Associations between health-related quality of life and physical domain scores with sociodemographic, parasitological, and clinical variables from multivariate regression analysis.

Multivariate regression models with random effects to account for clustering and a stepwise backward elimination approach were utilized to identify explanatory variables, which most significantly influence the children's overall quality of life and physical domain scores. Initial models included sociodemographic (e.g., sex, age group, socioeconomic status, residential area (rural or urban), and ecozone), parasitological (by infection intensity for each species investigated), and clinical (anemia, wasting, stunting, fever, hepatomegaly, and splenomegaly) variables. During stepwise removal, variable categories were combined, based on expert knowledge and logical deduction, before eventually eliminating the respective variable. Remaining explanatories were included at a significance level of p<0.2. Quality of life and domain scores were pre-transformed into a scale from 0 to 100, thus coefficients correspond to percentages of change. For variables on concurrent infections (Plasmodium-helminth co-infection and number of concurrent parasitic infections), the single parasite variables were exchanged by the concurrent infections variables but the same additional explanatories as for the single species models were used.
CI = confidence interval.
aReference categories for explanatory variables: sex = female; age group = 5–10 years; wealth quintile = wealthier quintiles (top 80%); ecozone = South; Plasmodium = no or low parasitemia (<1,000 parasites/µl of blood); anemia = not anemic; S. haematobium = no microhematuria; soil-transmitted helminths = not infected; any form of malnutrition = neither stunted, nor wasted, nor underweight (Z score>−2); Plasmodium-helminth co-infected = neither infected with any of the two; number of concurrent infections = not infected with any investigated parasite species.
bDomain 1 showed ceiling effects >15%. Tobit regression models were therefore built additionally for comparison. Except for ecozone and wealth quintile, the same significant relationships were identified in the tobit regression models as in the linear regression analysis presented above.
*Statistically significant (p<0.05).
Associations between health-related quality of life and physical domain scores with sociodemographic, parasitological, and clinical variables from multivariate regression analysis.