Prostration and the prognosis of death in African children with severe malaria

Abstract

Objectives: Malaria is still one of the main reasons for hospitalization in children living in sub-Saharan Africa. Rapid risk stratification at admission is essential for optimal medical care and improved prognosis. Whereas coma, deep breathing, and, to a lesser degree, severe anemia are established predictors of malaria-related death, the value of assessing prostration for risk stratification is less certain.

Methods: Here we used a retrospective multi-center analysis comprising over 33,000 hospitalized children from four large studies, including two observational studies from the Severe Malaria in African Children network, a randomized controlled treatment study, and the phase-3-clinical RTS,S-malaria vaccine trial, to evaluate known risk factors of mortality and with a specific emphasis on the role of prostration.

Results: Despite comparable age profiles of the participants, we found significant inter- and intra-study variation in the incidence of fatal malaria as well as in the derived risk ratios associated with the four risk factors: coma, deep breathing, anemia, and prostration. Despite pronounced variations, prostration was significantly associated with an increased risk of mortality (P <0.001) and its consideration resulted in improved predictive performance, both in a multivariate model and a univariate model based on the Lambaréné Organ Dysfunction Score.

Conclusion: Prostration is an important clinical criterion to determine severe pediatric malaria with possible fatal outcomes.

Keywords: Coma; Deep breathing; Mortality; Prostration; Severe malaria.

Fig. 1

Fatality rate vs presence of risk factors. Despite pronounced variation in fatality rates between the different studies (A) and between sites of the Observational Study (B), there is a general trend toward an increased risk of death with higher percentage of individuals recorded with one or more risk factors.

Fig. 2. Risk factor distribution associated with outcome.

(A, B) Proportional representation of the number of individuals recorded with a particular risk factor, stratified by outcome (death yes/no) and study site (all studies, left; observational study, right). (C, D) Venn diagrams illustrating a general trend toward being positive for multiple risk factors in children who died of malaria.

Fig. 3

Relative risk ratios of the four risk factors stratified by study site. There is significant variation in the risk of death associated with the presence/absence of the risk factors under consideration, spanning almost an order of magnitude (risk ratio displayed on a log scale). Note, all diseased individuals in the AQUAMAT study were positive for prostration, precluding the calculation of a meaningful risk ratio.

Fig. 4

Prostration is significantly associated with increased mortality risk. Comparison of estimated effect sizes (OR) and model performances based on all data (A, B) and ObservStudy, only (C, D). Different colors correspond to individual studies (blue: ObservStudy, red: TreatmentRCT, green: VaccineRCT) and individual sites, respectively (blue: Banjul, red: Blantyre, green: Kilifi, light blue: Kumasi, purple: Lambaréné, orange: Libreville). Pooled effect sizes are derived by inverse-variance meta-analysis. Receiver operator characteristic curves are based on generalized linear mixed models with (orange) and without prostration (blue). AUC, area under the receiver operator characteristic curve; OR, odds ratio.

Fig. 5. Outcome-stratified LODS-distributions.

(A, B) Whereas most of individuals who survived were recorded with no or only one risk factor (LODS 0 or 1), more than half of those who died had LOD scores of 2 or 3. (C, D) Prostration has a positive effect on model performance when included in LOD score; LODS2 does not include prostration. AUC, area under the receiver operator characteristic curve; LODS, Lambaréné Organ Dysfunction Score.