RMI is composed of 29 atolls and five islands with a total land mass of 70 square miles (sq mi) spread across ∼750,000 sq mi of ocean (Figure S1). The 2011 population of RMI was 53,158 (759 individuals/sq mi), 70% of which resided on Majuro atoll or Ebeye island (7,413 and 80,117 individuals/sq mi, respectively) [18]. Forty percent of the population was aged ≤14 years, and the sex ratio was 102 males to 100 females.

Surveillance data were collected during the outbreak, summarized weekly, and reported to WHO. After the outbreak had ended, a retrospective analysis of surveillance data was performed to: 1) describe the epidemiology of the 2011–2012 outbreak, including disease severity; 2) estimate the proportion of secondary DENV infections; 3) describe the molecular epidemiology of the DENV(s) responsible for the outbreak; and 4) identify the water containers producing vector mosquitoes.

Suspected cases identified at Majuro and Ebeye Hospitals were reported directly to MOH via the Dengue Surveillance Form (DSF; Figure S2A), which was implemented for the outbreak. DSF data were reported to MOH via short wave radio from all other health facilities. Case-patients’ ultimate severity of illness were captured with a second DSF (Figure S2B) that was completed upon patient discharge or follow-up evaluation.

Serum specimens were collected from all suspected dengue cases upon presentation to Majuro and Ebeye Hospitals and tested by RDT (Dengue Duo, Standard Diagnostics, Haryana, India). Convalescent specimens were collected upon discharge or during follow-up evaluation. For suspected dengue cases that presented to other health facilities in RMI, whole blood was collected at presentation and tested by RDT at the health facility. A convenience sample of both RDT-positive and-negative serum specimens were sent to CDC for confirmatory testing by multiplex, DENV-type-specific, real-time reverse transcriptase-polymerase chain reaction (RT-PCR) [19]. All specimens that tested negative by RT-PCR were further tested by anti-DENV IgM antibody capture ELISA (InBios International, Inc.; Seattle, WA). Unless a convalescent specimen was available, specimen diagnosis was based on the single acute specimen.

A suspected dengue case was an individual for whom a health care provider suspected dengue as the cause of illness and reported the case to MOH. A laboratory-positive case was a suspected dengue case for which: 1) DENV nucleic acid was detected by RT-PCR; 2) DENV non-structural protein 1 (NS1) was detected by RDT; or 3) anti-DENV IgM antibody was detected by RDT or ELISA. A laboratory-negative case was a suspected dengue case that had no laboratory diagnostic evidence of DENV infection. Dengue, dengue with warning signs, and severe dengue were defined according to 2009 WHO guidelines [2].

A representative sample of all laboratory-positive cases was selected to estimate the rates of primary and secondary infection. Cases were stratified by age group with optimal allocation for comparison between age groups. Sample size was calculated to estimate the proportion of secondary infections by age group based on data from the 2010 dengue epidemic in Puerto Rico [20], an error of 20%, 95% significance, and an expected 20% of specimens having insufficient specimen for testing. Of 218 selected acute (i.e., collected within five days of illness onset) specimens, 194 (89%) were tested at a dilution of 1∶100 to detect anti-DENV IgG antibody by ELISA using DENV-1–4 antigen and a cut-off value of OD₄₅₀≥0.15 [21], [22]. Secondary infection was defined by detection of anti-DENV IgG antibody in a laboratory-positive acute specimen. Primary infection was defined by lack of detection of anti-DENV IgG antibody in a laboratory-positive acute specimen.

Serum specimens in which DENV nucleic acid was detected and with CT values <25 were inoculated into cultures of C6/36 cells, in which growth of DENV was detected by RT-PCR and indirect immunofluorescence [23]. Viral RNA was extracted from culture supernatant using the Universal BioRobot System (Qiagen; Valencia, CA). The envelope glycoprotein (E) gene (1,485 base pairs) was successfully amplified and sequenced in one specimen. Multiple sequence alignment with 39 other Indonesian genotype strains obtained from GenBank was performed using ClustalW available in MEGA 5 (megasoftware.net). GTR+Γ+I4 was selected as the best nucleotide substitution model by MODELTEST v3.7. Phylogenetic relationship was inferred by 60 million Markov Chain Monte Carlo run in BEAST v1.7.5 (beast.bio.ed.ac.uk) under Bayesian skyline prior assuming a relaxed lognormal molecular clock. All ESS values reported >400. A Bayesian maximum credibility tree was constructed in TreeAnnotator (found in the same BEAST package) and visualized in FigTree v1.3.1. Posterior probabilities were used as statistic for internal nodes. Maximum likelihood phylogenetic trees with 1,000 bootstrap replicates were also inferred using MEGA 5; both trees rendered nearly identical tree topologies, confirming genetic relatedness. Genotype was referred to by previously described nomenclature [24], [25].

Relative abundance and species composition of juvenile mosquito populations in aquatic container habitats were assessed in 155 households on Majuro atoll during 3–8 November, 2011. Sampling focused on four communities (Rita, Uliga, Long Island and Laura) where the incidence of suspected dengue was high during 10 October–5 November 2011. All outdoor container habitats at selected households were inspected for the presence of standing water, juvenile mosquitoes, and if >20 pupae were present in the container. Mosquito specimens were retained from a subset of samples from each container habitat class and preserved in ethanol (larvae) or reared to eclosion (pupae) for identification to species.

All surveillance data were entered into a Microsoft Access (Microsoft Corporation, Redmond, WA) database at MOH. Rates of suspected and laboratory-positive dengue cases were calculated using population denominators from the 2011 RMI Census [18]. Unless otherwise stated, statistical differences in proportions were evaluated using the Chi-squared test. The nonparametric equality-of-medians test was used to compare median age between groups. Relative risk ratios (RR) were used to calculate differences between effect sizes. Graphs were produced in Microsoft Excel (Microsoft Corp., Redmond, WA). Statistical calculations were performed in EpiInfo7 (CDC, Atlanta, GA).

This investigation underwent institutional review at CDC and was determined to be public health practice and not research; as such, Institutional Review Board approval was not required. All specimens tested in RMI were not anonymized to facilitate reporting of diagnostic test results to clinicians. All specimens sent to CDC were anonymized prior to shipment.

The first identified laboratory-positive case had onset of illness on October 14, 2011 and resided in Majuro atoll (Fig. 1). The number of suspected cases began to increase steeply one week later and peaked on November 4. Cases continued to be identified at comparatively low levels through the first months of 2012, and the outbreak was declared over in February. A median of 8 (range: 0–54) suspected cases were identified per day during the outbreak.

Serum specimens were collected upon presentation from all patients with dengue-like illness. Specimens were collected a median of two days (25^th–75^th quartile: 1–3) post-onset of illness. Paired acute and convalescent specimens were available for 59 (3.7%) cases.

Of 1,603 suspected dengue cases identified during the outbreak (3.0% of RMI residents), 657 (41%) were laboratory-negative, 79 (4.9%) were not tested, and 867 (54%; 1.6% of RMI residents) were laboratory-positive. Of 1,379 specimens tested with the RDT in RMI, 626 (45%) were positive: 536 (86%) by detection of NS1, 73 (12%) by detection of anti-DENV IgM antibody, and 17 (2.7%) by detection of both NS1 and anti-DENV IgM antibody. From a convenience sample of 845 RDT-positive and -negative specimens sent to CDC and tested by RT-PCR, DENV-4 was detected in 482 (57%) and was the only DENV detected during the outbreak. Of the 363 specimens that tested negative by RT-PCR and were further tested by IgM ELISA at CDC, anti-DENV IgM antibody was detected in 45 (12%).

Laboratory-positive cases were identified from seven of the 25 inhabited atolls and islands that comprise RMI (Fig. 2). Among these regions, the median rate of suspected and laboratory-positive cases per population was 2.4% (range: 0.1–4.6%) and 0.6% (range: 0.2–2.7%), respectively; both were highest in Majuro atoll.

Most (72%) laboratory-positive cases were aged 10–29 years (Fig. 3A). The highest rate of laboratory-positive cases per population (3.4%) was in individuals aged 15–19 years, and lowest (0.4%) in individuals aged 0–4 or ≥70 years. Males and females were equally represented in all age groups except 0–4 year-olds (p<0.001), in which 79% of laboratory-positive cases were male.

From a representative sample of 194 laboratory-positive acute specimens, 184 (95%) had evidence of secondary infection (Fig. 3B). Rate of secondary infection by age group was 50% in 0–4 year-olds, and increased with age group until a plateau of ∼90% was reached in 10–14 year-olds.

Following virus isolation and E gene sequencing, Bayesian phylogenetic analysis demonstrated that the virus circulating in RMI is distinct from other DENV-4 previously identified in the Pacific islands, and instead belongs to clade II of the Indonesian genotype that is geographically associated with Southeast Asia (Fig. 4). The most closely related DENV-4 was isolated from cases that occurred in 2012: an individual from Japan with travel history to RMI, and a resident of Chuuk, FSM. Other closely related viruses were detected in Singapore in 2011 and Indonesia in 2004.

Laboratory-positive cases were not significantly associated with sex, age, or ethnicity (Table 1). Although more laboratory-positive (79%) than laboratory-negative (75%) cases met the 2009 WHO dengue case definition [2], this difference only approached statistical significance (p = 0.06). Laboratory-positive case-patients were 1.8 (95% confidence interval [95% CI]: 1.3–2.5) times more likely to be hospitalized than laboratory-negative case-patients (p<0.001). Laboratory-positive case-patients had dengue with warning signs significantly more frequently than laboratory-negative case-patients (p = 0.02). The most frequently identified warning signs among laboratory-positive case-patients were persistent vomiting (35%) and abdominal pain (59%); the latter was significantly more common in laboratory-negative cases (75%; p = 0.007). Fluid accumulation (e.g., pleural effusions, ascites) was only observed in laboratory-positive cases (Fischer’s exact test, p = 0.16). Concomitant rise in hematocrit and decrease in platelets was 4.3 times (95% CI: 2.2–8.4) more likely to occur in laboratory-positive than laboratory-negative case-patients (p<0.001). Six (0.7%) laboratory-positive case-patients had severe dengue. No fatal cases were reported.

Four infants aged <1 year tested positive with the RDT: three by detection of NS1, and one by detection of anti-DENV IgM antibody; all were hospitalized. One infant was a neonate male born to a febrile mother in whom NS1 was detected on the day of parturition. On day seven of life, the infant developed severe dengue with cyanosis and severe bleeding requiring blood transfusion. Both the baby and mother recovered and were ultimately discharged. As 111 pregnant women gave birth during the outbreak, the rate of documented vertical DENV transmission resulting in dengue during the outbreak was 0.9% (1/111). Four additional pregnant, laboratory-positive females were identified during the outbreak; all recovered without complication. No reported fetal losses or perinatal deaths were associated with maternal DENV infection during the outbreak.

An 11-year-old boy from Utrik atoll had onset of fever in early November and subsequently had streaks of blood in his stool. After being transported to Majuro Hospital, he was given intravenous ampicillin, cephazoline, and metronidazole for a presumptive diagnosis of typhoid fever; however, his fever persisted. White blood cell count, platelet count, hemoglobin, and hematocrit were normal. NS1 was detected on day 8 of illness, and Salmonella typhi was identified in a blood culture the following day. He was discharged home in stable condition after seven days of hospitalization.

An 18-year-old male in his third month of treatment for leprosy was admitted with a two-day history of fever and malaise. On examination he was febrile with patchy nodular skin lesions on the trunk and extremities. NS1 and DENV-4 were detected in a serum specimen. His hospital course was unremarkable and he continued treatment for leprosy, including a decreased dose of prednisone. He was discharged on day six of hospitalization.

Mosquito larvae or pupae were detected at 87% of inspected homes (i.e., House Index) and in 64% of containers with standing water outside homes (i.e., Container Index), with an estimated density of 413 mosquito-positive containers per 100 households (i.e., Breteau Index) (Table 2). Indoor inspections were performed at a subset of selected households; however, water-filled containers were rarely observed and all were negative for juvenile Aedes spp. mosquitoes (data not shown). Large water storage receptacles were the most abundant water-filled container (235 per 100 households inspected), and accounted for 32% of mosquito-positive containers and 50% of containers with >20 pupae. Discarded tires, domestic utensils (e.g., washing or cooking items, trash containers) and small refuse items (e.g., cans, disposable food packaging) collectively represented 44% of mosquito-positive containers and 29% of containers with >20 pupae.

A total of 738 Aedes spp. juveniles collected from 109 containers with larvae or pupae were identified to species (Fig. 5). Specimens of Ae. aegypti were identified in all juvenile mosquito samples collected from water storage containers, either alone or co-occurring with Ae. albopictus. Ae. albopictus was the most frequently identified species in samples collected from all other container classes except large refuse items. Specimens of Ae. marshallensis were rarely collected from artificial container habitats, but were present in almost half of samples from natural container habitats concurrently with Ae. albopictus. Culex spp. larvae and pupae specimens were present in 13 samples, with the greatest frequency in discarded tires (6 [27%] of 22 samples).