Data collected from a preliminary study, carried out at the Universidad del Valle, in which two storm drains were evaluated, were taken into account for the design and development of this study. One storm drain was modified by placing a gravel filter bed to prevent water retention and the other remained unchanged [11]. Based on the results obtained, a pilot trial was then carried out.

In selecting the study site, we took into account data from the National Health Institute showing that 75 % of dengue cases in 2012 came from 10 departments of the country, among them Valle del Cauca with 7 % [12]. Moreover, this region has been the center of previous dengue epidemics [13]. The city’s health department (Secretaria de Salud Pública Municipal de Cali or SSPM) found that in Commune 6, which includes the study area, dengue and severe dengue cases have increased more than expected [10]. Floralia is a residential neighborhood in the north of the city, at approximately 3° 29′ 44.06″ N and 76° 29′ 37.36″ W (figure 1). This neighborhood has 353 storm drains distributed across its area.

The sample size was chosen assuming a 90 % reduction in the mean pupae per storm drain. Further assuming 20 larvae per unmodified storm drain, and a negative binomial distribution with dispersion parameter . of 0.2 [14], a total of 42 storm drains (21 per arm) gave 90 % power for a two-tailed significance level of 5 % [15].

The selection of storm drains was done by assigning a numerical value to each of them on a digital map of the area and allocating them at random, subject to a minimum separation of 70-90 m. Moreover, backup storm drains were identified in case, on visiting the selected location in the field, it did not exist or could not be included for other reasons. However, on one occasion, this backup protocol was not followed, and a replacement drain was selected by convenience, after finding the selected one to be completely sealed. Once identified in the field, the selected storm drains were identified and marked.

Figure 1. Figure 1. Location of the study area-Floralia neighborhood, northern Cali Source: Own elaboration

The modification of the 21 selected storm drains consisted in installing a filter bed in order to avoid accumulation of water in the chamber of the water seal, which works as a mechanism for control of odors of Volatile Organic Compounds (VOCs) released from the sewer. This modification should therefore eliminate the potential of the drain to be a breeding site of the dengue vector A. aegypti. The first step was to fix a millimeter ruler in all storm drains, whether modified (M) or unmodified (UM), to determine the water level. For the modification, the overflow screen was drilled at the base, making two rows of holes, each row separated by 10 mm. A total of 15 holes were perforated at a spacing of 30 mm. The diameter of the perforation (15.87 mm) was smaller than the gravel of the filter bed, since the 3/4" gravel at the bottom had an average diameter of 19.05 mm. Once the overflow screen was perforated, geo-textile and hollow bricks were installed at the bottom of the storm drain (figure 2).

Two types of gravel, with a 19.05 and a 9.5 mm diameter, were used. The first type of gravel was placed at the bottom half of the storm drains, according to the height of the overflow screen which varied between 0.50 and 1.20 m, depending on the storm drain. After this, a diamond extruded plastic mesh was installed to separate the next layer of gravel with 9.5 mm in diameter (3/4") and a height of 0.12 m. This process was done in order to provide a filter media and avoid the accumulation of water by allowing a continuous flow of water and at the same time retaining solids inside the storm drain (figure 2).

Before performing the physical modification of the 21 storm drains, all 42 selected storm drains were completely cleaned, removing floating solids (e. g., dry leaves, sticks, and solid waste), supernatant water, and accumulated sludge. The drains were then pressure-washed to remove excess grime and any eggs of A. aegypti or other insects adhering to the walls.

Figure 2. Figure 2. Storm drain modification process. (a) Screen perforation, (b) Installation of brick and gravel 9.5 mm diameter (3/4"), (c) Installation of the diamond mesh d. Modified storm drain Source: Own elaboration

The evaluation consisted of the following two components:

Presence of standing water in the storm drains

This evaluation was to be made after a strong rainfall event (rain intensity greater than 100 mm/hr with a duration greater than 15 min). This took place at the end of March 2013 and the evaluation was carried out in the first week of April 2013, approximately 8 days after the rain event, since full larval development takes 5-7 days in the pertaining temperatures [5], [6] and [16].

Entomological inspections were performed using a larval dipper with a volume of 70 cm³. Ten samples were extracted from each drain with water, for a total of 700 cm³/storm drain. Each sample was poured into a plastic tray to facilitate the identification of any mosquito larval stages. A plastic dropper was used to transfer them to Ziploc^® plastic bags containing 200 cm³ of 70 % ethyl alcohol. These bags were sealed and labelled with the date and number of the storm drain evaluated (figure 3).

Figure 3. Figure 3. Collection of larvae and pupae from the evaluated storm drains Source: Own elaboration

After sampling, the material was transported to the Entomology Laboratory of the Universidad del Valle, where it was quantified and identified to the species level following the taxonomic keys of Forattini [17] y Gonzales and Carrejo [8]. Finally, for each treatment, the Aedic Index (AI) was calculated, expressed as the number of storm drains positive for immature stages over the number of sampled storm drains (equation 1).

1

According to the sample size calculation, negative binomial regression was used to analyze the main response variable of larvae per storm drain. The residual deviance was used to assess the fit of the model.

Costs of materials, instruments, and labor per modified storm drain were measured and extrapolated to the total cost of modifying all the storm drains in the study area (n = 353). Maintenance costs were not included.

Of the 5 modified storm drains with water, two were positive for larvae of A. aegypti. In the first one, 38 individuals were found: 14 third stage larvae, 10 fourth stage larvae, and 14 pupae. In addition, 40 larvae of Culex quinquefasciatus were found. In the second storm drain, 43 A. aegypti larvae were found: 37 second stage and 6 third stage. The remaining 16 storm drains were dry (table 2). For this treatment, the AI was equal to 5 %.

Of the 21 unmodified (control) storm drains, 19 were found to have standing water. Of these, 13 were positive for immature stages of A. aegypti. In one of these, 380 individuals were found, representing 56.8  % of the total juveniles found in UM storm drains. For this case, all individuals were in the second larval stage. Densities below 100 individuals were found in the other storm drains. An ecological niche overlap between A. aegypti and Culex quinquefasciatus was observed, with 6 and 14 individuals present in a single storm drain, respectively. Of the two dry storm drains, one had accumulated a large amount of sand, flushed from nearby construction sites, and the other had been manually cleaned. For UM storm drains, the AI was 62 %.

Table 2. Table 2. A. aegypti found in modified and unmodified storm drains in Floralia Source: Own elaboration

The average number of larvae per storm drain, including those with zero, was 3.2 in the modified ones, one tenth of the value of the unmodified ones (31.9), that is, a ratio of 0.10 or a reduction of 90 %. Using negative binomial regression, the 95 % confidence interval of the 0.10 ratio is 0.014 to 0.74 (a reduction between 98.6 and 26 %), with a p-value of 0.016. The residual deviance was 29.7, which is less than the degrees of freedom of 40, indicating a good fit of the model. This ratio of 0.10 corresponds to the 9.1 % of larvae in the intervention arm shown in table 2: (67/669) / (1 + (67/669)). The 95 % confidence interval for this percentage is from 1.3 to 42.5 %. This indicates that this intervention reduced the proliferation of the dengue mosquito vector A. aegypti.

Based on information collected on the costs of materials, tools, and labor per storm drain, an estimate was made of the cost per inhabitant of Floralia (table 1) should the modification be implemented in all the storm drains in the neighborhood (n = 353), the result being USD 1.28 (table 3).

Table 3. Table 3. Cost estimate for storm drain modification (representative exchange rate of 2013) Source: Own elaboration