation of bed numbers in the observation unit of patients involved (Figure 3). The first cluster of five patients had disease onset from April 29 through May 1; the second cluster of 14 cases began on May 4, and the third cluster of 12 cases began on May 11, 2003. In the third cluster, all the cases were in healthcare workers. The first cluster affected patients located in three neighboring beds in the observation unit of the ER (Figure 3). The second cluster affected patients located in four nearby beds and a fifth bed that was >3 m away. The distance between beds was approximately 1 m. None of the cases occurred in beds 9–18, which are separated by a half wall.

The index patient in the first cluster was an afebrile 73-year-old man who was admitted to the ER because of severe dyspnea (Figure 4, 1P). He was kept in the observation unit from April 23 through April 25, 2003. He was thought to have congestive heart failure and chronic obstructive lung disease and treated with aerosolized medication. He was admitted to the cardiology ward on April 25. A temperature of >38°C developed on April 27, and a chest radiograph taken on the same day indicated a new infiltrate. He was immediately transferred to a negative-pressure isolation room. He had not given this history, but after checking his health insurance card, we learned that he had visited hospital A on April 14 and April 15. He died on April 30. Sputum samples were positive for SARS-CoV RNA. On autopsy, he was found to have had an acute myocardial infarction. A small ground-glass density in the lung was compatible with viral pneumonitis.

The second patient was another 73-year-old man who had cough and fever for 2 weeks. He had been treated in the ER observation unit from April 24 through April 29. He returned to the ER on April 30 and was diagnosed with probable SARS. Because of the occurrence of two closely spaced cases of SARS in the ER observation unit (a non-SARS area), we immediately identified a potential outbreak. Contact tracing identified a cluster of cases in three patients and two nursing aids (Table 1, patient 2–6; Figure 4). All five cases were diagnosed as probable SARS, and results of three tests were positive for SARS-CoV RNA. The contacts were quarantined, and no tertiary cases emerged.

The second cluster began on May 8. A 46-year-old, otherwise-healthy woman (patient 8) was admitted with probable SARS. A week earlier she had taken care of her mother (patient 9) in the ER observation unit. She indicated that she did not have contact with other SARS patients, including those identified in the first cluster. Accordingly, we screened all patients who stayed in the observation unit from April 30 through May 8. This cluster affected six patients, three family members, and five nursing aids (Figures 2 and 4; Table 1, patients 7–20). Patient 17 was the only tertiary case.

The third cluster was noted on May 12, when the infection control team was informed that fever developed in three healthcare workers who had been isolated. The exact contact source could not be identified. Thus, we quarantined all the ER healthcare workers and suspended ER operations for 2 weeks. SARS related to the ER developed in 12 healthcare workers from May 11 through May 16 (Figure 2, solid lines). Six of the healthcare workers who became ill had close contact with SARS patients. However, patient contact and time of exposure were different. The healthcare workers were one desk clerk, two physicians, one radiology technician, and two nurses. All had followed infection-control precautions. Six other healthcare workers who became ill indicated that they did not have close contact with SARS patients. These workers were four nurses and two cleaners. These 12 healthcare workers differed from each other according to duty pattern, service time, work areas, and time of exposure to the unit (data not shown). Source and contact tracing failed to identify a common source. We therefore postulated that they might have acquired SARS through indirect contact.

On May 15, we collected 119 environment samples, including 100 surface samples and 19 air samples (Table 2). Nine samples were positive for SARS-CoV RNA. These included the buttons of the drinking water fountains in the triage and the observation unit; a bedside chair in the observation unit; the outlet of the central air supply; a table top, bedding and bed edge in a SARS area; and a bookshelf and bedding in the clean area. None of 19 air samples tested positive for viral RNA. The highest viral load was obtained from a bedside chair in the observation unit (2,570 to 25,700 copies per sample).

Targeted cleaning of the ER environment was performed. Follow-up surveillance was conducted on May 25. Nine samples were collected from previously contaminated surfaces, 21 samples from other areas in the ER, and 15 samples from SARS wards. All 45 samples were negative for SARS-CoV RNA. All personnel who had contact with SARS patients or their environments were reeducated on infection-control measures. Particular attention was paid to hand hygiene and routine environmental cleaning. The workload for healthcare workers was reduced. All patient beds were placed at least 2 m apart. No further cases of SARS related to the ER occurred after May 17, 2003.