GeoSentinel sites are specialized travel/tropical medicine clinics on 6 continents staffed by clinicians who are recruited on the basis of demonstrated training, experience, and publication in travel and tropical medicine literature. They contribute clinician-based information on all ill travelers seen, including travel history (additional detail is available from www.geosentinel.org) (9,14). The sites that account for most patient intake are within academic centers; several smaller volume sites (almost all with current academic affiliation) are in freestanding locations. The intake at sites reflects a mixed population of tertiary care and self-referred patients. Some sites are restricted to outpatients, and no one site limits its entire practice to ill travelers. To be eligible for inclusion in the GeoSentinel database, patients must have crossed an international border and be seeking medical advice at a GeoSentinel clinic for a presumed travel-related illness. Anonymous surveillance data that cannot be linked to an individual patient are entered into an SQL database at a central data center. Final diagnoses reported by physicians are used to assign diagnostic codes from a standardized list of >500 etiologic or syndromic diagnoses (9).

All returning travelers who reported to a GeoSentinel site in their current country of residence from October 1, 1997, to March 1, 2006, were eligible for analysis. Many GeoSentinel sites also enter data separately on immigrants with no other travel but the initial immigration trip. None of the patients in this immigrant dataset had a diagnosis of dengue acquired in the country of origin. The current study is restricted to traditional travelers, which also includes immigrants who subsequently traveled from their current country of residence.

Patients were excluded if no confirmed or probable diagnosis was reported. A case of travel-associated dengue was defined per current annual surveillance reports (15–17), which consider both probable and confirmed cases of dengue (18). A case of travel-associated dengue was defined as laboratory-diagnosed dengue in a resident of a non–dengue-endemic area who has traveled to a dengue-endemic area in the 14 days before symptom onset. Laboratory-diagnosed dengue was determined by isolation of dengue virus, virus antigen, or viral RNA, or a serum sample positive for either immunoglobulin (Ig) M or a very high titer of IgG by ELISA. All sites use best available reference diagnostics for their respective countries, which may include well-characterized commercial kits. GeoSentinel criteria for the diagnosis of malaria have been reported (19).

Analysis of dengue reports over time was based on proportionate morbidity (the number of patients with dengue fever as a proportion of the number of ill returned travelers visiting a GeoSentinel clinic in that month). Analysis of annual and monthly cycles was based on monthly proportionate morbidity aggregated over all years of data included in the analysis. Patients who were reported as having dengue were compared with all other ill returned travelers in GeoSentinel. A subanalysis, comparing dengue patients with malaria patients, was also performed. We used χ² or Fisher exact test as appropriate with a 2-sided significance level of 0.05. Data analysis was performed by using SAS statistical package version 9 (SAS Institute, Cary, NC, USA).

A comparison of the annual trends in illness from dengue as a proportion of all diagnoses in ill returned travelers showed sustained increases in dengue proportionate morbidity, represented by peaks that are both high and broad in 1998 and 2002. There was also a narrow peak in October 2003 and an increase in late 2005 (Figure 1). When dengue reports were segregated by region, the increases in 1998 and 2002 were found entirely in travelers to Southeast Asia; for 2003, in travelers to South Central Asia; and for 2005, in travelers to South Central Asia and Indonesia. These increases correspond to known epidemic years within local populations for those regions (20,21).

Since travel-related dengue was found to originate most commonly in Southeast Asia, more detailed analysis could be performed for that region. Dengue proportionate morbidity among ill returned persons who had traveled to Southeast Asia, which was a mean of 71 per 1,000 during the cumulative 1997–2006 period (Figure 2, panel A), was 159 cases per 1,000 ill returned travelers during 1998 and 2002 taken together (Figure 2, panel A); proportionate morbidity reached a peak of >200 cases per 1,000 ill returned travelers during June and July. Of the 264 Southeast Asian cases, 154 that were acquired in Thailand could be plotted separately (Figure 2, panel B). Dengue proportionate morbidity among ill persons who had traveled to Thailand, which was an average of 101 cases per 1,000 during the cumulative 1997–2006 period (Figure 2, panel B), was 257 cases per 1,000 ill returned travelers during 1998 and 2002 taken together (Figure 2, panel B) and was >500 cases per 1,000 ill returned travelers during the peak month of June (i.e., more than half of all ill travelers returning from Thailand had dengue).

Figure 3 shows month-by-month dengue cases as a proportion of all illness in ill returned travelers during the study period for each region separately. For Southeast Asia, dengue cases generally peaked in June and September in typical nonepidemic years. However, an examination of the outbreak years of 1998 and 2002 showed that seasonal patterns changed markedly when compared with nonoutbreak years; excess cases were seen for every month except January, and a high and sustained peak occurred from April through August (Figure 2, panel A). In Thailand, during the outbreak years, proportionate morbidity exceeded the mean 1997–2006 proportionate morbidity (Figure 2, panel B) for all months except January. Notably, the major peak of illness began in April, a time of minimal dengue activity in nonoutbreak years. The major epidemic peak in sentinel travelers preceded the epidemic pattern in the local population during 1998 and 2002, as reflected in Thai reports to the World Health Organization (20).

An examination of seasonality in travelers for other regions (Figure 3) showed that dengue cases were higher from September through December in South Central Asia (especially in India, which accounted for most of our cases in South Central Asia; Table 1). A sharp and consistent major peak can be seen each October throughout the study period. This closely tracks the monthly pattern in the Indian population during 2003–2006, years for which robust Indian national data are available (21). The number of dengue cases in travelers was higher from August through December in the Caribbean; it was highest in March in South America, especially in Brazil, which accounted for most of our cases in South America (Table 1). This peak is also consistently seen in the Brazilian population (22) each year. No evidence of a strong seasonal pattern was found in travelers to Central America and Africa, although the numbers of travelers to these regions in our sample were small.

The characteristics of travelers with dengue are compared in Table 2 with the characteristics of those with all other diagnoses. The mean age of dengue patients was 33.8 years; the male:female ratio was 1.17:1. Of the patients studied, 69.3% were traveling only for tourism, and the median trip duration was <28 days. Most of the dengue case-patients (61%) had a pretravel encounter, a significantly higher percentage than for ill returned travelers without dengue (53%; p<0.005). Dengue is overwhelmingly a disease of young adults 18–44 years of age. As expected, due to the short incubation period, >75% of dengue case-patients sought treatment within 2 weeks after return. In addition, significantly more dengue patients were hospitalized (24% vs. 6%; p<0.001), a level similar to the 25% rate reported in a study of European travelers (4).

During the study period, 1,605 (6.4%) ill returned travelers reported to GeoSentinel had been given a diagnosis of malaria. A comparison of the characteristics of travelers with dengue and those with malaria shows some important differences. Unlike dengue, which affects both sexes almost equally, malaria is more common in male travelers (11,23). Patients with malaria were less likely to have a recorded pretravel encounter. Duration of travel (median 34 days) was significantly longer than for travelers returning with dengue (median 28 days; p<0.05), although the difference was not large. Malaria was much more common in first- or second-generation immigrants visiting friends and relatives (Table 2).

Overall, the proportion of travelers with dengue in the GeoSentinel database (21 cases per 1,000 ill returned travelers) was less than the proportion seen with malaria (64 cases per 1,000 ill returned travelers; Table 1). This finding, however, was mostly due to the disproportionate numbers of travelers returning ill from Africa where malaria is highly prevalent (168 cases per 1,000 ill returned travelers) and where dengue is rare (3 cases per 1,000 ill returned travelers). A similar situation applies to Oceania where malaria (300 cases per 1,000 ill returned travelers) is significantly more frequent than dengue fever (36 cases per 1,000 ill returned travelers). For other regions, the proportionate morbidity due to dengue was higher than that due to malaria, except for South America, where proportionate morbidity was approximately equal (20 cases vs. 16 cases per 1,000 ill returned travelers).

Analysis of travel to several countries was possible. Dengue proportionate morbidity (101 cases per 1,000 ill returned travelers) was dramatically higher than that for malaria (5 cases per 1,000 ill returned travelers) in travelers returning from Thailand and exceeded that for malaria in travelers returning from Brazil and India.