In total, 2,304 and 2,430 households were randomly selected for participation in the 2004 and 2005 surveys, respectively. These households reported the presence of 2,721 and 3,123 children below the age of five years in 2004 and 2005, who were invited to participate in the survey. Of the invited children 70 and 42 were subsequently excluded because their dates of births revealed they were 60 months or older at the time of the survey. Dates of births were recorded from their under-5 clinic cards where available. Of the eligible children, 95.5% (2,532/2,651) and 96.3% (2,967/3,081) were enrolled. The main reason for non-enrolment was not being at home due to travel or work (4.4% and 3.4%). Only 3 (0.1%) and 9 (0.3%) children were not enrolled due to a refusal by their caretakers. A further 165 records were excluded from the final analyses because a) children were erroneously enrolled as part of a second or third mother-children pair within households of polygynous families in 2005 in Tone (147 records) or, b) because of data quality problems (18 records). In total, 5,334 children contributed to the results; 2,521 (from 1,749 households) in the 2004 survey and 2,813 children (from 1,828 households) in the 2005 survey. Three hundred and sixty three households were involved in both surveys (19.9% of 2005 households).
Characteristics of the study sample
The median number of children per EA was 29 in 2004 (range 11-51) and 31 in 2005 (range 17-54). District level characteristics of the enrolled population in the pre- and post intervention surveys are presented in Table . Both child and household level characteristics, overall, and by district, were comparable between surveys, with the exception that there were more households with more than 1 child in 2005 compared to 2004, and that caretaker education level was lower in Tone as compared to the other two districts. In 2005, 22.7% of enrolled children were below 18 months of age (i.e. an age group not targeted by the child health campaign).
Characteristics of 5334 children who participated in the 2004 and 2005 surveys
Pre- and post-intervention rainfall patterns
Monthly district level rainfall patterns were compared for the pre- and post-campaign year (Figure ). While rainfall patterns differ considerably between the three districts, similar rainfall patterns were observed within each of the 3 districts during 2004 and 2005.
Bed net ownership and use
Before the campaign nets of any kind were reportedly present in 5% to 12.7% of the households in the three districts (Table ). Less than 1% of children were reported to have slept under an ITN the previous night. Nine months after the campaign, net ownership and use had increased dramatically, with over 60% of the enrolled households reported to own an ITN in all three districts (Table ). Households in Tone that owned a net were more likely to hang their net than households in Ogou or Yoto (95.4% versus 75.7% and 62.9%, respectively, p < 0.001). Young children <5 years of age were present in 80.0%, 73.3% and 72.3% of eligible households in Yoto, Ogou and Tone, respectively. Based on the coverage survey results the community household level coverage of hanging ITNs (thus including households without young children) at the time of the survey was estimated to be 34.5%, 44.1% and 67.2% in Yoto, Ogou and Tone, respectively. ITN use during the previous night in the pooled group of children aged 0 - 59 months of age varied considerably between the three districts in the post-campaign survey, with prevalences of 35.9%, 43.8% and 80.6% in Yoto, Ogou and Tone, respectively (Table ). As the campaign only targeted children between nine to 59 months, and the post-intervention survey occurred nine months post-campaign, ITN use was also assessed by age group (Table ) and presence of a sibling who had been eligible for the campaign (Figure ). The use of an ITN was generally shared with other household members. Only 1.2% of children reportedly slept alone under an ITN. In 91.8% the mother reportedly slept with the child under the ITN, and 35% of children reportedly shared with both their mother and one or more siblings.
Reported and observed net coverage during the 2004 and 2005 surveys by district.
2005, current ITN use by age and presence of a targeted sibling <5 years of age in the household. Current ITN use was defined as reported ITN use during the previous night with a hanging ITN in the household.
Reported history of malaria and anaemia and related treatment
There was a marked reduction in reported fever cases and health care attendance in Ogou (Table ), but not in the other two districts. In Yoto, there was a smaller, but significant increase in reported fever cases for which health care was sought. The reported history of febrile episodes for which health care was sought, and the subsequent reported use of an anti-malarial or iron showed no decrease in 2005 compared to 2004 (Table ).
Reported history, health care seeking behavior, malaria and anaemia treatment in 2004 versus 2005 by district in 5331 children aged less than 5 years.
Overall, the prevalence of moderate-to-severe anaemia was 28% lower in 2005 compared to 2004 (PR = 0.72, 95%CI 0.62-0.84, adjusted for age group, SES and district) and there was a 0.35 g/dL (95%CI 0.25-0.45) increase in mean haemoglobin between the two survey points. Importantly, these differences between surveys depended on a combination of the district and the age category involved, in particular whether children were targeted by the campaign and were ≥18 months old or not at the time of the 2005 survey. When adding a three-way interaction term, survey × district × age group, to the model, its P-value was 0.048 for moderate to severe anaemia and 0.0008 for mean haemoglobin level. Because of this observed effect modification, results were also analysed stratified by both district and age group, as discussed below.
In the central district of Ogou, with the long, stable rainy season, and 43% of the children reportedly sleeping under ITNs in 2005, substantial differences were observed across most of the anaemia and malaria indicators. Comparable reductions among both children below and over 18 months of age (Table , Figure , Figure and Figure ), resulted in a pooled 50% and 42% reduction in prevalence of moderate to severe anaemia and clinical malaria, respectively, in children below five years of age with an increase of 0.6 g/dL in mean haemoglobin level. Little or no change was observed in parasite infection levels.
Multivariate morbidity endpoints in 2004 versus 2005 by district and age group targeted by the campaign in 5331 children aged less than 5 years.1
Crude prevalence of moderate to severe anaemia and mean Haemoglobin (Hb) levels in the 3 districts, presented by age group and survey, taking into account clustering at village level.
Crude prevalence of parasitaemia in the 3 districts, presented by age group and survey for any parasitaemia and parasitaemia ≥5000/μL.
Crude prevalence of P. falciparum (Pf) clinical malaria in the 3 districts, presented by age and survey.
In the southern district of Yoto in the coastal plain, with the lowest reported rate of current ITN use and community ITN coverage of the three evaluated districts, significant differences were only observed in the age group directly targeted by the campaign, i.e. children from the age of 18 months onwards (Figures , and and Table ). While moderate to severe anaemia and clinical malaria were reduced by an average 34% and 50%, respectively in this group in 2005, no significant differences were seen in the younger age category (Figure , Figure , Figure and Table ). Similarly, a significant increase of mean haemoglobin of 0.50 g/dL, 95%CI 0.34-0.66, Figure ) was only found in the targeted age group.
In the third, most northern district of Tone with its semi-arid climate and short rainy season, where over 80% of children were currently using ITNs and 40% of households had more than one campaign net hanging, no significant differences were observed in any age group. The younger age group up to 18 months showed a borderline significant reduction in moderate to severe anaemia prevalence (PR = 0.80, 95%CI 0.64-1.01), but this was not accompanied by a change in mean haemoglobin, or a difference in the prevalence of the assessed malaria markers (Figures , , and Table ).
Malaria morbidity in ITN users versus non-users in 2005
In a weighted analysis of the 2005 survey, children who were reported to sleep under the ITN and had an ITN hanging in their household (observed) had significantly less moderate to severe anaemia compared to other children who either did not sleep under a net or did not have an ITN hanging in their household (PR = 0.81, 95%CI 0.67-0.98, adjusted for age and district). SES and sex were not found to be confounders. While ITN use or lack thereof during the previous night does not necessary relate well to persistent use, further analysis at district level suggested a trend where the difference between ITN users and non-users with respect to moderate to severe anaemia was biggest in the district with the lowest community coverage level, and smallest in the district with the highest community coverage level (Yoto PR = 0.70, 95%CI 0.48-1.03, Ogou PR = 0.86, 95%CI 0.61-1.22 and Tone PR = 0.89 (0.65-1.21). Similarly, a difference between these same groups was also found in mean haemoglobin; ITNs users had higher mean Hb than nonusers: 0.17 g/dL, 95% CI 0.03-0.31, adjusted for age and district). There was no effect modification by district and/or age).