Prompt and accurate diagnosis and parasite density quantification are keys to effective malaria treatment. The malaria diagnostic and parasite density quantification methods currently in use are not without their shortcomings. Thus, there is the need for new methods with high sensitivity, specificity and predictive values that may complement other methods in use. In this study, a hundred (100) children (1–10 years) diagnosed of severe malaria were treated with 2.4 mg/kg body weight of artesunate intravenously at 0 hour and then 1.2 mg/kg body weight at 12, 24 and 48 hours, followed by artemether-lumefantrine combination therapy at doses of 4 mg/kg body weight of artemether and 24 mg/kg body weight of lumefantrine for three days, with 200 clinically healthy children of the same age range serving as control. The diagnostic potential of selected biochemical parameters in the serum of patients before treatment (day 0), 48 h of treatment (day 2) and 48 h after treatment (day 7) were evaluated. Using regression analysis, different relationships between the identified predictors and the predicted variable (parasite density) were tested. Serum glutathione reductase (GR) activity at > 85.78 U/l (cut-off value) was indicative of severe malaria and had the highest diagnostic values with 98.5% sensitivity, 96% specificity, 98.01% positive predictive value, 4.04% negative predictive value and 1576 as odds ratio, with an area under curve of 0.999, comparing favourably with microscopy (gold standard) at 95% confidence interval. This was followed by elevated serum glutathione peroxidase activity. Other parameters had relatively high sensitivity with relatively low specificity or vice versa. The polynomial quadratic equation, parasite density (x103 parasites/µl) = 6.96[GR] − 433.1 − 0.01[GR]2 had the best good-fit for prediction of parasite density, using serum GR activity. Thus, serum GR activity may be a very effective diagnostic and parasite density prediction tool for severe falciparum malaria in children.