Dengue Outbreak in Key West, Florida, USA, 2009

After 3 dengue cases were acquired in Key West, Florida, we conducted a serosurvey to determine the scope of the outbreak. Thirteen residents showed recent infection (infection rate 5%; 90% CI 2%–8%), demonstrating the reemergence of dengue in Florida. Increased awareness of dengue among health care providers is needed.

D engue is the most common mosquito-borne viral infection worldwide (1); however, it has been eliminated from the continental United States. No locally acquired cases have been reported outside the Texas-Mexico border (2,3) in >60 years, despite global increases in incidence and severity. Concern exists that dengue virus (DENV) may be reintroduced (4) because it is the most frequent cause of febrile illness among travelers returning from the Caribbean, South America, and Asia (5). Immigrants and visitors from dengue-endemic countries also provide opportunity for its reintroduction (6).
In September 2009, the Florida Department of Health (FDOH) was notifi ed that a person with suspected dengue had recently traveled to Key West, Florida. The Centers for Disease Control and Prevention (CDC) (Atlanta, GA, USA) confi rmed the diagnosis. Subsequently, dengue was confi rmed in 2 Key West residents without a history of recent travel. FDOH and CDC conducted an investigation to determine size and scope of the outbreak. Cases identifi ed through passive surveillance were reported (7).

The Study
Key West is a tourist destination with >2 million visitors annually (8). Old Town, the area with reported cases, has a population of 19,846 (9). Aedes aegypti mosquitoes, the usual vectors of dengue, are widespread in Key West, but Ae. albopictus mosquitoes are uncommon.
In September 2009, we surveyed Old Town residents to estimate the infection rate and identify risk factors using stratifi ed, 1-stage cluster sampling to randomly select 911 (15%) households within 1 km of the residence of the index case-patient. The area around the residences of the index case-patients was divided into 3 strata: strata 1 and 2 (within 200 m of each case-patient) and stratum 3 (201-1,000 m). Investigators asked household members (>5 years of age) for a blood sample and information on recent illness, travel, foreign residence, and risk factors for dengue. One adult per household completed a questionnaire concerning the household. Investigators revisited unresponsive households 3 times unless the homes were empty.
Serum specimens were screened by ELISA for denguespecifi c IgM and IgG (10,11). IgG-positive samples were tested by plaque reduction neutralization test with 90% cutoff (PRNT 90 ) against DENV serotypes 1-4 and West Nile virus (12). A >4-fold difference in titer between viruses was used to identify the infecting virus. Participants reporting febrile illness within a week were tested by reverse transcription PCR for DENV and West Nile virus and by nonstructural protein 1 ELISA for DENV antigen.
We classifi ed participants with laboratory-positive DENV infection as follows: acute, if positive with reverse transcription PCR or nonstructural protein 1 ELISA; recent, if IgM-positive ELISA and PRNT 90 results were consistent with DENV infection; and presumptive recent, if they had dengue-like illness within 3 months, IgG-positive ELISA, and PRNT 90 results consistent with DENV infection. We classifi ed participants as having previous DENV infection if they had IgG-positive ELISA and PRNT 90 results without recent febrile illness.
We weighted responses to account for sampling design using different probabilities of inclusion across strata and within-household participation rates, allowing for population inference (13). CIs accounted for sampling design and fi nite population correction factors. We used weighted logistic regression to assess risk factors for infection, and resulting inferences accounted for sampling design. Tests were performed in SAS version 9.2 (SAS Institute, Cary, NC, USA); p = 0.10 was signifi cant.
Informed consent was obtained from all participants >18 years of age. Assent from the minor and informed consent from a parent were obtained for minors.
Persons who kept windows open >50% of the time, lived in a household with >50% of the yard covered with vegetation, had a bird bath, or reported receiving mosquito bites at work were more likely to be infected with DENV (Table). Persons who used air conditioning >50% of the time, emptied containers of standing water weekly, and used N,N-diethyl-m-toluamide-containing repellents were less likely to be infected. African Americans were more likely (19%) to be infected than whites (4%; p = 0.09).

Conclusions
Approximately 3%-5% of Old Town residents were infected with DENV during July-September 2009, and surveillance identifi ed new cases through October (7). Additionally, 63 cases, caused by the same strain of DENV-1, were reported there in 2010. All infections in Key West seem to have been locally acquired. Our fi ndings that less frequent use of air conditioning, leaving windows open, and yard vegetation were risk factors for infection agree with results of previous dengue studies (2,3,14).
A specifi c route and time of introduction of dengue to Key West cannot be identifi ed. The island has many tourists from dengue-endemic countries and a well-traveled population. Because the clinical features of dengue are often nonspecifi c, transmission was likely ongoing before the index case-patient's condition was diagnosed and FDOH was notifi ed. Without diagnosis of this case, the outbreak may not have been recognized. Efforts to inform local healthcare providers about the identifi cation, diagnosis, and clinical management of dengue are ongoing.
A useful discovery was the evidence of previous infections with other serotypes among residents, which indicates that a segment of the population is at risk for secondary infection, which is associated with severe dengue (15). The survey population was well traveled, placing them at risk of acquiring severe disease during travel to dengue-endemic countries or of different dengue serotypes being introduced into Key West. Other risk factors include the age, immune status, and genetic background of the infected person (15).
This investigation was performed quickly after outbreak identifi cation while transmission was ongoing, which allowed acute infections to be identifi ed. Study limitations included the circumstances that many households could not be contacted and many residents refused to participate.  We attempted to minimize participation bias through our sampling scheme. Also, presumptive infections may have been misclassifi ed, which makes the 3% infection rate more likely to be accurate. Other misclassifi cation because of prior fl avivirus infection is possible, but was minimized by careful interpretation of laboratory results. This outbreak and its continuation into 2010 demonstrate the potential for reemergence of dengue in subtropical areas of the United States where Ae. aegypti mosquitoes are present. Awareness among healthcare providers should be increased for optimal patient management and to limit outbreaks. Minimizing the effects of future dengue epidemics depends on personal protection against mosquitoes, mosquito control, early diagnosis, appropriate testing, and prompt reporting of suspected cases to public health authorities. *OR, odds ratio; DEET, N,N-diethyl-m-toluamide. †Weighted percentages are reported, reflecting the stratified, 1-stage cluster sampling design. Responses were weighted to account for the different probabilities of household inclusion across strata, within-household participation rates, and interhousehold clustering of infections.
‡Significance level, p = 0.10. Weighted logistic regression models were used to assess risk factors for recent infection, and CIs were based on the modeling accounted for the sampling design.