Socio-demographic data of guardians
A total of 274 subjects were interviewed to identify factors that influence child health in malaria-endemic districts in the Central region of Ghana. Majority of the guardians (73.4%) were above 25 years of age with 216 (78.8%) and 58 (21.2%) representing Females and males respectively. The educational level of the participants ranges from informal education to tertiary education. Participants with informal education were 52 (19.0%), middle school/JSS 106 (38.7%), secondary school/vocational education 97 (35.4%), and tertiary education 12 (4.4%). The main occupations of the study participants were trading/business 92 (33.6%), pension 9 (3.3%), farming 55 (20.1%), hairdressing/seamstress 42 (15.3%) whilst the remaining 20 (7.3%) were unemployed (Table 1).
Clinical data of children
274 children whose Guardians were enrolled in the study were diagnosed with malaria and presented in mild to severe clinical conditions. About 190 (69.34%) of the children were within the age category 0-5 years whilst 84 (30.66%) were within 6-15 years. The median clinical presentations were; parasite density 89923 (64056-149541) parasites/mm3, haemoglobin concentration 8.40 (4.70-11.60) g/dl, and platelet count 126 (87-146) x10^9/L of blood (Table 2).
Factors influencing child health
The factors that influence child health were identified using both clinical data and interviews. The clinical condition of the children whose guardians were enrolled in the study ranges from mild clinical condition 154 (56.20%) to severe clinical condition 34 (12.41%). The main clinical symptoms of the children that were identified by the guardians included high body temperature 233 (85.04%), less activeness 153 (55.84%), loss of appetite 189 (68.98%), and excessive crying 154 (56.20%). Most of the guardians sought treatment for their children at the hospital 89 (32.5%) whilst the rest resorted to drug/chemical shops 60 (21.9%), clinic/health centres 58 (21.2%) and pharmacy shops 51 (18.6%). Majority of the guardians waited 2-3 days after clinical symptoms of their wards before seeking medical treatment 181 (66.1%). Guardians also adopted some form of malaria prevention intervention such as the usage of mosquito net, mosquito spraying, and mosquito coils (Table 3).
Effect of guardians’ socio-demographic characteristics on factors of child health
Guardians’ socio-demographic characteristics were modelled to test for its canonic effects on the factors that influence child health using multivariate GLM statistics. The multivariate GLM test showed significant association between guardians’ socio-demographic characteristics and the factors that influence child health [CSOC F (6,516) =41.059a, Wilks’ L=0.45, p<0.0001; FPOCCH F (12, 682.895) =12.541, Wilks’ L=0.59, p<0.0001; DOOSCS F (6, 516) =2.506b, Wilks’ L=0.944, p=0.021; CS F (3, 258) =36.523, Wilks’ L=0.702, p<0.0001; and MCID F (9, 628.055) =57.908, Wilks’ L=0.23, p<0.0001] (Table 4). To identify the specific guardians’ socio-demographic characteristics that have an overall effect on the factors that influence child health, the associations were further tested using type-III ANOVA statistics. The analysis indicated that CSOC (F=98.997, SS=33.224, df=2), FPOCCH (F=29.838, SS=20.028, df=4), DOOSCS (F=7.655, SS=2.569, df=2), CS (F=109.977, SS=18.455, df=1) and MCID (F=11.827, SS=5.954, df=3) were all influenced by the occupational status of the guardian, p<0.0001. On the other hand, CSOC, FPOCCH, and MCID were also influenced by educational level of guardians with p values of <0.0001, 0.076, and <0.0001 respectively. It was also revealed that FPOCCH and MCID were the only factors that were influenced by the guardian age category, p<0.0001. However, the gender of the guardians did not show any statistical significance with the factors influencing child health (Table 5).
Patterns between guardians’ socio-demographic characteristics and factors influencing child health
The Wilk’s Lambda statistics showed that there is some significant proportion of variance in the factors that influence child health which could not be explained by the guardians’ socio-demographic characteristics (Table 4). To further model the association to have some insight into the underlying patterns within the data, we employed multilayer perceptron network analysis. The model predicted strong confidence indicating a suitable pattern between guardians’ socio-demographic characteristics and the factors influencing child health (SSE=13.899, 186 (67.9%) in training, and 88 (32.1%) in testing from the total sample of 274). The relative error for scale dependents in training sets were OCSG=0.043, EDULG=0.058, SOG=0.0, and AGECOG=0.050 with only 3.8% (SSE=13.899, error=0.038). The overall misclassification of training sets was 4.3% (SSE=9.193, error=0.043) compared to testing sets with scale dependents determined OCSG=0.053, EDULG=0.047, SOG=0.0, and AGECOG=0.08. This predicts 96.2% and 95.7% accuracy for the association between guardians’ socio-demographic characteristics and factors that influencing child health respectively (Table 6). The graphical presentation of the association between guardians’ socio-demographic characteristics and the factors influencing child health is shown in figure 2. The figure 2 gives a complete insight into the underlying patterns such as biases and hidden factors (H [1:1-9]). The nature of the underlying factors on the associations either reduces the effects of the hidden factors (as shown by blue lines; synaptic weight <0) or increases the effects of the hidden factors (as shown by grey lines; synaptic weight >0). This analysis provides a potential tool for data mining or knowledge discoveries that can impact interventions aimed at improving child health in malaria-endemic areas.
Impact of Socio-economic progenitors on child health
The multilayer perceptron analysis showed that the effects of guardians’ socio-demographic characteristics on child health are influenced by some hidden factors (Figure 2). For further elucidation of the interactions, we propose a network of socio-economic progenitors that impact on child health. Understanding the elements that affect child health can serve as the foundation for the promotion of health interventions. Therefore, the awareness of socio-economic factors has become an indispensable tool to model effective interventional measures. To significantly reduced malaria prevalence among children in the malaria-endemic areas, our ability to predict and evaluate accurately the determinants that can either affect the introduction of health-related interventions negatively or positively is required. Our socio-economic progenitor network provides easy visualisation, identification, prediction and evaluation of factors that affect health-related interventions. The model predicts health-seeking behaviours, issues related to health services, social, cultural, and economic factors as the major hidden factors that affect the association between guardians’ socio-demographic characteristics and the factors influencing child health in malaria-endemic communities (figure 3).