Study area
The survey was conducted in a 600 square meter transect at Bomani town, Msambweni location, Kwale County, Kenya. Bomani is an upcoming urban center with a population density of 958/square km [24]. The site is located 60 kilometers south of Mombasa city. The coastal climate is tropical hot and humid throughout the year with annual temperature of 23 ºC to 34 ºC and average relative humidity of 60–80%. The study area is approximately 2 kilometers from the Indian Ocean seashore and very close to the main hospital in Kwale County; Msambweni County referral hospital. The main activities of the people in the area are small scale farming and fishing. Residents in the study site store water for domestic use in various containers because tap water supply is not reliable. Water supply is mostly from wells, boreholes and harvested rain water, that supplement inadequate pipe water supply.
Larval Habitat Census
A larval habitat census was conducted from June 2 to June 17, 2017 to map and document all water receptacles within a 600 by 600 m area in Bomani town, Msambweni location, Kwale County. Potential larval habitats in outdoor domestic environment of every house located within the selected area were inspected for mosquito larvae and pupae. The larval habitats were classified into different habitat types as described by Ngugi and others [19]. All pupae and a sample of larvae (3rd and 4th instars) from positive larval habitats were collected with the aid of pipettes and ladles [25], counted and recorded on field-data forms. Water from large larval habitats was first sieved and mosquito samples were placed in a white plastic tray with some water from which the immatures were pipetted. Mosquito samples were placed in 10 ml falcon tubes and/or Whirl-pak® plastic bags (Nasco, Fort Atkinson, WI), labeled, and taken to the Vector Borne Disease Control Unit (VBDCU) laboratory at Msambweni County referral Hospital. Immature mosquitoes were reared in 200 ml plastic cups under laboratory conditions at an average temperature of 28.15 ± 1.8°C and relative humidity of 80.9 ± 6.3%, and larvae were fed on TetraMinbaby® fish food (Melle, Germany). Emerged adults were identified to species using standard taxonomic keys [26]. Ae. aegypti (L) subspecies were morphologically distinguished using keys by Edwards (1941) [26], Mattingly (1958) [27] and Huang (2004) [28].
Containers Management Practices And Characteristics
A total of seven container types were identified and classified based on their use and material: drums, tires, pots, small domestic containers (SDC), buckets, jerrycans and others [19]. Drums were defined as 100–500 litre capacity plastic or metal water storage containers. Pots included flower vases and water storage vessels made of clay. Small domestic containers included small plastic food containers, tins, bottles, plates, cans, cooking pots (sufuria) and jars. Others included polythene bags, fallen leaves, coconut shells, hoof prints, drains, gutters, septic tanks, shoes, cisterns, sinks and animal feeding containers (AFC). The AFCs, ranged from small 1 liter bird watering and feeding containers made of plastic or cut tires, to a medium 30 liter plastic container for watering cattle. For each breeding habitat, data was collected on the location within the outdoor domestic environment (Frontyard, backyard and others including bushes, gardens, dumpsites), container size or capacity (small: <25 liters; large: >25 liters), capable of being moved (movable; not movable), exposure to sunlight (fully shaded from sunlight; partially shaded from sunlight; fully exposed to sunlight), purpose of the water in the water storage containers (domestic uses; no purpose), evidence of covering (covered; not covered), water source, and frequency of water refilling.
Estimation of Ae. aegypti larval habitats productivity
A total of 83 representative habitats were randomly selected from the 664 potential larval habitats identified during the habitat census and marked with indelible ink for ease of identification. Daily productivity was estimated in the 83 selected representative larval habitats for 30 consecutive days. The selected larval habitats included 17 buckets, 9 drums, 9 jerrycans, 5 others, 8 pots, 16 SDCs and 19 tires. At every habitat daily for 30 days during the two sampling periods, quantification of the numbers of Ae. aegypti immatures was done following methods as described by Chadee and others and Ngugi others [19, 25]. The number of larvae and their stages of development was recorded for each habitat as well as the number of pupae. All pupae were removed and allowed to emerge in the laboratory. All emerged adults were identified to sub-species level by morphological features [26]. During each sampling visit, records were made on the location within the outdoor domestic environment (Frontyard, backyard and others including bushes, gardens, dumpsites), container size or capacity (small: <25 liters; large: >25 liters), capable of being moved (movable; not movable), exposure to sunlight (fully shaded from sunlight; partially shaded from sunlight or fullly exposed to sunlight), purposes of the water in the water storage containers (domestic uses; no purposes) evidence of covering (covered; not covered), water source and frequency of water refilling or emptied.
Data analysis
Data analysis was conducted independently for the 3 datasets in this study; the baseline survey, the wet and dry season longitudinal datasets. Descriptive analyses were used to explore the data. The number of pupae per habitat was the output variable for the baseline survey and for both the wet and dry season longitudinal surveys. Pearson dispersion statistic was used to assess for over-dispersion for all the 3 datasets. All three datasets were highly over-dispersed with 86% zero counts in the baseline survey and over 90% zero counts in both the wet and dry season surveys. A zero-inflated negative binominal regression (ZINB) was therefore considered appropriate to test association with Ae. aegypti pupae infestation [29]. The ZINB model fits a negative binomial component and a zero inflated component, since we were interested with the association with Ae. aegypti pupae infestation estimated by the negative binomial component, only its regression coefficients transformed into incidence rate ratios (IRR) were reported. Before being included in the ZINB models, correlations among the predictors were assessed using Spearman’s Rank correlation coefficient and those with strong correction (> 0.7) excluded in the analysis. For the baseline survey, factors assessed for association with risk of Ae. aegypti pupae infestation included larval habitat type (abbreviated as ‘Habitat type’), location within the outdoor domestic environment (abbreviated as ‘place’), container size, capable of being moved (abbreviated as ‘move’), exposure to sunlight (abbreviated as ‘shade’), purpose of the water in the water storage containers (abbreviated as ‘water purpose’), evidence of covering (abbreviated as ‘cover’), water source and frequency of water refilling (abbreviated as ‘filled’). All these factors were included in univariate ZINB regression models and those with P-values < 0.1, were included in the multivariate ZINB regression model.
For the wet and dry season datasets, we assessed the daily pupal productivity of the larval habitats by running both univariate and multivariate ZINB models controlling for the multiple daily observations using robust standard errors. A similar approach to the one used in baseline survey above, was used to fit multivariate ZINB models for wet and dry season. The pupal productivity predictors were habitat type, place, container size, shade, water purpose, cover, habitat filled, water source and habitat stability. The predictor ‘Move’ was not included in this analysis because no pupae were recorded from the unmovable larval habitats. Habitat stability was defined as the number of days a habitat contained water during the 30-day sampling period. Habitats were classified as stable if they had water for at least 7 days in wet season and 20 days for the dry season. The habitat stability cutoffs were selected to maximize the number of habitats within each category and varied among the wet and dry seasons. Statistical analyses were carried out using SAS software version 9.3 (SAS Institute Inc., Cary, North Carolina, USA) and Stata version 15.1 (StataCorp, College Station, TX, USA).