3.1. Description of the Study Population
In total, 2609 children under five years who had full covariate information were used in the current analysis.
Table 1 shows that 77% of the children are positive for anemia and about 15% for malaria.
Analysis by sex showed that there were more male children (51.36%) and more male headed households (85.17%) than female. The distribution of children by place of residence shows that 71.18% are in rural areas and only 28.82% live in urban areas. However, children who are in big cities and those in secondary cities are almost equal, (14.64%) and (14.18%), respectively. Regarding the analysis by administrative region, the majority of children are in the region of Boke (17.90%), followed by Nzerekore (16.67%), and the region of Mamou has the lowest percentage (8.78 %). In terms of natural region, 24.76 % of children lived in Maritime Guinea, 22.35% in Middle Guinea, 22.08% in Upper Guinea, 20.54% in Forested Guinea, and 10.27% in Conakry. The highest percentage of children (29.86%) were between 48 and 59 months, and the lowest (6.52%) were between 0 and 11 months. According to the socio-economic status of the household (Wealth index), 24.68% had a disadvantaged standard of living (poor level) and 12.04% of households had a good level (rich level). Additionally, 74.74% of the mothers of the children and 66.73% of household heads have not attended formal education. In the majority of the households (98.93%), mosquito nets were hung up, and 98.43% of the nets observed were in good conditions. The ethnic group distribution of the household heads shows that 36.80% were Peul, 25.49% were Malinke, 16.67% were Soussou, 7.17% were Guerze or Kono or Mano, 5.98% were Kissi, 2.38% were Toma, and 5.52% other ethnicity. It is the household heads of the Muslim religion who are more represented (84.44%). In terms of potable water source in the households, 78.42% of them had improved water sources. Thus, 66.73% did not treat drinking water. The analysis of the number of people in the household shows that households with 1–5 people are 40.17%, 36.14% are between 6 and 8, and 23.69% have 9 or more people. 27.06% of households have access to electricity. The household heads who had their own radio and television were 48.41% and 25.53%, respectively. The main materials most used as roof, wall exterior, and floor were metal sheets (74.01%), cement, stone with lime cement, brick, or cement block (70.18%), and cement, grout, or carpet (51.59%), respectively.
Table 2 shows the findings of the bivariate analysis (anemia in children versus administrative region) and binary logistic model by including non-spatial random effects. These results indicate that the prevalence of anemia among children varies according to the region of residence. From the bivariate analysis, the region of Nzerekore (85.29%) has the highest prevalence. On the other hand, in the region of Labe, the prevalence is lower (69.39%).
Table 3 presents the results of bivariate analysis (unadjusted) and binary logistic regression. All the interpretations of the models were done using the odds ratio and corresponding 95% credible intervals.
In view of the results of bivariate analysis, we noticed that the variables associated with the status of anemia in children are: place of residence, administrative region, natural region, age of the child, standard of living of the household, mother’s level of education, ethnicity of household head, religion of household head, household’s access to electricity, and whether the household head has their own television.
Indeed, children from rural areas were more likely to be anemic (OR: 1.59, CI [1.31 1.93]) when compared to those from urban areas. The same observation was made by comparing children from rural areas and those from big cities. Rural children were more likely to have anemia (OR: 1.59, CI [1.24 2.03]). The results also indicate that children in the region of Nzerekore were more likely to have anemia (OR: 1.54, CI [1.09 2.18]) compared to those from Boke region. It appears that children from Conakry were less likely to be anemic (OR: 0.61, CI [0.44 0.84]) than those of Maritime Guinea. In addition, children in the age group of 48–59 months were less likely to be anemic (OR: 0.51, CI [0.34 0.78]) than children in the 0–11 months age group. Regarding education level of mother, the analysis shows that the children of mothers in an advanced level (secondary school or above) were less likely to have anemia (OR: 0.61, CI [0.47 0.79]) compared to the children of mothers that do not have formal education (no educational attainment). As for the standard of living of the household, children from rich households were less likely to be anemic (OR: 0.52, CI [0.38 0.71]) compared to their counterparts in poor households. Children whose household head is of the Peul ethnic group were less likely to have anemia (OR: 0.66, CI [0.50 0.86]) compared to the children whose head is Soussou. Results also indicate that children whose household head is animist or no religion were less likely to be anemic (OR: 0.80, CI [0.61 1.04]) compared to the children whose head is Muslim. If the household did not have electricity, children in the household were more likely to have anemia (OR: 1.50, CI [1.23 1.83]) than those in the household with electricity. Children whose households do not own a television are more likely to have anemia (OR: 1.51, CI [1.23 1.84]) compared to those whose households own a television.
3.4. Factors Associated with Anemia in Children from the Spatial Models
In
Table 4, we have the factors associated with anemia among children in Guinea after controlling for the non-spatial random effects (M2), spatial random effects (M3), and both non-spatial and spatial random effects (M4). These models were implemented using WinBUGS version 1.4 (MRC Biostatistics Unit, Cambridge, UK)
Significant risk factors shown in
Figure 1 of the model incorporating the non-spatial random effects (M2) were included in the binary logistic model. For example, in the model (M1), children aged 48–59 months were less likely to have anemia (OR: 0.46, CI [0.30 0.71]) compared to those who are younger (0–11 months). Adjustment for non-spatial random effects have provided a protective effect against anemia for this age group of children, which reduces the odds of being anemic by 53% (OR: 0.47, CI [0.29 0.70]). Children whose mothers attained secondary school or above in education had a reduced chance of being anemia positive 33% (OR: 0.67, CI [0.49 0.90]) compared to children of mothers who did not have formal education (no educational attainment). Children who are under the responsibility of household heads from Peul ethnic group after controlling for non-spatial random effect were associated with anemia among children as well. They are less likely to have anemia (OR: 0.57, CI [0.41 0.78]) than their counterparts whose leaders are Soussou.
In view of the importance of the mother’s education level and the ethnicity of household head factors, the stratified results for these variables are shown by the standard of living of the household in
Figure A3.
Figure 2 illustrates the results of the model by including non-spatial random effects (M2). We have areas that are perceived as high and low prevalence of anemia among children. The Nzerekore region (yellow color) has the highest prevalence, and the Conakry region (maroon color) has the lowest prevalence. However, seeing these prevalences, all areas are considered to have a high prevalence of anemia among children. This map was obtained from the results of the Bayesian analysis.
Appendix ATable A1 presents the posterior means (relative risk) and standard deviation (sd) for non-spatial random
v and spatial random effects
u. The results presented as relative risk of the non-spatial random effects for the best fitting model are given in the map (
Figure A4). The map shows that the region of Nzerekore (yellow color) has high relative risk (0.11). Moreover, a few clusters (Conakry, Boke, and Kindia ) with moderate relative risk (0.0 to 0.1) are seen in the map (green color). Regarding the posterior standard deviations of the non-spatial random effects, the results show that the region of Nzerekore tends to be higher (0.14) than the other regions. This means the within-region variation of anemia tends to be higher than the rest of the regions after accounting for all the covariate effects. The map of the posterior means for spatial random effects also shows that the regions of Nzerekore, Boke, Kindia, and Conakry (green color) had a moderate relative risk, between 0.0 and 0.1 (
Figure A4).