The study was undertaken between November 2000 and June 2003 during which five child health days occurred in each parish. Because of a delay in funding, almost 12 months, rather than 6 months, elapsed between the first two child health days. presents descriptive statistics on the child health days and the percentage of children classified as underweight at each round. The absence of any apparent difference in the percentage of underweight children or in the mean Z scores between the treatment and control groups does not indicate the absence of an effect, as new children entered the project and study at each round, whereas others did not always attend.
Dates of child health days in rural Uganda, sample sizes, proportion of children classified as underweight (−2 SD below reference vales), and mean Z scores of weight for age of children in treatment and control parishes
shows that at least two measurements of body weight were made on 14 940 treated children and 13 055 control children, and that there was a statistically significant difference in extra weight gained of 154 g (95% confidence interval 96 to 214, P < 0.01). This is equivalent to an extra 166 g per year (16 to 316 g) or nearly 10% of average initial body weight.
Average weight gain, months in programme, and number of visits to child health days in Uganda for all children with two or more measurements
shows the results of six regression models, five using total weight gain and the sixth using weight gain per month. Model 1 indicates that weight gain was greater in children who attended more child health days but that children in the parishes where albendazole was given gained 55 g (9 to 104 g) more weight per visit than children in control parishes. The robustness of this model was then tested in other models.
Variables generated by regression models examining effect on weight gain of attendance at child health days in rural Uganda where albendazole was provided. 95% confidence intervals are corrected for cluster sampling effects
Model 2 shows that the interval between the first and last measurements was highly correlated with total weight gain, as expected, but the difference in total weight gain per visit between study groups did not change although the magnitude of both did change.
Model 3 shows the effect of attending child health days in which the treatment effect is divided into three groups: about twice a year, annually, or at a longer interval. The children treated twice a year gained more weight than children treated less often.
Model 4 controls for initial weight. No biological interpretation can be assigned to the coefficient of initial weight in this model since any measurement error in the initial weight is also in the dependent variable, total weight gain, leading to a bias towards minus one for that coefficient. This initial weight variable, however, picks up unexplained variance without biasing the variable of interest. It indicates that there was no bias.
Model 5 repeats model 4 but uses a value for initial weight predicted from the height of children aged more than 2 years, their sex and age, and fixed effects of the parish. This was done because the initial weight used in model 4 is correlated with the outcome, gain in weight. The prediction equation had an R2 value of 0.782. As recumbent length was not recorded for children less than two years old, not every child weighed had height recorded, so the sample size was reduced. Neither models 4 nor 5 led to an appreciable change in variable estimates.
Model 6 uses the variables in model 3, but standardises the outcome per unit time by using weight gain per month as the outcome variable. It shows an average additional weight gain of 13.8 g per month among children treated about twice a year, but it was not significant if the interval between treatments was a year or more. In model 6 the interval between the first and last measurement was negatively associated with monthly weight gain. This probably reflects the fact that weight gain declines with age so that, as the interval between the first and last measurements increases the longer participants were studied, the velocity of weight gain decreases. The length of time each child was studied has a strong correlation with the total weight gain, but not with weight gain per month.
shows that the percentage of households who reported deworming their children before the programme started in 2000 was similar in both treatment and control parishes. By 2003 three times as many children in the treatment parishes had been dewormed, mostly at the child health days, but this practice had also increased by 50% in the control parishes. Most treatments for the control group were obtained from private clinics or shops, where the average cost was Ugandan shillings (USH) (SD 705) 748 or about £0.20 (€0.30, $0.40). No significant difference was found in the average number of child health days attended by children in the treatment and control parishes.
Table 4 Proportion of children aged 1 year or older who were reported to have been dewormed in samples of households in two surveys three years apart in parishes in rural Uganda in which deworming was being provided (treatment) at child health days and in parishes (more ...)