1.4 million LLINs were distributed from July 8 to 14, 2006, primarily in Nyanza and Western provinces, in association with a measles vaccination campaign. Western and Nyanza provinces were chosen to receive nets first because of a high burden of malaria-related morbidity and mortality. The remaining 2.0 million nets were distributed September 23 to 24, 2006 in malaria-affected districts in Rift Valley, Central, Eastern, and Coast Provinces. The target population to receive the nets was children under five years of age. Batches of LLINs were sent to numerous distribution centers (health facilities, schools, etc.) where caretakers would receive one net for each CU5 that accompanied them. At some centers, the children received other health interventions appropriate for the locale (measles vaccination, deworming medication, vitamin A, and iron tablets).

There are 46 Districts in Kenya where malaria transmission occurs. Twenty of these Districts were randomly chosen with probability proportional to size within the following three malaria transmission zones that formed the sampling strata for the survey: areas of endemic malaria - 10 districts from Nyanza, Western, and Coast provinces (2006 estimated population: 11.9 million); epidemic malaria - five districts from the Rift Valley (2006 estimated population: 8.6 million); highland/seasonal malaria - five districts from Eastern and Central Provinces (2006 estimated population: 9.2 million). The region containing Nyanza, Western, and Coast Provinces had higher sampling fractions because they contained all but one of the endemic districts in the country and are where the majority of malaria-related morbidity and mortality occurs.

Within each district, five enumeration areas were selected with probability proportional to size. A sample of 21 households was selected for each enumeration area, giving a target sample of 2,100 households. The sampling frame within each cluster was developed through the use of global positioning system (GPS) linked with hand-held computers (PDAs) to allow rapid preliminary mapping of enumeration areas, thereby allowing selection of a simple random sample of houses [13]. Sufficient numbers of districts and enumeration areas were sampled in the Nyanza and Rift Valley provinces and the other two regions to produce statistically reliable estimates and confidence intervals.

The survey was conducted from October 17 to November 1, 2006 (Figure 1). Questions and methodology were similar to the 2003 Kenya Demographic and Health Survey (KDHS) to facilitate comparisons [17]. The survey instrument asked questions about the household, the bed nets owned by the household (if any), whether a CU5 or a woman of reproductive age had slept under each specific net during the previous evening, malaria knowledge and awareness of public media messages related to the campaign, and socio-economic questions concerning characteristics of the household. Surveyors noted whether they had actually observed the nets and if so, if they were hanging. Although pregnant women were not specifically targeted by the campaign, they were included in the survey to measure the effectiveness of the targeted campaign for CU5s in reaching them. Socioeconomic data used to construct a wealth index employing the same definitions used in the 2003 KDHS [17].

PDAs with GPS were used for data collection. The questionnaire was constructed and programmed into the PDAs with appropriate skip patterns and data checks. After a one-week training period, 11 teams of four persons each collected the data in the field over a two week period. Upon survey completion, PDAs were returned to Nairobi, where data were aggregated into a desktop database.

The survey was conducted at the end of the dry season and the early stage of the short rainy season which is not the peak time of malaria transmission. As a result it was not expected to provide the maximum measure of net usage. The objective of this survey was to assess ownership and retention of bed nets distributed during the campaign and to assess bed net ownership in general.

A household was defined as everyone who shared a meal at a common gathering place, using the 2003 Kenya DHS definition [17]. An ITN was defined as any bed net that had been treated with insecticide in the last six months. LLINs were counted as ITNs. At the time of the survey, there were only three types of LLINs available in Kenya: SupaNet ExtraPower (socially marketed by Population Services International), Permanets (available commercially and via the campaign), and Olyset nets (available commercially and via the campaign). Pre-campaign bed net ownership levels were based on ownership of non-campaign bed nets. Changes in ownership of these nets between the campaign and the survey were assumed to be negligible. Post-campaign bed net ownership included all types of bed nets. Permanet and Olyset campaign nets had distinctive labeling and packaging. To aid identification, the PDAs used for survey data collection had pictures of each net, and its packaging and labeling, integrated into the computerized questionnaire. Equity ratio was defined as the ratio of intervention coverage proportions in the poorest quintile to the coverage in the wealthiest quintile.

Consistent with the sampling design, the SAS survey procedures (SAS v9.1.3, SurveyFreq, SurveyLogistic, and SurveyMeans) were used to produce estimates and standard errors using the sampling weights that accounted for the sampling design and the different sampling fractions for each transmission zone. Estimates are precise (half-width of 95% confidence interval) to 2.5% on a national basis (i.e., for all areas with malaria in Kenya), 2.5-4.0% on a regional basis (the three sampling strata based on provinces grouped by the nature of their malaria endemicity), 5% for each wealth quintile, and 6% for CU5. The survey was not designed for making precise statements on bed net ownership or usage for pregnant women. However, precision for women of reproductive age was 5% on a national basis.

Comparisons of indicators before and after the campaign were done by computing the difference in the indicators and computing 95% confidence intervals that accounted for the survey design. The Rao Scott Chi square statistic was used for p-values. Projected numbers of various types of bed nets (ITNs, campaign nets) were computed by multiplying the number of a particular type of bed net in the survey household by its sampling weight and summing this product over all households.

The protocol was approved by the Kenya Medical Research Institute Institutional Review Board and the Centers for Disease Control and Prevention. Informed written consent was obtained from each respondent.

From October 17 through November 1, 2006, members of 2,059 households were interviewed out of 2,100 targeted. Of the 100 enumeration areas, 85% were classified as rural and 15% as urban, mirroring the characteristics of the 46 campaign districts. On a weighted basis, households with CU5s represented 51.7% of all households in the survey population. Among sampled households, 78.0% had a woman of reproductive age; of these, 37.1% did not have a CU5 (95% CI: 34.7, 39.6%). An estimated 8.3% of the women in the population reported themselves to be pregnant.

The survey estimated that 3.63 million LLINs were distributed (95% CI: 3.39, 3.87 million), of which 3.38 million were distributed to CU5s (93.9%) (Table 1). The campaign increased the number of bed nets owned by 50%, from 7.2 to 10.9 million. Households owning only campaign nets were 13.8% of all households, which was over one-third of all households receiving campaign nets. Pre-campaign ITN household ownership could be estimated as 36.9% = 50.7 (post-campaign ITN/LLIN HH ownership) -13.8% (households that only owned campaign nets in the survey), and similarly pre-campaign household ownership of any bed net could be estimated at 53.1% (66.9-13.8%). Prior to the campaign, socially marketed nets represented 68.7% of all nets owned. This represents approximately 5.0 million nets (7.24 million* 0.687). After the campaign, they represented 46.0% of all nets owned. Figure 2 presents the distribution of bed net brand ownership before and after the campaign.

The campaign increased ITN ownership for HH with CU5s by 28.7% (from 46.7% to 74.4%, p < .0001). It increased ITN ownership for HH with women of reproductive age by 18.6% (from 40.3% to 58.9%).

Prior to the campaign, the percent of all nets in households with CU5s that were ITN increased with the wealth quintile. However, even in the wealthiest quintile, less than a third of the homes owned ITNs. After the campaign, the relationship between ITN ownership and wealth disappeared. The smallest increase in ITN ownership was in the wealthiest quintile, 25.5%,. The other wealth quintiles had increases of nearly 40% or more. The increases in all wealth quintiles were statistically significant (p < .001 for all quintiles, Table 2).

The equity ratio for all households receiving campaign bed nets was 1.59 (95% CI: 1.29, 1.89). This means that the poorest households were 59% more likely to receive campaign bed nets than the wealthiest households. The equity ratio for CU5s, women of reproductive age, and pregnant women sleeping under an ITN the previous night were all less than 0.8. This means that despite the campaign, the target populations in the wealthiest quintiles were all 20% or more likely to be sleeping under an ITN. Equity ratios with 95% confidence intervals are presented in Figure 3.

There were 3141 bed nets in the 2059 households in the survey. Over half of households surveyed (n = 1080) owned ITNs. One third of all nets were campaign nets. During the survey, 70% of all nets were seen by the interviewers. Summary data with 95% confidence intervals are presented in Table 3.

Among all bed nets, 66.2% were either treated in the last six months (ie, were ITNs) or were LLINs. More than half (54.8%) of all nets were LLINs. Among all bed nets, 43.1% were obtained at no cost. Nearly two-thirds (67.1%) of all nets were hanging. About a third (30.6%) of all nets were obtained at health facilities (other than as a part of the campaign), and 24.8% were obtained from dukas (small stores), hawkers, kiosks, or textile shops. Almost ninety percent (88.4%) of nets were classified as being in good condition (no holes, tears). For the most popular brand of socially marketed bed net (a non-LLIN), 30.6% had been treated in the last six months.

About two-thirds (66.9%) of all households had at least one bed net of any kind. Over forty percent (41.9%) of households had more than one bed net. Over half (50.7%) of households owned an ITN, and 39.1% of all households had received a campaign LLIN. ITNs were hanging in 41.9% of all households. In households owning ITNs, 82.6% had at least one ITN hanging. Households that owned only campaign nets were 13.8% of all households, which was over one-third of all households receiving campaign nets. Summary data with 95% confidence intervals are presented in Table 4.

Households that had neither a CU5 nor a woman of reproductive age comprised 19.2% of all households. In this group, bed net ownership was 16.2% for any ITN, and 33% for any type of bed net. Urban households were more likely than their rural counterparts to have nets of any type. (ITNs: Odds Ratio: 1.37, 95% CI: 1.06, 1.77; Any Net: Odds Ratio: 2.35, 95% CI: 1.70, 2.35). Among all households that received campaign nets, 94.8% had retained all nets and 97.9% percent had retained at least one campaign net. There was no significant variation in retention by malaria transmission zone status.

There were 1,727 CU5s living in 1113 households in the survey. Approximately two-thirds (67.5% ) of households with CU5s, received campaign LLINs. The percent of these households receiving campaign LLINs did not vary significantly by region. Just under ninety percent (87.2%) of households with CU5s had a bed net of any type, with 74.4% owning an ITN. Summary data with 95% confidence intervals for households with CU5s are presented in Table 5.

Just over half (51.7%) of CU5s (95% CI: 48.8, 54.7%) slept under an ITN during the previous night. This proportion did not vary significantly by age or sex. There was significant variation by malaria zone/geographic region: endemic: 50.3%; epidemic: 41.4%, highland/seasonal 67.9% (p < .0001). In households with CU5s owning ITNs, 69.0% of children slept under them the previous night, and in households with CU5s owning ITNs that were hanging, 83.5% of children slept under them the previous night. Having received a campaign bed net greatly increased the likelihood that a child under 5 slept under an ITN the previous night (67.7 vs. 28.8%, p < .0001). Overall, 56.5% of the ITNs used by CU5s were campaign nets. Summary data with 95% confidence intervals for CU5s are presented in Table 6.

There were 2,053 women of reproductive age (15-49 years, WORAs) living in 1,613 households. There were 182 pregnant women living in 180 households in the survey. Among pregnant women, 58.4% slept under a net of any type during the previous night (95% CI: 50.9, 65.9%); however, only 36.3% of pregnant women slept under an ITN. (95% CI: 29.0, 43.7%). Among all WORAs, 33.0% slept under an ITN during the previous night (95% CI: 30.6, 35.4%). In households that owned one or more ITNs, 55.9% of women of reproductive age slept under an ITN during the previous night. For pregnant women, this number was 56.7%. Summary data for WORAs with 95% confidence intervals are presented in Table 7.