Associations between meteorological variables, vector indices and dengue hospitalizations in Can Tho, Vietnam: a field survey

Introduction Dengue is a significant cause of morbidity and mortality in Can Tho, a province in the Mekong Delta in Vietnam. In this region, average temperatures have increased by 0.5°C since 1980, and river levels have risen. In a time-series analysis, we previously found that relative humidity was the most important meteorological predictor for dengue hospitalizations in Can Tho. To better understand proximate factors mediating this association, this study examines weather variables in relation to dengue hospitalization rates, vector indices, container productivity and larval elimination and mosquito avoidance behaviors. Methods Four hundred households were sampled bimonthly for one year in Can Tho. Vector indices of the immature forms of the dengue vector, Aedes aegypti, and the productivity of different types of household containers were determined. Dengue hospitalization rates were determined for the study period. Associations between these variables and mean temperature, relative humidity, precipitation, and the number of hours of sun were estimated using mixed effects Poisson regression analysis. Relative productivity of containers was determined by collecting Ae. aegypti pupae using a sweep method and adjusting by a calibration factor. Ae. aegypti larval density risk factors were determined using multivariate generalized estimating equations with a negative binomial distribution. To examine possible mechanisms mediating the relationship between climate, vectors and dengue, we also interviewed households about mosquito avoidance and larval elimination behaviors. Results The house-(HI), container-(CI), Breteau (BI), and pupal (PI) indices were associated with relative humidity (1-month lag, IRRHI=1.10 (95% CI 1.06, 1.13) per 1% increase), IRRCI=1.10 (95% CI 1.02, 1.19), IRRBI=1.17 (95% CI 1.14, 1.21), IRRPI=1.12 (95% CI 1.10, 1.14)). Vector indices were also associated with precipitation (1-month lag) and to a lesser degree, hours of sun and mean temperature. Ae. aegypti larval density was associated with not cleaning water storage containers (RR=2.50, 95% CI 1.59, 3.66), not having access to municipal waste pick-up (RR=3.15, 95% CI2.09, 4.75), disheveled clothes in the home (RR=1.85, 95% CI 1.24, 2.74) and season (RR[rainy season]=3.10, 95% CI 2.18-4.48). The most productive containers were water storage containers (relative pupal productivity 87%). Dengue hospitalization rates were associated with relative humidity (2-month lag, IRR=1.11 (95% CI 1.06, 1.17) per 1% increase). Only the PI (1-month lag) was significantly associated with dengue hospitalization rates (IRR 1.04, 95% CI 1.00, 1.07). Mosquito avoidance behaviors were more frequent in the dry season (92.5% vs. 86.0% of interviewees endorsed one or more forms of mosquito prevention, p<0.001). There was also less use of larval elimination strategies (39.2% vs. 50.5%, p<0.001) during the rainy versus the dry season. Conclusion Our study reveals a strong effect of relative humidity on vector indices and dengue hospitalization rates. This may be due to the mosquito’s vulnerability to desiccation, and the association warrants further study. Our findings also demonstrate, however, that during the rainy season when mosquito prevention is most needed, the use of fans, repellant coils and maintenance of water storage containers is actually reduced. Water storage containers were by far the most productive of pupae, and should be targeted in vector control activities. Author summary Climate plays an important role in the geographic distribution and burden of disease due to dengue, owing to the vector and virus’ sensitivity to temperature, humidity, and rainfall. In the Mekong Delta in Vietnam, where dengue poses a significant health burden, average temperatures have increased by 0.5°C since 1980. To better understand the influence of climate on dengue, this study examines its influence on dengue hospitalization rates, vector breeding behavior and human mosquito avoidance behaviors. We sampled 400 households every 2 months for one year for the presence of the dengue vector, Aedes aegypti, and the productivity of different types of household containers. Human mosquito avoidance behaviors, such as the use of fans, mosquito repellant, and larval elimination strategies were also recorded. The association between dengue hospitalizations, mean temperature, relative humidity, precipitation, and the number of hours of sun were established, and risk factors for the abundance of Ae. aegypti larvae were determined. We found that relative humidity is positively associated with the presence of Ae. aegypti immature forms, and that large jars used for water storage serve as the most important source of this vector. We also determined that people engage in mosquito avoidance/larval elimination strategies more frequently in the dry season versus the rainy season, despite increased vector breeding and dengue hospitalizations during the rainy season. This temporal disconnect between peak vector activity and dengue hospitalization rates vis-à-vis mosquito control strategies is a potential area for intervention.


20
Dengue is a significant cause of morbidity and mortality in Can Tho, a province in the Mekong Delta 21 in Vietnam. In this region, average temperatures have increased by 0.5°C since 1980, and river 22 levels have risen. In a time-series analysis, we previously found that relative humidity was the most

41
Results Previously, we reported on the associations between dengue hospitalizations in this region and 118 climate between 2004 and 2011 in a time-series analysis [7]. We found that the dengue 119 hospitalization rate in Can Tho was significantly associated with relative humidity with a lag of one 120 month. To better understand the entomological and behavioral factors that may be contributing to 121 this association, we conducted this prospective study. Specifically, we analyze indices of immature 122 forms of Ae. aegypti, dengue hospitalization rates, container productivity, and mosquito 123 avoidance/larval elimination behaviors by weather variables.

131
City of Can Tho (10.0333 (Fig 1). The former lies in the heart of Can Tho city and is urban, while the latter is suburban/rural.

205
Risk factors for Ae. aegypti larval density were modeled using generalized estimating

209
Larval and pupal indices and counts were used to determine container productivity. Since 210 jars and cement tanks hold large volumes of water, the absolute count of larvae and pupae 211 collected in a sweep in these containers was adjusted by a calibration factor (C-2Fs) assuming that 212 water level is at two thirds the total capacity, as described in [13].

266
To examine the association between vector indices and meteorological factors in more 267 detail, mixed effects Poisson regression coefficients were estimated for 0-, 1-, and 2-month lags

268
( Table 2). The most pronounced associations were seen with relative humidity. Relative humidity 269 was positively associated with virtually all vector indices at all lags. Precipitation was also strongly 270 associated with all indices at 1-month lag, and variably associated at 0-and 2-month lags. The

271
monthly hours of sunshine were negative associated with HI, BI, and PI at 1-month lag.

272
Temperature was significantly correlated only with PI, and most strongly at a lag of 2-months. PI   (Table 4a). In the dry season, 6,291 containers were 310 sampled (Table 4b). A total of 1,688 Ae.aegypti pupae were collected in the rainy season, with 311 dramatically less in the dry season (263 pupae). Since water storage containers and cement tanks 312 hold large volumes of water, the absolute count of pupae collected in a sweep was adjusted by a 313 calibration factor (C-2Fs) assuming that water level is at two-thirds the total capacity[13]. In both 314 seasons, the most productive containers were the water storage containers with a relative 315 productivity of 87% and 88% in the rainy and dry seasons, respectively. This was followed by 316 cement tanks (relative productivity 10% and 7% in the rainy and dry seasons). Other containers 317 such as buckets, vases, and miscellaneous other containers were common, but none of these 318 yielded many larvae nor pupae relative to the jars. The dengue hospitalization rate varied by ward and by month (Figure 4)

336
The dry season had a mean monthly hospitalization rate of 8.5 (95%CI 5.0, 12.1), whereas the 337 rainy season had a mean of 12.8 (95%CI 8.6, 17.1) (p=0.13). Dengue rates were significantly 338 associated with most of the meteorological variables at 0-, 1-and 2-month lags (Table 5a). Models 339 with rainfall and relative humidity with a lag of 2-months had the lowest AIC. The only significant 340 association between dengue hospitalization rates and vector indices was observed for PI with a 2-341 month lag (Table 5b).  Mosquito avoidance behaviors by season

356
The percentage of people endorsing the use of mosquito avoidance behaviors was higher in the dry 357 season (92.5%) than the rainy season (86.0%) ( Table 6). The most commonly used method was  . Some strategies to address these priorities may provide multiple health benefits,

464
such as improving water supply and infrastructure, such that water storage will no longer be 465 necessary. Prospectively measuring the impacts of such mitigation efforts on vector-borne disease 466 indicators will provide valuable insight into the full extent of benefits conferred.

467
In conclusion, our study sought to link dengue and weather by examining multiple levels 468 within the chain of causation, namely vector indices, pupal productivity, dengue hospitalization rates 469 and mosquito avoidance and elimination measures. Our results indicate that relative humidity is a 470 key weather variable in this area where temperatures are consistently within an optimal range for 471 dengue transmission. We also found that large water storage containers are the source of the 472 majority of Ae. aegypti pupae, and that these containers are maintained less frequently during the 473 rainy season. Climate change projections forecast rising temperatures and flooding in this region of

474
Vietnam. This will likely render this region vulnerable to water shortages, leading to more reliance on storing water near the domicile. Further studies are warranted on how these factors will influence 476 not only dengue but also other the transmission risk of other arboviruses vectored by Ae. aegypti.