Study area
The study took place in Kano state (Northwest region) and included the 24 local government areas (LGA) covered by the two campaign waves, which took place in May and July 2009, respectively (Figure 1). During Wave 1, the campaign covered 21 LGAs, while during Wave 2, 23 LGAs were covered.
Study population
This cross-sectional household survey utilized a stratified two-stage cluster sampling design. The strata were the two areas covered in the different campaign waves. Each stratum was considered a survey domain. A total of 60 clusters were selected, including 30 from each stratum. No urban/rural stratification was done but clusters were defined as urban or rural based on their categorization in the 2006 census. Seventeen (17) households were selected from each cluster, resulting in a total sample of 1,020 households in the campaign area.
The sampling procedure of the required number of households was done in two stages including:
Stage one: selection of clusters
The household registration lists from the distribution campaigns were used for the selection of clusters. A cluster was defined as a community and selection was carried out independently for each distribution wave using a two-step procedure: first a cumulative list of registered households by ward was compiled and 30 clusters from each of the two strata were selected using systematic sampling with probability proportionate to size (PPS). Second, a list of all communities and the number of registered households was compiled for each selected ward and the required number of villages was selected again using PPS.
Stage two: selection of households
Within each selected community 17 households were selected using the following methodology: For small communities (less than 100 compounds), the field team mapped the entire village and from the compiled list of eligible households the supervisor randomly selected 17 households with equal probability for each household. Following the household definition used in the distribution campaign, which was "a wife with her direct dependents", a compound was divided into several households depending on the number of wives. The husband was assigned to the first wife's household. For large communities (more than 120 compounds), the equal size section-approach was used. With the help of local chiefs, the community was divided into sections with approximately equal number of compounds. The supervisor then selected randomly one of these sections within which all households were mapped and selected as above.
Data collection
Data collection took place from October 19 to November 4, 2009, corresponding to five months after the first campaign and three after the second campaign. The data collection was done using a questionnaire adapted from Malaria Indicator Survey (MIS) Household Questionnaire [13]. The questionnaire was composed of six sections including household roster, household characteristics, campaign net distribution, nets received during the campaign, nets owned by the household, and nets previously owned by the household. The questionnaire was pre-tested in 30 households in a community that was not selected for the survey and corrections were made before the training of the field team.
Prior to the fieldwork, community mobilization activities took place. This was critical; specifically, it attempted to ensure that the researchers did not create further expectation of another distribution campaign after the survey. Such expectations could have potentially influenced some households to under report their mosquito net possession in hopes of receiving additional nets.
Each selected household was visited and the head of household or one of his/her adult dependents was interviewed. In case no appropriate respondent was found at the house, a new visit was scheduled later that day. At least three attempts were made to reach a respondent before dropping the household without replacing it. The main respondent was the head of household or his/her adult dependents except for the section on the mosquito net receipt at the delivery point, for which the person who collected the mosquito net was interviewed.
To ensure high data quality, the team supervisor reviewed all questionnaires daily for completeness and possible inconsistencies and ensured that missing information was corrected while still in the field. In addition, spot-checks were performed on 12% of interviews conducted by each fieldworker.
Data processing and analysis
Data entry was done using QPS software with double entry of all records. Both data sets were then compared and any discrepant records were verified using the original questionnaires. After the first stage of cleaning, the data set was transferred to STATA 10 for further consistency checks, preparation of data files and analysis.
Two types of analysis were performed. The first one included two binary response logistic regression models. Model 1 assessed the effect of socio-economic status (wealth quintile- highest quintile used as reference) of the household on ITN (including both conventional ITNs and LLINs) ownership after the campaign, controlling for several covariates including campaign wave, education level of head of household, size of household, presence of pregnant women in the household, presence of under fives in the household, and whether the household was present at the mosquito net distribution point. The second model was used to assess the effect of socio-economic status on ITN use of individuals from households which owned at least one ITN. Only members who slept in the household the night preceding the survey visit were included in the analysis. Covariates controlled for included place of residence, gender, education of the head of the household, size of the household, ratio ITN/household member, age, and use of mosquito repellents (aerosol, coils and herbs)
The second set of analyses included equity analysis using the Lorenz concentration curve and index [14] to assess the relative fairness of the distribution in terms of household ITNs ownership and use considering the wealth quintile. The concentration curve plots the cumulative proportion of the outcome variable (ITNs ownership and ITNs use) against the cumulative proportion of the sample population ranked by socio-economic status (wealth quintiles). The curves are compared to the diagonal or equity line. A curve above the diagonal line will indicate the concentration of the outcome among the poor while a curve below will means that the outcome is concentrated among the rich. If there is an equal concentration among poor and rich the curve matches with the diagonal line [14].
The concentration index ranges between -1 and 1. An index of 0 reflects equitable distribution of the outcome between poor and rich while a concentration index of more than 0 suggests that the outcome is more prevalent among the poor. Conversely, a negative index indicates that the outcome is more concentrated among the rich. An index closer to 0 therefore expresses more equitable distribution of the outcome among household with different socio-economic levels [14].
Ethical clearance
This paper used data from the Kano post mosquito net free distribution campaign survey conducted on the behalf the Federal and State Malaria Control Program. Because this was part of the programmatic activity, ethical clearance was exempted. Informed consent was obtained from each participant.