Testing the Effectiveness of Community-Based Dengue Vector Control Interventions Using Semiparametric Mixed Models

Abstract

Objective:

To evaluate the results obtained though intersectoral coordination and community empowerment in one study carried out during 6 years in Playa Municipality, Cuba.

Methods:

A longitudinal assessment comparing one intervention and one control area was conducted. The intervention encompasses two main stages separated by two dengue outbreaks. The first stage, focused on strengthening intersectoral coordination, was initiated in January 2000. In August 2003, a complementary strategy, focused on community empowerment and was initiated in half of the intervention area. In the control area, routine dengue control activities continued without additional input. We used entomologic surveillance data from January 1999 to December 2005 to assess the effectiveness. We computed the Breteau index (BI) per health area and the 95% confidence interval for the difference between the BIs at each time point. A semiparametric mixed model to capture the evolution in time of Aedes aegypti larval densities was fitted.

Results:

The BI in the control area showed the lowest value before starting the intervention. This was reversed 1 year after launching intersectoral activities for dengue control in the intervention area. In spite of spraying actions in all areas, the differences in BI between intervention and control areas remain significant until December 2002. Although for the next 2 years no differences were observed, they become significant again in December 2004, which corresponds with the implementation of the complementary community-based vector control strategy in the intervention area.

Conclusions:

The model fitted identified monotonous trends over time and reversal trends at particular moments. The confidence bands indicate sections with significant differences between areas. Our data increase the evidence that the intersectoral coordination and community empowerment strategy for A. aegypti control is effective.

Introduction

D engue fever and dengue hemorrhagic fever (DHF) are major public health problems with almost one half of the global human population at risk (Farrar et al. 2007). Transmission is increasingly endemic, with epidemic outbreaks in the Americas, Southeast Asia, and the Pacific (Gubler 2005). Each year, an estimated 50 to 100 million new dengue cases occurs globally; of them, 500 000 are DHF, and 25 000 are fatal (Guzman and Kouri 2002).

There is no specific management for dengue, as a vaccine is not yet available. Prevention is restricted to controlling the main vector Aedes aegypti, the main vector in tropical and subtropical regions. Reduction of the mosquito population can be achieved by making water-holding containers unsuitable for breeding or larval development and by adulticiding with insecticides (PAHO 1994). Although with strategies based on insecticide application some successes have been achieved in controlling dengue, these are not sustainable, because they are too costly (Gubler and Clark 1996). Thus, the vector control strategies in recent decades are essentially based on environmental management, increasingly with community involvement and intersectoral action (Kay 1994, Parks et al. 2004, Sanchez et al. 2005). However, Heintze et al. (2007a), in a systematic review of the achievement of community-based dengue control programs, conclude that there is weak evidence for community effectiveness in dengue control. This may be due to lack of well-designed studies or because community-based interventions deliver less than expected (Heintze et al. 2007b). Erlanger et al. (2008), in a systematic review of the effect of various dengue vector control intervention on entomological parameters in developing countries, report that most of the studies demonstrated a significant effect on the entomological index, but of widely varying magnitude.

Earlier, we reported 6 years' experience with an intervention study in Playa municipality, Havana City, and the result obtained though intersectoral coordination and a community empowerment approach to dengue control (Perez et al. 2007, Sanchez et al. 2009). We found that the process indicators, community outcomes, and entomological data consistently reflect a trend to increase in effectiveness from the implementation of the intersectoral intervention alone to community empowerment and intersectoral coordination. In our previous analysis, analysis of variance models that considered time periods according to the intervention strategies were employed. Thus, a continuous time scale was not used, whereas an analysis which carefully models captures the evolution over time of A. aegypti larval densities would be of particular interest.

In this article, we present an application of the semiparametric mixed model (Ruppert et al. 2003) to capture the form of the evolution in time of A. aegypti larval densities in three areas of Playa Municipality in Cuba. Identifying a suitable parametric model (e.g., linear, quadratic etc.) to appropriately describe the evolution of of A. aegypti larval densities with time may not be straightforward. A more flexible approach is to use the semiparametric mixed model approach that allows nonlinear relationships to be handled without restrictions of parametric models. In this approach, a set of knot points on the time axis is selected, and these act as “pivots” where small pieces of the model are joined. If an adequate number of knots is identified, then treating the knot points as random, and specifically, from a normal distribution with mean 0 and a certain variance, will result in a smooth fit of the A. aegypti larval densities (see Ruppert et al. 2003 for details). As pointed out by Ruppert et al. (2003) and Maringwa et al. (2008a, 2008b), the advantage of using a semiparametric mixed model is that we do not need to specify the mean structure that describes the dependency of the BI on time in advance but instead we use a flexible and data-driven approach to model this relationship. We examined the effectiveness on BI, an entomological index, of routine vector control, and the effect of adding intersectoral coordination and community empowerment approaches to dengue prevention estimate the delay between the start of the intervention and statistically significant changes in BI. Applications of this nature have appeared in literature, for example, Verbyla et al. (1999), Ruppert et al. (2003), Ngo and Wand (2004), and Durban et al. (2005). Here, we closely follow the approach in Maringwa et al. (2008a).

Materials and Methods

Study site

The study was carried out in Playa Municipality, in the north-west of Havana City (Fig. 1). Among the nine Consejos Populares* (CP) of Playa Municipality, we selected CP4, the one with the highest A. aegypti infestation levels, for a pilot intervention started in September 2000. This area measures 4.5 km², has 27,030 inhabitants, and is served by the Polyclinic ‘26 de Julio’. We chose CP5, with an area of 4.0 km² and a population of 14,219, as the control area. It has similar socio-ecological conditions. In August 2002, after the dengue outbreak, the municipal health authorities decided to extend the intervention to other fours areas. We selected CP6, with an area of 4.3 km², 16,096 inhabitants and served by the Polyclinic ‘1ero de Enero’ to assess the results of this extension strategy.

FIG. 1.

Study area. Playa Municipality, Havana, 1999–2005.

Interventions

Routine control program

Standard A. aegypti control actions comprising entomological surveillance, source reduction through periodic house inspection, periodical application of larvicides, health education, and enforcement of mosquito control legislation through the use of fines were carried out by vector control technicians. During the dengue outbreaks that occurred in 2000 and 2001, vector control actions were augmented with outdoor spraying in a radius of 100 m² around houses of confirmed cases. In addition, the frequency of inspection cycles was changed from bimonthly to monthly. In January 2002, an “Intensive Campaign” was launched. It involved people at all levels, from the Head of State to households. Massive environmental management, sanitation efforts, and the periodic elimination of adult mosquitoes were carried out. The frequency of inspection cycles was reduced to biweekly. The “Intensive Campaign” lasted until March 2002, and routine A. aegypti control actions continued to the end of the study period.

Intervention area

The intervention in CP4 encompasses two main stages separated by the dengue outbreaks. In the first stage, the aim of the intervention was to strengthen intersectoral coordination at CP level (Sanchez et al. 2005). The intervention was focused on changing the concept of the existing intersectoral management from a technocratic toward a participatory approach that involves the community in all phases of planning and implementation. Social scientists introduced participatory methods to facilitate dialog in the Health Council. This council subsequently developed an intersectoral plan for dengue prevention. A. aegypti control methods included elimination of useless containers, covering of water tanks, and cleaning public areas. In the second stage, started in August 2003, a complementary strategy that focused on community empowerment was initiated (Sanchez et al. 2009). This phase eventually involved almost half (5 out of 12) of the circunscripciones in the intervention area. The family doctors involved as leaders started the implementation of the proposed strategy once they felt ready to put the acquired knowledge and skill into practice. Due to this, the incorporation of circunscripciones was gradual: one in August 2003 and two in March and August 2004. Community working groups at the circunscripción level, composed of formal and informal leaders from existing community organizations, were motivated to work on dengue control and take responsibility for managing vector control activities in their areas. The community empowerment intervention targeted five participatory processes: capacity building, community dengue surveillance, social communications, behavioral change, and participatory evaluation. The main practices promoted at individual and collective level were scrubbing water tanks, repair and correct covering of water tanks, protection of useful containers from filling with rainwater, and the elimination of useless containers.

Extension area

After the 2002 dengue outbreak, the municipal health authorities organized short courses for family doctors and epidemiologists in the CP6. Topics such as dengue prevention, intersectoral coordination, community participation, and strategic planning were emphasized. After the training, family doctors carried out a community diagnosis and developed intersectoral plans that included sanitation activities and the delivery of locally developed informative and educational materials on eliminating useless containers and covering water tanks.

Control area

In the control area, the routine dengue control program continued to be carried out without interference.

Entomologic information

We used entomologic surveillance data from January 1999 to December 2005 that were recorded by the National Vector Control Program in the pilot, extension, and control areas. As elsewhere in Cuba, vector control technicians inspected every house for aquatic stage of A. aegypti. Each water container was recorded and examined for mosquito larvae and pupae. When positive, a sample was taken and sent to the municipal laboratory to identify the species. The system was subject to quality control by municipal supervisors, who revisited one-third of the households. Data collected in all inspection cycles were compiled by the epidemiologist at the municipality. Information on the number of inspected premise and positive containers (with A. aegypti larvae or pupae) was extracted. We computed the Breteau index (BI, number of positive container per 100 premises) per health area.

Data analysis and model description

Let g_A (t) and g_B (t) be the expected values of BI in area A and B at time t, respectively. We wish to test the hypothesis that the BIs in the two areas are equal to a two-sided alternative that the BIs are different. We formulate the hypotheses: \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland,xspace}\usepackage{amsmath,amsxtra}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} \begin{align*}H_0 :g_A ( t ) = g_B ( t ) \\ H_1 :g_A ( t ) \neq g_B ( t ) \end{align*} \end{document}

We assume that the BI change smoothly over time, that is, g_A (t) and g_B (t) are smooth functions of time, which can be estimated from the semiparametric mixed approach. In particular, we assume the following model for the BI in area k \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland,xspace}\usepackage{amsmath,amsxtra}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} \begin{align*}BI_k ( t ) = g_k ( t ) + \varepsilon_k ( t ) \tag{1}\end{align*} \end{document}

Let tj ( \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\pagestyle{empty}\DeclareMathSizes {10} {9} {7} {6}\begin{document}$$j = 1 \ldots N$$\end{document} ) denote the time points, where BI is measures; here, N=84, the number of A. aegypti inspection cycles between January 1999 and December 2005. As shown in Maringwa et al. (2008b), the expected value of BI(t) can be formulated as a mixed model of the form \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland,xspace}\usepackage{amsmath,amsxtra}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} \begin{align*}BI_k ( t ) = \beta_0 + \beta_1 t + \mathop\sum_{j = 1}^J b_j ( t - \tau_j ) _ + + \varepsilon_k ( t ) \qquad &(2).\end{align*} \end{document}

Here, \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\pagestyle{empty}\DeclareMathSizes {10} {9} {7} {6}\begin{document}$$j = 1 \ldots N$$\end{document} are prespecified set of knots, and the truncated function in equation (2) is given by \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland,xspace}\usepackage{amsmath,amsxtra}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} \begin{align*}( t - \tau_j ) _ + = \begin{cases}t - \tau_j \quad t > \tau_j \\ 0 \qquad \quad { \rm elsewhere}\end{cases}\end{align*} \end{document}

and \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\pagestyle{empty}\DeclareMathSizes {10} {9} {7} {6}\begin{document}$$b_1 \ldots b_J$$\end{document} are a set of r random effects for which we assumed b_j ∼N(0; σ_j), \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland, xspace}\usepackage{amsmath, amsxtra}\pagestyle{empty}\DeclareMathSizes {10} {9} {7} {6}\begin{document}$$j = 1 , \ldots , J$$\end{document} ; in our case, we used J=60.

As shown in Maringwa et al. (2008b), in matrix notation, the model in (2) can be expressed as follows: \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland,xspace}\usepackage{amsmath,amsxtra}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} \begin{align*}BI_k ( t ) = X \beta_k + Zb_k + \varepsilon_k ( t ) \qquad &(3).\end{align*} \end{document}

Here, X is a known design matrix for the fixed effect β _k, \documentclass{aastex}\usepackage{amsbsy}\usepackage{amsfonts}\usepackage{amssymb}\usepackage{bm}\usepackage{mathrsfs}\usepackage{pifont}\usepackage{stmaryrd}\usepackage{textcomp}\usepackage{portland,xspace}\usepackage{amsmath,amsxtra}\pagestyle{empty}\DeclareMathSizes{10}{9}{7}{6}\begin{document} \begin{align*}X = \left[ \begin{matrix}1 \ & t_1 \\ 1\ & t_2 \\ \vdots\ & \vdots \\ 1\ & t_N\end{matrix} \right]\end{align*} \end{document}

Results

Figure 2 shows the fitted profile for BIs in the intervention, extension, and control areas. In 1999, the intervention area had the highest BI values, but a clear decreasing trend was observed over time with the introduction of the intervention actions. In the extension and control areas, a peak was observed just before the dengue outbreak in September 2000. During the intensive vector control campaign in the second epidemic period in 2001, all areas appeared to have similar BI values. Afterward, at the end of the study period, a slight increase was observed in the BI in all three areas, that it was significantly more marked in the extension area.

FIG. 2.

Fitted Breteau indices and 95% confidence bands in the intervention, extension, and control areas. Playa municipality, Havana, 1999–2005.

Figure 3 shows the fitted models for the differences in A. aegypti larvae densities between areas. With regard to the difference between intervention and control area, the BIs in the control area showed the lowest values before starting the intervention. This was reversed 1 year after the launch of the intersectoral activities for dengue control in the intervention area. In spite of spraying actions carried out during the intensive campaign that started in January 2002 in all areas, the differences in BIs between intervention and control areas remain significant until December 2002. Although for the next 2 years, no differences were observed, they become significant again in December 2004, >1 year after the launch of the complementary community-based vector control strategy at circunscripcion level in almost half of the intervention area.

FIG. 3.

Fitted differences in Breteu indices (BIs) in intervention and extension/control area and 95% confidence bands in Playa municipality, Havana, 1999–2005.

The differences between intervention and extension areas remain nonsignificant during a longer time, which coincides with the introduction of intersectoral actions in the extension area. At the end of the study period, the difference between areas becomes nonsignificant, but this could be linked to the uncertainty about the functional relationship and the sparseness of the data.

Discussion

The semiparametric model fitted in this study allowed to illustrate the differences in larval A. aegypti infestation indices among areas with different control strategies. We consider the BI as a continuous normally distributed random variable that might change over time. The use of the linear mixed model presented in this article permits us to compare between the BI of any two transmission areas at each time point without the need to average the BI over a relatively long period of time. This means that we were able to identify trends over time. Using the confidence bands, one is able to identify specific sections where the bands do not overlap with the x-axis (i.e., y-axis=0), thus indicating significant differences. In our data from the interventions in Playa, the effect to move toward intersectoral and community empowerment A. aegypti control becomes evident 1 year after beginning of the intervention. Although the three areas reduced the infestation levels after global intensive campaign, only the intervention area, with a complementary community empowerment approach, showed promising results in term of sustainability.

Since only adult female Aedes mosquitos are directly involved in dengue transmission, the classical larval indices have been criticized as impact measures of interventions (Nathan et al. 2006). However, they continue being widely used as an indicator in surveillance and evaluation of interventions (Heintze et al. 2007a, Erlanger et al. 2008). Further, the BI was strongly associated with transmission in previous studies carried out in Playa municipality (Winch et al. 2002; Avila Montes et al. 2004). The interventions carried out at the CP level were linked to higher level structures; contamination between areas cannot be excluded. Although this would be in itself desirable, it could result in an underestimation of the effect of the interventions.In addition, entomological data collected through the routine dengue control program have some limitations: different vector control technicians inspected the areas, the procedures used may not have been completely standardized, and the intensity of the routine control inspection activities varied with the occurrence of the dengue outbreaks. Additionally, as also illustrated here, entomological success does not necessarily lead to complete expression of disease transmission.

Several studies of community-based dengue control have been reported in the literature, most of them showing a statistically significant effect on the entomological index, but none specifically uses a longitudinal assessment (Leontsini et al. 1993, Wang et al. 2000, Espinosa-Gomez et al. 2002, Lardeux et al. 2002, Winch et al. 2002, Avila Montes et al. 2004, Nam et al. 2005, Phuanukoonnon et al. 2005, Kroeger et al. 2006, Toledo et al. 2007, Toledo Romani et al. 2007, Vanlerberghe et al. 2009). It is not easy to relate those studies to the present one, as there is a huge variation in the design, the type, length, and intensity of interventions as well as in the accurate indicators to measure the effect on A. aegypti infestation. In 2006, a systematic review of community-based dengue control strategies (Heintze et al. 2007a) identified only 11 papers that compared intervention with nonintervention settings, and almost all studies had shorter observation periods (<1 year). Our study covered a 6-year period, a time frame that allows to better appraise the consistency and durability of effects.

Semiparametric mixed models should, in our opinion, be used more often to evaluate the effects of complex longitudinal interventions. Our study reinforces the evidence that the intersectoral coordination and community empowerment strategy to A. aegypti control is effective.

Footnotes

Acknowledgments

The authors gratefully acknowledge the family doctors, formal and informal leaders, and residents of Playa Municipality who participated in the study. The research was funded by the Cuban Ministry of Public Health and through a framework agreement between the Institute of Tropical Medicine (Antwerp) and the Belgium Directorate-General for Development Cooperation. The study received ethical clearance from the Ethics Committee of IPK and the municipality health authorities.

Authors' Contributions

L. Sanchez conceived the study, participated in fieldwork and data analysis, and drafted the article. J. Maringwa and Z. Shkedy participated in the statistical analysis and drafted sections of the article. M. Castro participated in the design of study, the fieldwork, and the collection of the data. N. Carbonell participated in data analysis, and P. Van der Stuyft participated in the design of the study, data analysis, and drafted the article. All authors participated in the interpretation of the data and critically revised subsequent drafts of the article.

Disclosure Statement

No conflicts of interest to be declared.

*

The Consejo Popular (CP) is a governmental structure between the municipal level and the “circunscripción,” the lowest level of local government. The CP is composed of representatives of the circunscripciones, community organizations, the Health Area, and other sectors.

References

Avila Montes

, Martinez

, Sherman

, Fernandez

Evaluation of an educational module on dengue and Aedes aegypti for schoolchildren in Honduras. Rev Panam Salud Pub, 2004; 16:84–94.

Durban

, Harezlak

, Wand

, Carroll

RJ.

Simple fitting of subject-specific curves for longitudinal data. Stat Med, 2005; 24:1153–1168.

Erlanger

, Keiser

, Utzinger

Effect of dengue vector control interventions on entomological parameters in developing countries: a systematic review and meta-analysis. Med Vet Entomol, 2008; 22:203–221.

Espinosa-Gomez

, Herandez-Suarez

, Coll-Cardenas

Educational campaing versus malathion spraing for the control of Aedes aegypti in Colima, Mexico. J Epidemiol Comm Health, 2002; 56:148–152.

Farrar

, Focks

, Gubler

, Barrera

et al. Editorial: towards a global dengue research agenda. Trop Med Int Health, 2007; 12:695–699.

Gubler

The emergence of epidemic dengue fever and dengue hemorrhagic fever in the Americas: a case of failed public health policy. Rev Panam Salud Pub, 2005; 17:221–224.

Gubler

, Clark

GG.

Community involvement in the control of Aedes aegypti. Acta Trop, 1996; 61:169–179.

Guzman

, Kouri

Dengue: an update. Lancet Infect Dis, 2002; 2:33–42.

Heintze

, Garrido

, Kroeger

What do community-based dengue control programmes achieve? A systematic review of published evaluations. Trans R Soc Trop Med Hyg, 2007a. 101:317–325.

10.

Heintze

, Garrido

, Kroeger

Reply to comment on: what do community-based dengue control programmes achieve? A systematic review of published evaluations. Trans R Soc Trop Med Hyg, 2007b. 101:631–632.

11.

Kay

BH.

Intersectoral approaches to dengue vector control. Gaoxiong Yi Xue Ke Xue Za Zhi, 1994; 10:S56–S61.

12.

Kroeger

, Lenhart

, Ochoa

, Villegas

et al. Effective control of dengue vectors with curtains and water containers covers treated with insecticide in Mexico and Venezuela: cluster randomized trials. BMJ, 2006; 332:1247–1252.

13.

Lardeux

, Sechan

, Loncke

, Deparis

et al. Integrated control of peridomestic larval habitats of Aedes and Culex mosquitoes (Diptera: Culicidae) in atoll villages of French Polynesia. J Med Entomol, 2002; 39:493–498.

14.

Leontsini

, Gil

, Kendall

, Clark

GG.

Effect of a community-based Aedes aegypti control programme on mosquito larval production sites in El Progreso, Honduras. Trans R Soc Trop Med Hyg, 1993; 87:267–271.

15.

Maringwa

, Geys

, Shkedy

, Faes

et al. Application of semiparametric mixed models and simultaneous confidence bands in a cardiovascular safety experiment with longitudinal data. J Biopharm Stat, 2008a. 18:1043–1062.

16.

Maringwa

, Geys

, Shkedy

, Faes

et al. Analysis of cross-over designs with serial correlation within periods using semi-parametric mixed models. Stat Med, 2008b. 27:6009–6033.

17.

Nam

, Kay

, Yen

, Ryan

, Bektas

Community mobilization, behaviour change and biological control in the prevention and control of dengue fever in Viet Nam. Dengue Bull, 2005; 28:57–61.

18.

Nathan

, Focks

, Kroeger

Pupal/demographic surveys to inform dengue-vector control. Ann Trop Med Parasitol, 2006; 100:S1–S3.

19.

Ngo

, Wand

. Smoothing with mixed model software. J Stat Soft, 2004; 9:1–56.

20.

Pan American Health Organization (PAHO). Dengue and Dengue Hemorrhagic Fever in the Americas: Guidelines for Prevention and ControlScientific publication No. 548.Washington: PAHO, 1994.

21.

Parks

, Lloyd

, Nathan

, Hosein

et al. International exeriences in social mobilization and communication for dengue prevention and control. Dengue Bull, 2004; 28:1–7.

22.

Perez

, Lefevre

, Sanchez

et al. Community participation in Aedes aegypti control: a sociological perspective on five years of research in the health area “26 de Julio”, Havana, Cuba. Trop Med Int Health, 2007; 12:664–672.

23.

Phuanukoonnon

, Mueller

, Bryan

JH.

Effectiveness of dengue control practices in household water containers in northeast Thailand. Trop Med Int Health, 2005; 10:755–763.

24.

Ruppert

, Wand

, Carroll

RJ.

Semiparametric Regression. New York: Cambridge University Press, 2003.

25.

Sanchez

, Perez

, Cruz

, Castro

et al. Intersectoral coordination, community empowerment and dengue prevention. Six years of controlled intervention in Playa municipality, Havana, Cuba. Trop Med Int Health, 2009; 14:1356–1364.

26.

Sanchez

, Perez

, Sosa

et al. Intersectoral coordination in Aedes aegypti control. A pilot project in Havana City, Cuba. Trop Med Int Health, 2005; 10:82–91.

27.

Toledo

, Vanlerberghe

, Baly

, Ceballos

et al. Towards active community participation in dengue vector control: results from action research in Santiago de Cuba, Cuba. Trans Roy Soc Trop Med Hyg, 2007; 101:56–63.

28.

Toledo Romani

, Vanlerberghe

, Perez

, Lefevre

et al. Achieving sustainability of community-based dengue control in Santiago de Cuba. Soc Sci Med, 2007; 64:976–988.

29.

Vanlerberghe

, Toledo

, Rodriguez

, Gomez

et al. Community involvement in dengue vector control: cluster randomised trial. BMJ, 2009; 338:1477–1480.

30.

Verbyla

, Cullis

, Kenward

, Welman

SJ.

The analysis of designed experiments and longitudinal data by using smoothing splines. Appl Stat, 1999; 48:269–311.

31.

Wang

, Chang

, Wu

, Ho

CM.

Integrated control of the dengue vector Aedes aegypti in Liu-Chiu Village, Ping-Tung County, Taiwan. J Am Mosq Control Assoc, 2000; 16:93–99.

32.

Winch

, Leontsini

, Rigau-Perez

, Ruiz-Perez

et al. Community-based dengue prevention programs in Puerto Rico: impact on knowledge, behavior and residential mosquito infestation. Am J Trop Med Hyg, 2002; 67:363–370.