We obtained demographic data and travel-related information (last port of embarkation and flight times) by direct interview and retrospective review of clinical notes. Duration of travel, which was calculated from flight times and transit time within airports, was categorized into 4 groups: short haul (<3 hours), medium haul (3–6 hours), long haul (6–15 hours), and extra-long haul (>15 hours). Clinical data were collected for symptoms and date and time of symptom onset (to the nearest hour in most instances and to the nearest 8-hour block [midnight to 7:59 am, 8:00 am to 3.59 pm, or 4:00 pm to 11:59 pm] if the patient was unsure). Symptom onset time was further categorized as before embarkation, during travel, or after disembarkation in Singapore. We also categorized hospitalized case-patients on the basis of whether their symptoms met US Centers for Disease Control and Prevention (CDC) ILI criteria (body temperature >37.8°C with either cough or sore throat in the absence of an alternative diagnosis) and WHO ILI criteria (body temperature >38°C with either cough or sore throat in the absence of an alternative diagnosis) (11,12).

A confirmed case of pandemic (H1N1) 2009 was defined as an ILI in a patient with a temperature >37.5°C, cough, rhinorrhea, sore throat, or myalgia and with laboratory confirmation of infection by real-time reverse transcription–PCR performed on respiratory samples (sputum or combined nasal and throat swab specimens). An imported, travel-associated case was defined as above but occurred in a person with recent travel outside Singapore who had arrived in Singapore during the containment period and had illness onset within 10 days of arrival.

We compared proportions by using the χ² test. Means were compared by use of t tests for dichotomous variables and 1-way analysis of variance for multichotomous variables. Stata 10.0 for Windows (StataCorp LP, College Station, TX, USA) was used for all statistical analyses.

From April 27 through June 27, 2009 (epidemiologic week 25), 152 persons with confirmed pandemic (H1N1) 2009 were admitted to TTSH. Of the 152 patients, 116 met the definitions for having an imported, travel-associated case; the rest either did not have a history of travel or had onset of symptoms >10 days after their travel ended. These 116 cases span 5 weeks (epidemiologic weeks 21–25) from when the first confirmed case-patient arrived in Singapore on May 26, 2009.

Demographic and travel data for patients with cases of imported pandemic (H1N1) 2009 are shown in Table 1. Infections involved equal numbers of male and female travelers and occurred predominantly in young travelers (mean age 28.5 years; 70% were <30 and 7% were >50 years of age). Half of the travelers were Singaporeans returning from abroad, and the other half were travelers from other countries.

As shown in Figure 1, the regions where travelers were exposed to pandemic (H1N1) 2009 virus changed rapidly during the 5-week period of study. Most of the cases in epidemiologic weeks 21 and 22 (May 24 through June 6) were in patients who acquired their infection in the United States; however, within 2 weeks, this had changed dramatically, with a large proportion of cases coming from Australasia in epidemiologic week 23 (June 7–13). Within a week, this situation changed yet again; most infections originated from Southeast Asia by epidemiologic weeks 24–25 (June 14–27).

The first cases of pandemic (H1N1) 2009 detected among travelers arriving from the Philippines (week 22), Thailand (week 23), and Indonesia (week 25) indicated potential community transmission in those countries earlier than official announcements. Infections from Southeast Asian countries accounted for 29% of imported cases during the containment phase; the Philippines were linked to 23% cases, mostly in weeks 24–25.

Travel duration reflects the distance between the regions of exposure and Singapore (Table 1). Of the 116 case-patients, 19% had a short travel duration (<3 hours), and 20% had a long travel duration (>15 hours). Although most case-patients became ill after arrival at their destination, 25% were ill before travel and boarded their flight despite symptomatic illness, and 15% became ill while traveling.

Doctors based at the airport referred 15 (12.9%) of the 116 patients to TTSH; thermal scanners used to screen arriving passengers had detected fever in 14 of these 15 patients. Of the remaining 101 patients, 51 (44%) self-reported to the screening center at TTSH and 50 (43%) were referred by doctors in the community.

At the time of examination, 72% of case-patients had temperatures >37.5°C. However, only 61.2% had temperatures >37.8°C and 54% had temperatures >38°C, the temperature criteria used in the US CDC and WHO ILI case definitions, respectively. Considering the entire symptom complex, 51% of patients would have fulfilled ILI criteria as defined by the US CDC, and 44% would have fulfilled the WHO criteria (11,12).

As the pandemic shifted toward Asian ports of embarkation, the number of case-patients with travel durations of <8 hours increased (Figure 2, panel A). The time of symptom onset relative to arrival in Singapore is shown in Figure 2, panel B; the figure does not include information for patients who were symptomatic before embarkation. Time of symptom onset was progressively closer to the time of arrival in Singapore for those arriving from longer distances (p = 0.001). Patients with longer travel durations were also more likely to have onset of symptoms before arrival (p = 0.04).

Port of embarkation, clinical symptoms, and duration of travel did not predict delay to isolation (Table 2). However, case-patients referred to TTSS by airport doctors had a shorter time to isolation (0.76 days) than self-referred patients or those referred by other sources (1.6–1.9 days). The number of case-patients referred by airport doctors increased over the 5-week period, but they represent only 12% of all travel-associated cases (Figure 3). Although the mean duration to isolation did not increase significantly over the study period, total numbers of case-patients with delays to isolation increased as the volume of travel-associated cases rose.

Given Singapore’s position as a major travel hub, with passenger traffic at Changi International Airport exceeding 37 million in 2008 (13), there is an ever-present risk of importation and subsequent community transmission of emerging respiratory infections. Such importation and transmission occurred during the SARS epidemic of 2003 (14) and remains a concern during the current pandemic of pandemic (H1N1) 2009 virus. In any epidemic, every imported case may start a cluster of locally transmitted cases, but the number of imported cases and delays to isolation are particularly important at the start of the epidemic in determining speed of spread within the community. Improving detection and shortening time to isolation of infectious persons could modify the outbreak curve and allow more time for improving community preparedness.

Improving detection and shortening the time to isolation of sick persons is the rationale for using airport thermal scanners. Our data show that for the minority of cases detected by airport thermal scanners, detection does result in a hospital referral by an airport doctor and shorter time to isolation. However, intrinsic limitations of airport thermal scanners are that passengers have to become symptomatic before disembarking from a flight and have a fever high enough to be detected. Our data show that >30% of case-patients from all flights >3 hours had symptom onset before arrival, but overall, only 12% of all case-patients were detected by thermal scanners, suggesting that thermal scanners detected 40% of those symptomatic patients. This early detection and isolation may still have a valuable adjunctive role, especially in the initial phase of outbreaks. Situations favoring the use of airport thermal scanners include short-incubation diseases and geographically distant outbreak epicenters, such that arriving passengers have been on a long-haul flight. However, if the converse were true, with transmission occurring in nearby countries and passengers arriving from short-haul flights, symptoms would develop in most passengers who become ill after entry and, thus, would be missed by airport thermal scanners.

The fact that one fourth of the case-patients in our study boarded a plane after becoming ill and traveled despite having symptoms illustrates the role of travelers in disseminating infection in a highly interconnected world. It raises the question of whether exit screening should be considered. However, the effectiveness of exit screening will depend on the role of asymptomatic persons in transmission, and such screening will still miss persons who are incubating the infection. Exit screening would severely hinder international travel, and because of its questionable efficacy, it may not be justified and may be contrary to the intent of the International Health Regulations 2005 (15)

Because 44% of the case-patients in this study were self-referred and 43% were referred by community physicians, prevention efforts should focus on other strategies. For example, health advisories should be distributed to arriving passengers, encouraging them to seek medical care if ill, and steps should be taken to improve detection by physicians who see patients in their clinics. The importance of clinical judgment, epidemiologic history, and access to diagnostic testing is emphasized by the substantial minority of case-patients who would not have met the WHO or US CDC ILI criteria but who did have laboratory-confirmed infection. However, our study did have limitations: clinical features were assessed at a single time point (on arrival at the hospital screening center) and there was potential recall bias.

Our data demonstrate how swiftly situations can change in a fast-moving pandemic; affected areas shifted from the Americas to Australasia to Asia within a matter of days. These rapid shifts pose a tremendous challenge to health authorities responsible for outbreak management, and they emphasize the narrow window of opportunity during which interventions can slow an epidemic. The monitoring of travelers fulfills a vital sentinel surveillance function, providing an early indicator of community transmission in countries even before transmission has been officially confirmed. Understanding how travel-associated infections propagated the first wave of this pandemic yields rich insights into how health authorities might respond to future outbreaks of emerging respiratory infections.