Study design and sample size

The study followed a cluster randomised trial design with cross-sectional entomological surveys. A cluster was defined as a geographical area that included at least 100 private households, but also non-residential buildings and public spaces. A buffer zone of 100 metres between clusters was assured to prevent spillover effects.

The sample size was calculated as required for cluster randomised intervention studies¹³ using 10 clusters with 100 households for each study arm taking into account the following assumptions reported in previous studies:^4,14 mean numbers of pupae per person pre- and post-intervention in treated clusters were assumed to be 3.0 and 0.3, respectively;⁴ negative binomial distribution considering an overdispersion of the dependent variable; dispersion coefficient of 0.02; and intra-cluster coefficient of 0.05. This reflected a desired power of 80% with significance level set at 5%.

Selection, pairing and random allocation of clusters to study groups

To obtain a sample of 20 clusters, a map of Girardot was overlaid with a grid containing 200 squares as described by Quintero et al.¹⁵ Simple random numbers from Epi Info™ 2000 software (CDC, Atlanta, GA, USA) were used to select 20 squares, from the total of 196 identified (Figure 1).

Open in a separate window

Figure 1.

Map of study clusters of Girardot, Colombia, with cluster number and area. The figure shows selected clusters included in the study, they are listed from 1 to 20 for easy identification. Black squares represent intervened clusters with LLIN, grey ones represent control groups. The grey shade inside squares shows location of 100 households corresponding to each cluster.

Before random allocation of clusters to study groups (10 to control and 10 to intervention) a two-stage cluster analysis was used to define homogeneity between clusters regarding the following variables: entomological indices in wet and dry seasons, cluster socio-economic stratum, pupae per person index (PPI), Breteau index (BI), number of containers per household, size of cluster (in hectares) and inhabitants per cluster. Two groups of clusters were defined: one of 16 and the other of 4. Clusters were randomly allocated within the two groups based on a public raffle. Half (8) of the 16 clusters in one stratum and half (2) in the other were randomly distributed by raffle in the two study arms.

Interventions

Situational analysis of the first phase of the project¹⁵ showed that low concrete water containers (‘albercas’) located in the backyards of houses were the most permanent productive container type for vector development. Stored water is used for washing, house cleaning and personal hygiene, and water shortages and cost savings are the main reasons for storage (García-Betancourt T, Higuera D, González-Uribe C et al., manuscript submitted).

PermaNet 2.0^® (long-lasting insecticide net treated with deltamethrin 50 mg/m², Vestergaard-Frandsen, Lausanne, Switzerland) used as curtains on windows and doors as well as LLIN covers for household water container covers were selected as suitable intervention tools based on literature;^4–6,16,17 natural populations of Aedes aegypti in Girardot being susceptible to pyrethroids,¹⁸ an insecticide authorised by public health authorities; and water containers (albercas or plastic containers of >200 L)^10,19 with no protection against mosquito oviposture being the main productive sources of Aedes aegypti pupae (70% of all pupae were found in these containers) (Alcalá L, Quintero J, González-Uribe C et al., manuscript submitted). In addition, involvement of the community in all stages of the intervention encouraged acceptance of the project²⁰ and represented an income-generating activity for local people.

The interventions were applied at the household level (Table 1). Houses received: curtains for windows and doors: 2852 curtains (about three per household) and 947 door curtains (one per house) were sewn by 10 seamstress from white bed-nets of PermaNet 2.0, WHOPES approved²¹ (Figure 2A) and covers for water containers of >200 L (Figures 2B and ​and2C).2C). Two types of water containers were designed through workshops within the community as described in Garcia-Betancourt T et al.²² and manufactured by 4 microenterprises and 10 seamstress. The design depended on the dimensions and shapes of the water containers. For cylindrical tanks, PermaNet 2.0 circular lids were made by seamstress with an elastic band and waterproof material for fixing. Square and rectangular covers consisted of aluminium frames with a sliding mechanism fixed to PermaNet 2.0 and were made and installed by small to medium enterprises (SMEs).

Open in a separate window

Figure 2.

Curtains in windows, rectangular container cover and circular container cover. The Figure shows three intervention tools. (A) a curtain inside a household in Girardot. (B) the rectangular water container cover for rectangular or square cement containers of at least 200 L capacity; the cover is made of an aluminum frame, LLIN netting and a sliding mechanism for opening both doors. (C) the circular cover made of rubric, LLIN netting and rubber, for plastic and cement cylindrical containers that can store more than 200 L. This figure is available in black and white in print and in colour at Transactions online.

First, curtains were installed immediately after baseline entomological surveys by the community in 922 intervention houses from July to August 2013. They were fixed by homeowners with a white string and two nails beneath existing curtains (in some homes) and usually supervised by research staff. At least one door per house and all unprotected and in-use windows were covered. However, it was not possible to cover the openings above windows if present (Figure 2A).

The second intervention, water container covers, were implemented in the same houses that were previously allocated to curtains, 8 months after baseline surveys. A total of 354 lids (1.2 per household) in 4 clusters representing 303 households out of 385 have been installed so far from October to March 2014.

In addition, during 2013 the vector control programme continued to deliver the routine interventions regarding larvae control through larvicides (Abate, American Cyanamid Co., Princeton, NJ, USA), health education, and occasional public space spraying of an ultra-low volume of Malathion (Southern Agricultural Insecticides Inc., Palmetto, FL, USA) in both control and treatment clusters.

Data collection

Effectiveness of the first and second interventions: impact on entomological indices

The first intervention with LLIN curtains alone (with incomplete coverage of holes in the walls) (Figure 2) showed a reduction in entomological indices in intervention clusters compared to controls); the differences were statistically significant for BI only (p=0) (Table 3). BI fell from 14 to 6 (57%) in the intervention group and from 8 to 5 (38%) in the control group. PPI increased in 17% of the intervention houses but fell by 22% in control houses. The additional intervention with LLIN covers for water containers showed a significant reduction in vector densities measured through the PPI (p=0.01) as shown in Table 3. In the intervention group, the PPI showed a decline of 71% (from 0.75 at baseline to 0.22 at second follow up), compared with 25% (from 0.40 to 0.30) in the control group (Table 2 and Table 3).

The majority of containers that produced >78% of pupae were containers >200 L; after the intervention with LLIN covers the pupae productivity decreased from 970 to 388 (60%) and in the control group from 394 to 339 (16%). The other container types were responsible for <12% pupal production, especially small buckets (Table 2).

Coverage and satisfaction towards intervention

Out of the 947 eligible households, 97.4% (922) households received 3138 window and door curtains (2216 for windows and 922 for doors). After 9 weeks, 74.1% (577/779 of houses with available data) of the households were still using at least one curtain; 74.9% (432/577) of residents reported a willingness to pay and 89.6% (517/577) of householders would recommend them to friends and neighbours. A total of 78.0% (450/577) of the residents reported good quality of the material and 32.1% (185/577) rated the quality of the curtains 5 (excellent) on a 1 to 5 scale. After two follow-ups (29 weeks), the percentage of use decreased to 45.2% (417/922) because the monitoring period coincided with the Christmas holidays and new year seasons, when 34.8% (77/221) inhabitants removed curtains in order to decorate and paint their houses; 11.7% (26/221) washed them and 6.3% (14/221) traveled or migrated to other places.

Out of 385 eligible households, 78.7% (303) households received 354 water container covers for 303 rectangular and 51 circular containers. After using them, 60.1% (182/303) of residents reported a willingness to pay for the covers, 83.2% (252/303) would recommend them to friends and neighbours, and 89.8% (272/303) would be happy to receive them again if they were free; 26.4% (80/303) reported less use of do-it-yourself insecticide sprays and indicated that the median cost of the sprays was US$8.

The 10 FGD showed that participants were impressed by the number and diversity of dead insects (mosquitoes, cockroaches, domestic flies, grasshoppers and other domestic pests) below curtains; they would all recommend the interventions to others. Mild cutaneous rash was reported by the seamstresses who manipulated the LLIN material without using gloves or other protective gear (2/10 tailors). This effect was also reported by 2.7% of residents (21/779) who received curtains.

Feasibility

Our study did show that the intervention package was effective in reducing PPI, but the feasibility of any control programme does require additional elements such as people's, satisfaction with the control services, and affordable costs.

Acceptance was achieved by the use of participatory strategies reported by García-Betancourt et al.²² Partnerships were established among local authorities, vector control services and the community. Satisfaction towards intervention was enhanced as people observed the dead insects below the curtains, which convinced them of its efficacy. This is similar to the findings reported in Haití,¹⁷ Venezuela⁴ and the Guatemala study.⁶ Yet in spite of people's positive perceptions, after some time they forgot about the benefits of the interventions and stopped using them: this means that there is a need for continued encouragement and monitoring of the intervention programme to sustain LLIN coverage which has a coverage-dependent effect on vector densities, as reported by Vanlerberghe et al.¹⁰

Based on the preliminary results of the intervention, vector control staff and local policy makers were very much in favour of continuing the novel vector control tool. National decision makers expressed an interest in financing the extension of the study to larger parts of the city and measuring whether reductions in dengue incidence could be achieved.

As a general rule, community-based interventions require more time and resources than conventional institution-based interventions because of longer socialisation and negotiation processes that are needed to achieve social participation and to respond to community expectations. The median costs of the intervention package in Girardot were US$48 per household: considerable if compared with similar interventions implemented in other countries. Rizzo et al.⁶ reported a median cost of US$5.31 per household for the same intervention package in Guatemala, excluding the costs of the netting materials that were donated. Baly et al.²⁹ estimated median costs of US$8.84 and US$6.68 per household cover with LLINs in Venezuela and Thailand, respectively, also excluding the costs of the LLIN materials. However, the cost savings by families where dengue cases have been averted are also substantial. A recent study showed that households have out-of-pocket expenditure for dengue prevention of US$13.27 but an indirect cost of US$197.10 for dengue outpatients.³ Thus, the median costs of the intervention in our study are high, but when compared to local out-of-pocket expenditures or indirect costs, the investment in the described intervention package seems to be warranted.

Limitations

Our impact evaluation is based on a short observation period; we do not know for how long after a 5 and 12 month observation period the effect will be sustainable. The efficacy, along with other vector control tools, will depend on multiple factors such as appropriate actions of social mobilisation to achieve long lasting behaviours, positive perceptions, perceived effectiveness, durability of materials used for interventions, coverage attained and local contextual and environmental conditions. However, the preliminary results are encouraging for policy makers.

Baseline and follow-up surveys took longer than initially planned, spreading over 3 months, as some houses were only occupied during holiday seasons, therefore field workers spent extra time revisiting empty houses to ensure participation. This could have biased our results as infestation rates can change according to season. However, our previous study¹⁵ demonstrated that pupal productivity is independent of rainfall, as no significant differences were found between seasons. Indoor low water container covers alone produce more than 70% of Aedes pupae during the wet and dry season and there are very few productive water containers outdoors.

Working with local seamstresses and SMEs that produced the aluminium frames for covering water tanks was beneficial. However, in the case of the container covers, SMEs were not prepared for such a high workload and resource utilisation. Therefore, frequent interruptions between measuring water container size and installation of covers were the rule.

Authors' contributions: JQ and GC conceived the study; JQ and GC designed the study protocol; TGB, DG and LA coordinated field work and data recollection; JQ, CGU, SC, TGB, LA and DG analysed and contributed to data interpretation; JQ drafted the manuscript; HB advised the development and analysis of entomological components; CGU, GC and HB critically revised the manuscript for intellectual content. All authors read and approved the final manuscript. GC is guarantor of the paper.

Acknowledgments: The authors thank local leaders, stakeholders and field workers, especially Sandra Marta Lozano, the field supervisor, without whom this project would not have been possible; the research team of Ecosalud ETV Colombia in Bogota, who supported technical and field activities; the health authorities of Girardot, particularly Dr Magda Caicedo, for supporting the fieldwork and other activities of the project; the residents of Girardot for welcoming our team; and to Eduardo Alfonso who coordinated the cost analysis.

The Special Programme for Research and Training in Tropical Diseases (TDR) at the World Health Organization, in collaboration with its Regional Office for the Americas (PAHO), formed a partnership with the Ecosystem and Human Health Program of the International Development Research Centre (IDRC) of Canada to develop the research programme under the title ‘Towards Improved Dengue and Chagas Disease Control through Innovative Ecosystem Management and Community-Directed Interventions: An Eco-Bio-Social Research Programme on Dengue and Chagas Disease in Latin America and the Caribbean’ [IDRC grant Project Number 104951-001].

Funding: This investigation received financial support from the UNICEF/UNDP/World Bank/WHO Special Programme for Research and Training in Tropical Diseases (TDR) with the aid of a grant from the International Development Research Centre (IDRC), Ottawa, Canada.

Competing interests: None declared.

Ethical approval: This study received clearance from the Ethical Review Committee of Fundación Santa Fe de Bogotá and WHO's Research Ethics Review Committee (ERC) under protocol ID: A90296. Each householder approved the intervention and signed a written informed consent form. The LLIN products are approved by the WHO Pesticide Evaluation Scheme (WHOPES) and by the Colombian Ministry of Health.