Dengue fever and Aedes aegypti risk in the Galápagos Islands, Ecuador

Introduction Dengue fever is an emerging infectious disease in the Galápagos Islands of Ecuador, with the first cases reported in 2002 and periodic outbreaks since then. Here we report the results of a pilot study conducted in two cities in 2014: Puerto Ayora (PA) on Santa Cruz Island, and Puerto Baquerizo Moreno (PB) on Santa Cristobal Island. The aims of this study were to assess the social-ecological risk factors associated with dengue and mosquito presence at the household-level. Methods In 2014 we conducted 100 household surveys (50 on each island) in neighborhoods with prior reported dengue. Adult mosquitoes were collected inside and outside the home, larval indices were determined through container surveys, and heads of households were interviewed to determine demographics, prior dengue infections, housing conditions, and knowledge, attitudes and practices regarding dengue. Multimodel selection methods were used to derive best-fit generalized linear regression (GLM) models of prior dengue infection, and the presence of Ae. aegypti in the home. Results We found that 24% of PB and 14% of PA respondents self-reported a prior dengue infection, and more PB homes than PA homes had Ae. aegypti. The top-ranked model for prior dengue infection included human movement – travel between neighborhoods, between islands, and to the mainland; demographics including salary level and education of the head of household, and increase with more people per room in a house, house condition, access to water quality issues, and dengue awareness. The top-ranked model for the presence of Ae. aegypti included housing conditions, including the presence of window screens and air conditioners, mosquito control actions, and dengue risk perception. Discussion/conclusion To our knowledge, this is the first study of dengue risk and Aedes aegypti in the Galápagos Islands. The findings that human movement within and between islands, and to and from the mainland, were important to reported dengue cases confirms concerns of this route of introduction and repeated transmission.

on the islands [12]; however, less attention has been paid to emerging pathogens in humans. 74 Prior studies of West Nile Virus risk on the Galápagos focused on the impacts on endemic bird 75 species [13][14][15]. However, emerging infections in humans are a growing concern due to the 76 increasingly urban resident population and the large number of international tourists. 77 Only three mosquito species have been reported to be present at the Galápagos islands, 78 namely Aedes aegypti, Aedes taeniorhynchus and Culex quinquefasciatus [16,17] The Ae. The primary means of preventing dengue transmission in Ecuador is through vector 112 control, reducing the density of Ae. aegypti in high-risk households, since a dengue vaccine is 113 not yet available for widespread use . Dengue control is conducted by the Ministry of Health 114 through repeated cycles of ultra-low volume fumigation of neighborhoods throughout the rainy 115 (peak transmission) season; indoor residual spraying in and around homes with suspected dengue 116 cases; routine visits to homes to apply larvicide (temefos/abate) to water-bearing containers, and 117 to destroy larval habitat in and around homes. The central vector control team for the Galápagos 118 is based in Puerto Ayora, due to the greater burden of disease, and a smaller team is based in the only sites in the Galápagos where cases of autochthonous dengue have been reported. 143 including type, use, source of water, and location inside or outside the home. All pupae and a 160 sample of larvae were reared to adults in the laboratory to confirm species identification. 161

Statistical models 162
Survey data were used to identify social-ecological variables associated with the 163 occurrence of self-reported dengue fever and vector presence. We hypothesized that both self-164 reported dengue and mosquito presence were associated with one or more of these factors (see and Attitudes were associated with the number of self-reported dengue cases on both islands 178 (Table 1). The second model selection process examined which survey factors related to Housing 179 Conditions, Knowledge and Attitudes, and Mosquito Abatement Practices were influencing the 180 presence of Ae. aegypti within the homes of survey participants (Table 1). 181 Some factors collected in the initial survey were excluded from model variable 182 candidates due to missing or uninformative data (i.e., the responses were identical across 183 households). Selection was run to convergence using glmulti's genetic algorithm (GA); models 184 were ranked using Akaike's Information Criterion (AIC) corrected for small sample size (AICc). 185 For each suite of variables in our hypotheses, a best model was obtained (Tables 3 and 4) Ae. aegypti per 100 homes) in PB were 20 and in PA were 6. Breteau Indices (number of 206 containers with juvenile Ae. aegypti per 100 homes) in PB were 26 and in PA were 6. A greater 207 proportion of surveyed containers were found with Ae. aegypti juveniles in PB than in PA 208 (p=0.019). The predominant characteristics of containers positive for juvenile Ae. aegypti (n=16) 209 were: low water tanks made of cement or plastic (92%); containers that were completely or 210 partially uncovered (92%); containers located outdoors (85%); containers that were shaded 1 1 (85%); containers filled with tap water as opposed to rain water (100%); and containers intended 212 for domestic use (i.e., used for cooking, cleaning, laundry as opposed to abandoned containers) 213 (77%). 214

Risk perceptions and practices 215
Most households on PA (82%) and PB (94%) reported that dengue was a serious problem 216 in their community (p = 0.1) and a severe disease (PA = 86%, PB = 98%, p = 0.06) ( Table 1). 217 Significantly more PB households reported that it was difficult or impossible to prevent dengue 218 (p = 0.02, Table 1). The majority of heads of households knew that dengue was transmitted by a 219 mosquito (94% on both islands), and most people had received information about dengue 220 prevention (64% on both islands). However, few people had participated in dengue prevention 221 campaigns (PA = 12%, PB = 10%). Sources of dengue information were similar between islands, 222 with media (TV, newspaper, radio) as the primary source of information on both islands, and 223 social networks as the least likely source of information (Table 1) The top-ranked model of a prior self-reported case of dengue in the household 234 (AICc=69.89, κ = 12.97) included the following suite of positively associated variables: the 235 number of people per room in the home, the head of the household earning more than minimum 236 wage, the head of the household with secondary education or higher, having a house in good 237 condition, household members who travel to the continent, household members who travel 238 between islands, frequent interruptions in the piped water supply, and being aware of dengue 239 cases in their community (Table 3, Figure 4). Having a female head of household, visiting other 240 neighborhoods daily, and no screens on doors and windows were negatively associated with self-241 reported dengue. Nineteen additional models were found within 2 AICc units of the top model 242 (Supplementary Table 3). 243 The top-ranked model to predict the presence of Ae. aegypti in households (AICc=63.00, 244 κ = 12.41) included the following positively associated variables: use of mosquito nets in the 245 home, bad patio condition, air conditioning, no screening on windows or doors, and the 246 perception that dengue is a difficult disease to prevent (Table 4, Figure 4). Negatively associated 247 variables included covering water containers, bad house condition, and the perception that 248 dengue is a problem. Eight additional models were found within 2 AICc units of the top model, 249 comprising alternating selections of similar variables to the top model (Supplementary Table 4). The important variables in the mosquito presence model indicate that housing conditions, 281 dengue risk perception, and prevention practices are important risk factors. We found that homes 282 were more likely to have Ae. aegypti if they had no screens on windows or doors, if they did not 283 cover water containers, and if they perceived that dengue was difficult to prevent ( Figure 4). 284 These risk factors are consistent with prior studies from Ecuador, Taiwan, and India [27,43,44]. 285 Interventions to address these factors include dengue awareness and community mobilization 286 campaigns [45], water container covers, and programs to provide low-cost screening to 287 homeowners. Paradoxically, we found that the use of mosquito nets was positively correlated 288 with Ae. aegypti presence. This may be as simple as a causal reverse in a correlation -bet nets 289 are more likely to be used when mosquitoes are perceptibly present. However, it may instead be 290 the case of the wrong intervention for the vector. Ae. aegypti have a small range and will bite 291 during the day (in contrast to other mosquito genera such as Anopheles), so bed nets may not be 292 an effective barrier to dengue transmission. Another unexpected finding was that air conditioning 293 was positively associated with Ae. aegypti presence. This result was counterintuitive, because 294 one would expect homes with air conditioning to have closed windows and subsequently fewer 295 mosquitoes and lower dengue risk, as shown in prior studies [46]. However, it is possible that 296 water buildup and puddles created by air conditioning units could create mosquito habitat, and 297 this requires further investigation, including understanding the context in which air conditioning 298 units are installed, versus used. If rooms are cooled for a few hours a day only, and in the 299 evening, windows are used to cool houses, an air conditioning unit is a false signal of closed 300 window behavior. A more comprehensive study should include AC practices in questionnaires. season, when this study was conducted. The primary focus in these areas should be the creation 325 of sealed (Aedes-proof) water storage containers used by households and located outdoors in the 326 patio. This result highlights the importance of integrated household water management strategies 327 in regions that are water-scarce and at risk of dengue risk. 328 We found that there were differences in self-reported prior dengue infections, vector 329 abundance, prevention strategies, sources of information, and risk perception between PA and 330 PB. This suggests that there may be differences in a number of factors between the islands, 331 including disease burden, community outreach programs, community awareness, and/or access 332 to information between the two sites. The high proportion of homes with juvenile Ae. aegypti on 333 PB indicate that there was significant risk of another dengue outbreak, even during the low 334 transmission season when this study was conducted. Islands. The findings that human movement within and between islands, and to and from the 390 mainland, were important to reported dengue cases, confirms concerns of this route of 391 introduction and repeated transmission. Bolstering surveillance of both tourism and cargo routes of entry would be useful to mitigate potential further introductions. The identification of both 393 sources of knowledge and perceptions, and our assessment of both the importance of knowledge 394 of prevention, and a pervasive lack of involvement in control campaigns, point to targets for 395 policy and action. We found that, similarly to studies conducted in mainland Ecuador, housing 396 condition, and water supply, access, and storage related behaviors are important. The water 397 connection is particularly poignant for the Galápagos, where access to freshwater is a concern. 398 Given the geographic challenges faced in distributed island vector management, this is a 399 complicated and unique setting for dengue management, but similarities with other studies, in 400 terms of targets for interventions identified in this study, provide useful information for 401 establishing combined outreach and direct intervention efforts in the public health arena. 402 403 Declarations 404

Ethics approval and consent to participate 405
The study protocol was reviewed and approved by the Institutional Review Board (IRB) of the 406 Universidad San Francisco de Quito and SUNY Upstate Medical University. The protocol was 407 also approved by the Agencia de Regulación y Control de la Bioseguridad y Cuarentena para 408 Galápagos (ABG) prior to study start. Heads of households (>18 years of age) were consented by 409 trained study technicians and signed an informed consent form prior to study start. 410

Consent for publication 411
Not applicable 412

Availability of data and material 413
The datasets used and/or analyzed during the current study are available from the corresponding 414 author on reasonable request.

Competing interests 416
The authors declare that they have no competing interests 417