High effective coverage of vector control interventions in children after achieving low malaria transmission in Zanzibar, Tanzania

The study was conducted during June-July 2009 in two districts of Zanzibar: North A district (on Unguja island) and Micheweni district (on Pemba island). Zanzibar is an archipelago off the coast of mainland Tanzania. It previously had high and stable transmission of Plasmodium falciparum malaria, with malaria burden peaking during the rainy seasons; the long or heavy rains (Masika) between March/April and May/June and the short rains (Vuli) from October to December. However, in recent years there has been a dramatic decrease in malaria prevalence as a result of implementation and reinforcement of different malaria control measures [3, 4].

In May 2005, the overall ITN use in children under five in Zanzibar was documented at 40%, with the Micheweni district having the lowest under-five ITN use (<10%) [30]. Consequently, retreatment campaigns were carried out in Micheweni during 2005, followed by a targeted mass distribution of long-lasting insecticidal nets (LLINs) in August 2005. The distribution campaign was scaled-up to other districts in early 2006. The LLINs distributed were blue rectangular Olyset® nets that were free of charge to all children under five and pregnant women. Free mass distribution of LLINs also took place from 2008 till 2009. In this distribution all households received two LLINs, except for households with a single resident that received only one net. The majority of LLINs were blue rectangular Olyset® nets, but some were white LLINs that cannot be easily distinguished from conventional nets.

From 2006 till 2009, there have been four rounds of IRS, targeting all households in Zanzibar (excluding Stone Town), with the last one implemented approximately six months before the survey was conducted. The insecticide used for IRS is the synthetic pyrethroid lambda-cyhalothrin (ICON).

A cross-sectional household survey using two-stage cluster sampling technique [31] was used. The selected sampling units, the shehias (the smallest administrative unit which comprises several villages), were the same as those randomly selected in a previous survey in 2006 [8]. However, the 22 shehias chosen in 2006 had been restructured by the Zanzibar authorities into 32 shehias due to population growth; hence 32 sampling units were used in this survey. Households were randomly selected from these sampling units in proportion to shehia size. Assuming a proportion of 50% of under-five children sleeping under bed nets, and accounting for a cluster effect of two, a sample size of 192 under-five children was needed to determine LLIN use with an absolute precision of ±10% and a 95% confidence interval. Since the survey was done in conjunction with the annual malaria cross-sectional survey conducted in the two districts, the total number of caretakers interviewed was larger than needed (560 in total), and some households had more than one under-five child. If a household on the sampling list could not be found, or consent could not be obtained, it was replaced by another household from a reserve list. Participation consent by community members was generally high.

Data were collected in North A and Micheweni districts right after the rainy season through household interviews with caretaker, preferably mothers, using a structured questionnaire with closed and open-ended questions in Kiswahili. Nineteen trained interviewers, who were all health professionals, received one week of training on the interview technique. The caretakers were interviewed in their homes, and asked about different perceptions and beliefs on the malaria situation in Zanzibar, vector control interventions (ie, bed nets and IRS), current vector control coverage, and intention to continue the use of vector control interventions in the event that malaria burden further decreases.

Data were single entered in CSPRO 4 by a data entry clerk, checked for consistency and errors, and analysed in STATA 10. Open-ended answers were coded into different categories and analysed quantitatively. Frequencies and proportions of perceptions and bed-net usage were computed. Factors associated with under-five LLIN use were identified using bivariate logistic regression. All variables with a p-value ≤0.25 in the bivariate analysis were included in the multiple logistic regression model after they were checked for colinearity. p-values were adjusted for cluster effect on the shehia level using the STATA svy command.

The study population was grouped into socio-economic quintiles based on an asset index, which was created using principle component analysis (PCA) [32]. The index was based on type of floor, walls and roof, source of water and light, type of toilet and cooking facilities, and owning 20 different assets. LLIN and IRS effective coverage in under fives belonging to different socio-economic quintiles was compared.

Most questions in the structured questionnaire were dichotomous closed questions with "yes" or "no" answers. These included the questions on perceived risk of malaria, perceived usefulness of bed nets and IRS in preventing malaria and mosquito bites, perceived usefulness of combining preventive measures and continued use of preventive measures after malaria further decreases. Although giving a dichotomous answer to these qualitative-natured questions could cause a bias, it was deemed as an appropriate way to quantify these perceptions. Another option would have been to use other quantitative techniques, such as Likert-type scales, which could potentially have given more nuanced responses. However, following the dichotomous questions, open-ended questions were used, where the participants could further explain or justify their answers.

Use and willingness to use malaria prevention might have been overestimated due to desirability bias. This bias may have been increased due to the fact that the survey was identified with the Zanzibar Malaria Control Program and the interviewers were health professionals. However, an attempt to minimize this risk was made through emphasizing the importance of creating a comfortable environment during the interview.