Clinical Prediction Models for Pediatric In-Hospital Death Risk in Congolese Severe Malaria Children Using Machine Learning Based-Algorithms

Severe Malaria

Severe malaria, Paediatric in-hospital mortality risk, Machine learning-based algorithms, Predictive model, Plasmodium falciparum, Multiple Organ Dysfunction Syndrome (MODS), Metabolic acidosis

The goal of this observational study is to optimize the management of severe childhood malaria, based on understanding and controlling the severity factors of the disease in Congolese children aged 2 to 9 years (the age group at risk of developing various severe forms of malaria), admitted to the paediatric intensive care units (PICU). The main question it aims to answer is whether the scores or models used to predict the severity of severe malaria and the associated risk of mortality accurate enough to warrant early interventions, including treatments, on their own? Thus, investigators aim to fill three knowledge gaps associated with the following hypotheses: Hypothesis-1: Children with severe malaria show signs of disease severity based on their severity scores on admission. Higher severity scores on admission are associated with a higher risk of mortality. Hypothesis-2: Validation of the predictive power and transferability of severe malaria severity scores to additional independent populations is needed to support their clinical utility. Hypothesis-3: The severity of the clinical and biological changes induced by plasmodium depends not only on the ability of the parasite to invade and grow in the host organism, but also and above all on the number of parasites present in the host (parasitemia). For any child admitted to the PICU and meeting the inclusion criteria, as part of clinical care, investigators proceeded before any treatment: 1. An arterial blood sample of 3 ml by puncture of the radial artery for instant arterial blood gaz as well as for venous biochemistry, including albumin, phosphate, chlorine, magnesium, urea, creatinine and total bilirubin dosages, and, 2. A one-drop finger pulp blood test for parasitemia measurement and the rapid diagnosis test for plasmodium falciparum. Then, the diagnostic parameters of acid-base disorders will be calculated, including AG (anion gap), AGCAP (AG corrected for albumin and phosphate plasmatic concentrations), SIG (Strong ion gap), SBE (Standard base excess) and SBDCAP (Standard base deficit corrected for albumin and phosphate plasmatic concentrations).

Background: Severe malaria has associated with a high risk of paediatric hospital mortality in resource-constrained countries, which remains deplorable. Improved methods of risk-stratification can assist in referral decision making and resource allocation. Investigators aim to i) create prediction model for in-hospital mortality risk among children presenting with severe malaria and compare its predictive performance to the current models, ii) validate the latter, and iii) assess the plasmodium-induced changes in clinical and biological parameters. Methods: This is a retrospective study of data collected prospectively during a period from January 30, 2017 to August 01, 2025, from children with severe malaria, admitted to the PICU of the Monkole Hospital Center (MHC) and the Kimbondo Pediatric Center (KPC), all in Kinshasa, DR. Congo. Baseline clinical and laboratory variables were collected on enrolled children. The primary outcome is death up to 1 week post-admission, and the second outcome, the length of stay in pediatric intensive care following admission for severe malaria. Machine learning algorithms will be employed to accomplish the three specific research objectives. Expected Results: In line with research objectives, the following results are expected: 1. The prevalence of Multiple Organ Dysfunction Syndrome (MODS) and metabolic acidosis in children presenting with severe malaria will be determined. 2. A novel model for predicting associated mortality risk of severe malaria will be developed: 1. This novel model will be based on predictors of disease severity and will measure: * The degree of severity of MODS and metabolic acidosis. * The length of stay for severe malaria in the PICU * The risk of death following hospitalization for severe malaria * The influence of parasitemia on disease severity 2. The performance of the proposed novel model will be measured 3. The predictive nomogram and scoring system will be associated with it. 3. Investigators validate and compare the performance of existing models for predicting severe malaria severity against the proposed novel model. Together, research data will provide proof of principle supporting early interventions and treatment choices in children presenting with severe malaria.

No related papers found yet.

Democratic Republic of the Congo