Clinical profiles and predictors of intensive care unit transfer among children hospitalized with severe Plasmodium falciparum malaria in a tertiary care hospital in Libreville, Gabon: a retrospective study Authors: Bridy Chesly Moutombi-Ditombi, Aude Mariella Lembet Mikolo, Jacques Mari Ndong Ngomo, Noé Patrick M’bondoukwé, Coella Joyce Mihindou, Dimitri Ardin Moussavou Mabika, Denise Patricia Mawili-Mboumba, Simon Jonas Ategbo, Marielle Karine Bouyou-Akotet Published: 2026 ABSTRACT BackgroundSevere Plasmodium falciparum malaria remains a major cause of paediatric morbidity in sub-Saharan Africa. While mortality has declined in many settings, severe disease requiring escalation of care persists. Data on predictors of intensive care unit (ICU) transfer among children with severe malaria remain limited in Central Africa. This study aimed to describe the clinical and biological profiles of paediatric severe malaria in an urban Gabonese setting and to identify factors associated with ICU transfer as an early marker of clinical deterioration.MethodsA retrospective analytical study was conducted at the Centre Hospitalier Universitaire Mère-Enfant Fondation Jeanne Ebori (CHUMEFJE) in Libreville, Gabon. Medical records of children below 17 years and hospitalised between January 2021 and July 2022 with microscopically confirmed P. falciparum malaria were reviewed. Severe malaria was defined according to WHO 2014 criteria. ICU transfer among survivors was the primary adverse outcome. Clinical, laboratory, and demographic variables were analysed using univariate and multivariable logistic regression. A cumulative count of WHO severe malaria criteria was used as an indicator of disease severity.ResultsAmong 3.009 hospitalised children, 480 (15.9%) met WHO criteria for severe malaria and were included. The median age was 6 [1-10] years. Overall mortality was low (0.7%), while 8.5% (n = 41) required ICU transfer. Neurological manifestations (64.2%) predominated, particularly prostration (49.2%), impaired consciousness (10.8), and coma (4.2%). ICU transfer was significantly associated with delayed consultation (p = 0.01) and neurological signs (p < 0.01). In multivariable analysis, impaired consciousness (aOR: 14.86; 95%IC [5.58-42.40], p < 0.01) and coma (aOR: 53.3; 95%IC [10.9-178.1], p < 0.01) remained the strongest independent predictors of ICU transfer, whereas isolated biological abnormalities such as severe anaemia or hyperparasitaemia were not. The risk of ICU transfer increased markedly with the number of concurrent severe malaria criteria, especially beyond three criteria (aOR: 10.59; 95%IC [2.38-42.87], p < 0.01).ConclusionICU transfer was frequent and primarily driven by neurological and accumulated severity features. Assessing the number of concurrent WHO criteria offers a pragmatic tool for risk stratification. Although conducted in a referral centre capturing the city’s transmission heterogeneity, multicentre studies are needed to validate these predictors across diverse healthcare systems.Clinical trial numberNot applicable. Introduction Plasmodium falciparum malaria remains a major cause of paediatric morbidity and mortality worldwide, with most cases occurring in sub-Saharan Africa. Despite substantial progress over the past two decades, malaria continues to exert a heavy burden on health systems, particularly among children. In 2024, an estimated 282 million malaria cases were reported globally, of which over 90% occurred in Africa, accounting for most malaria-related deaths [1]. In endemic settings with perennial transmission, children under five years of age have historically been considered the most vulnerable group. However, recent evidence suggests a changing epidemiological pattern, with an increasing proportion of severe malaria cases occurring among older children [2, 3]. This shift may reflect evolving transmission dynamics, partial immunity acquisition, and uneven coverage of preventive interventions, which often prioritise younger age groups. Severe malaria in children is characterised by a wide range of clinical and biological manifestations whose prognostic impact depends not only on their nature but also on their simultaneous occurrence within the same patient. Clinical studies conducted in both endemic and non-endemic settings consistently show that adverse outcomes are more closely related to the accumulation of multiple severity features than to any single criterion considered in isolation [4, 5]. In a recent multicentric study from India, Adigun et al. (2025) reported that neurological impairment, severe anaemia, metabolic disturbances and respiratory compromise frequently coexisted and that the risk of unfavorable outcomes increased with the number of severe manifestations present at admission [6]. Similar observations have been made in East Africa, where prospective and retrospective hospital-based studies from Uganda demonstrated that children presenting with multiple WHO severe malaria criteria were more likely to experience clinical deterioration, prolonged hospitalisation or death [3, 7]. Earlier work from West Africa also highlighted that the clinical expression of severe malaria varies with age and transmission intensity, and that younger and older children may accumulate different combinations of severity features, with important prognostic implications [8, 9]. In non-endemic settings, studies of imported paediatric malaria have similarly shown that the presence of several WHO severity criteria is strongly associated with intensive care unit admission and complications, supporting the clinical relevance of cumulative severity assessment across settings [10, 11]. Together, these findings suggest that severe malaria should be approached as a graded clinical continuum, and that evaluating the number of concurrent severity criteria may provide a more clinically meaningful framework for risk stratification and early identification of children at high risk of poor outcomes, including intensive care unit (ICU) transfer. In such contexts, transfer to the ICU represents a critical clinical decision point, reflecting acute deterioration and the need for advanced life-support measures. ICU transfer is a clinically meaningful outcome that often precedes death and reflects acute clinical deterioration, and it has been used alongside with mortality in predictive models of patient deterioration to enhance statistical power for identifying high-risk patients [12–14]. However, data on the clinical and biological determinants of ICU transfer among children with severe malaria remain scarce in sub-Saharan Africa, particularly in Central African settings. Gabon is characterised by stable and perennial malaria transmission, with Plasmodium falciparum as the predominant species. In Libreville, the capital city, the epidemiological landscape is marked by significant heterogeneity linked to varying levels of urbanisation. Recent data collected from the malaria sentinel surveillance site between 2012 and 2023 among febrile children have revealed the presence of transmission hotspots across different neighborhoods [15]. Between 2021 and 2022, the prevalence of clinical malaria was estimated at 35–36% with higher rates in older children [16]. While several studies have reported on severe malaria in southeastern Gabon, data from Libreville, the capital city, and from specialised paediatric referral centres remain limited [17, 18]. Understanding the factors associated with escalation of care in this setting is essential for optimising triage, improving early management, and guiding allocation of limited intensive care resources. The present study aimed to describe the clinical and biological profiles of children hospitalised with WHO-defined severe Plasmodium falciparum malaria at the Centre Hospitalier Universitaire Mère-Enfant Fondation Jeanne Ebori (CHUMEFJE) in Libreville, Gabon, and to identify factors associated with transfer to the intensive care unit as an early marker of clinical deterioration. Methods Study design and setting This was a retrospective, hospital-based analytical study conducted at the Centre Hospitalier Universitaire de Libreville-Fondation Jeanne Ebori (CHUME-FJE), a tertiary referral hospital dedicated to maternal and child health, located in Libreville, the capital city of Gabon. The paediatric department includes both a general paediatric ward and a paediatric intensive care unit (ICU), allowing stepwise management of severe cases within the same facility. The hospital serves as a major referral centre for Libreville and surrounding peri-urban areas, receiving children referred from primary and secondary healthcare facilities. Libreville is an urban setting characterised by perennial malaria transmission with seasonal fluctuations, primarily due to Plasmodium falciparum. However, several studies have documented significant changes in malaria epidemiology over the past decade, including an increase in malaria prevalence (reaching 36%) and a shift towards older children, likely reflecting declining transmission intensity and delayed acquisition of immunity [16, 19]. In this urban context, malaria transmission remains heterogeneous, influenced by environmental conditions, peri-urban expansion, and vector ecology, resulting in variable exposure across populations. All children were initially managed in the general paediatric ward. The paediatric intensive care unit (ICU) is located within the same department, allowing stepwise escalation of care. Transfer to the ICU was decided by the attending physician in cases of clinical deterioration requiring invasive monitoring, mechanical ventilation, or advanced supportive care not available in the general ward, including management of refractory seizures, haemodynamic instability, or prolonged impairment of consciousness. No standardised protocol for ICU admission was in place during the study period; transfer decisions were therefore based on clinical judgement. Malaria diagnosis was based on routine laboratory procedures available at the hospital, and patient management followed national malaria treatment guidelines. The paediatric unit is staffed by paediatricians, residents, and trained nursing personnel, ensuring continuous clinical monitoring of admitted children. Study population and eligibility criteria Medical records of children aged under 18 years hospitalised between January 2021 and July 2022 in emergency and general paediatric ward, with microscopically confirmed Plasmodium (P.) falciparum malaria were reviewed. Records were eligible for selection if hematological and/or biochemical laboratory results were available. Files from neonates admitted to the neonatal unit and children hospitalised for non-malarial febrile illnesses were excluded. Children were classified as having severe malaria based on the World Health Organization (WHO) 2014 criteria [20]. Only data from children meeting WHO-defined criteria for severe malaria were included in analyses assessing factors associated with ICU transfer. Data collection Electronic case report form developed by Research Electronic Data Capture (RedCap) was used for the electronic record of the following variables [21]: sociodemographic data (age, sex), insecticide-treated nets (ITNs) use; clinical information at admission (time to consultation, clinical signs and symptoms); laboratory test results such as haemoglobin (Hb) level, Plasmodium parasitaemia, blood creatinine, blood glucose; and outcomes (discharge, transfer to ICU, or hospital death). Data completeness and consistency were checked weekly by a dedicated data management team from the Department of Parasitology and Mycology of the Université des Sciences de la Santé, Gabon. Missing data were not imputed, and analyses were conducted using available data for each variable. Laboratory procedures Malaria diagnosis was confirmed based on laboratory records indicating that Giemsa-stained thick and thin blood smears were routinely examined using the Lambaréné method by trained laboratory technicians [22]. Haematological parameters were measured using a Yumizen H550 analyser (HORIBA ABX SAS; Montpellier, France), while biochemical tests, including renal and blood glucose levels, were performed using a Cobas C311 analyser (Roche® Diagnostic). Operational definitions Severe malaria was defined according to the World Health Organization (WHO) 2014 criteria [20]. These include: severe anaemia (Hb < 5 g/dL), prostration, shock, incoercible vomiting, convulsions (≥ 2 within 24 h), impaired consciousness (Blantyre coma score < 3 or pediatric Glasgow coma score < 10), clinical jaundice, pulmonary oedema or acute respiratory distress syndrome, haemoglobinuria, hyperparasitaemia (P. falciparum parasitaemia > 10%), hypoglycaemia (< 2.2 mmol/L or < 40 mg/dL). Acute kidney injury (AKI) was defined using age-adjusted serum creatinine thresholds in accordance with the kidney disease: Improving Global Outcomes (KDIGO) criteria for paediatric populations [23]. As baseline creatinine values prior to admission were unavailable, absolute creatinine values adapted to age were used. Haemoglobinuria was detected macroscopically, based on the observation of dark reddish or port wine-coloured urine. Total bilirubin was only available for three patients; thus, jaundice was defined based on the presence of yellow coloration of mucous membrane recorded in patient file. Hyperlactatemia and acidosis were not assessed as these measurements are not part of routine laboratory analysis in the CHUMEFJE. Age was analysed both as a continuous variable and categorised into two groups (< 5 years and ≥ 5 years), in line with standard paediatric malaria risk stratification. Consultation delay was defined as the time interval between symptom onset and first hospital presentation and was categorised into ≤ 3 days, 4-7 days, and > 7 days. Prostration Inability of the child to sit, stand, or feed (drink or suckle) normally according to age. Impaired consciousness Blantyre Coma Score (BCS) between 3 and 4, or paediatric Glasgow Coma Scale (pGCS) score between 9 and 14, indicating reduced responsiveness without meeting the criteria for coma. Coma Blantyre Coma Score ≤ 2 or paediatric Glasgow Coma Scale (pGCS) ≤ 8. Shock Defined by clinical signs of impaired peripheral perfusion, such as cold extremities, a capillary refill time > 3 s, or a weak and rapid pulse. Clinical jaundice Defined by the presence of yellow coloration of the sclerae or mucous membranes recorded in the patient’s clinical file. Acute respiratory distress Characterised by laboured breathing, including deep breathing (acidotic breathing), nasal flaring, or intercostal recession. Single malaria infection Patients with Plasmodium parasites detected by microscopy. Co-infection Patient with Plasmodium parasites and other infection confirmed. The number of severe malaria criteria was used as a composite indicator of disease severity. For each child, all WHO severe malaria (SM) criteria present at admission were coded as binary variables (present/absent) and summed to obtain an integer count reflecting the cumulative severity burden. When this variable was analysed as a continuous measure, an effect estimates was expressed by additional severe malaria criterion. Additionally, WHO severe malaria criteria were classified according to their frequency in the study population. Criteria observed in ≥ 10% children whereas those present and in < 10%, were defined as non-frequent. This distinction was used to characterise common versus rare severity profiles and to analyse parterns of co-occurrence and their association with ICU transfer. Outcome definition ICU transfer was used as the primary adverse clinical outcome and was defined as referral to the intensive care unit during hospitalisation in the absence of death. A favorable outcome was defined as hospital discharge without ICU transfer. Fatal outcome was defined as in-hospital death occurring during the same episode of severe malaria. Sample size An exhaustive, consecutive sampling method was employed for medical file selection. The minimum sample size required was estimated using the Schwartz formula:\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\boldsymbol{\mathrm{Schwartz}}\:\boldsymbol{\mathrm{formula}}:\:\boldsymbol{\mathrm{n}}=\left[\boldsymbol{\mathrm{z}}^{\boldsymbol{\mathrm{2}}}\times\:\boldsymbol{\mathrm{p}}\left(\boldsymbol{\mathrm{1}}-\boldsymbol{\mathrm{p}}\right)\right]/{\boldsymbol{\mathrm{d}}}^{\boldsymbol{\mathrm{2}}}$$\end{document} Where: n is the sample size, Z = 1.96 for a 95% confidence interval, p = 20.6%, the hospital prevalence of severe paediatric malaria reported in Franceville, Gabon, d = 5%margin of error [17]. The minimum medical files to be included for analysis was thus 246. Statistical analysis Continuous variables were summarised using medians and interquartile ranges (IQR) or means and standard deviations (SD), as appropriate, and categorical variables as frequencies and percentages. Comparisons between groups were performed using the Mann-Whitney U test or Student’s t-test for continuous variables, and the chi-square test or Fisher’s exact test for categorical variables. Collinearity between impaired consciousness and coma was assessed using variance inflation factors and correlation coefficients. Sensitivity analyses were conducted using coma scores (Blantyre or paediatric Glasgow scales) as ordinal variables in logistic regression models. Factors associated with ICU transfer were first explored using univariable logistic regression. Variables with a p-value < 0.20 or considered clinically relevant were included in a multivariable logistic regression model to estimate adjusted odds ratios (aOR) with 95% confidence intervals (95% CI). The cumulative severity burden was defined as the total number of WHO severe malaria criteria present in each child and analysed both as a continuous variable and as a categorical variable (1-2 vs. ≥ 3 criteria). The association between cumulative severity and ICU transfer was assessed using logistic regression. Pairwise associations between severe malaria criteria (coded as binary variables) were analysed using the phi (ϕ) coefficient, derived from 2 × 2 contingency tables, and complemented by odds ratios (ORs) to estimate the magnitude of co-occurrence. These associations were visualised using a heatmap, where colour intensity reflected the strength and direction of association between criteria. Only observed combinations were included. This approach was used to identify patterns of co-occurrence and clustering of severity criteria within the study population. All analyses were performed using R software (version 4.5.1), and a two-sided p-value < 0.05 was considered statistically significant. Results Study population Between January 2021 and July 2022, a total of 3,009 children were hospitalised at CHUMEFJE. Among them, 591 had microscopically confirmed malaria, of whom 480 (15.9%) met the WHO criteria for severe malaria and were included in the analysis (Fig. 1). Fig. 1Flow chart of study Flow chart of study The median age of the study population was 6 [1-10] years, with 52.5% of the children being under five years age and the sex ratio was 1.06. The sickle cell was reported among 5% (n = 24). Only 22.5% (n = 108) of children usually slept under an insecticide-treated net (ITN), with significantly higher use among children under five years compared to older age groups (p < 0.01). Previous antimalarial self-medication was reported in 10.8% (n = 52) of cases and rates were comparable between older and young (Table I). The median delay between symptom onset and hospital consultation was 4 [3-6] days; it was not significantly different according to age groups. Among them 149 (31.0%) were co-infected. (Table 1: General characteristics of the study population according to age). Severity criteria were predominantly neurological manifestations (89.5%, n = 430), followed by severe anaemia and hyperparasitaemia (Table 1: General characteristics of the study population according to age). When stratified by age, children under five years were significantly more likely to present with prostration (p < 0.01), severe anaemia (p < 0.01), and hyperparasitaemia (p = 0.01). In contrast, children aged five years and older more frequently presented with cerebral malaria including coma (p = 0.01), impaired consciousness (p = 0.01), as well as AKI (p = 0.01). (Table 1: General characteristics of the study population according to age). In sensitivity analyses using neurological scores as ordinal variables, decreasing BCS scores in children under five years were strongly associated with ICU transfer (OR per unit decrease: 5.36; 95%CI: [2.73-10.51]; p < 0.01). Similarly, in older children, decreasing paediatric Glasgow Coma Scale scores were associated with increased risk of ICU transfer (OR: 1.93; 95%CI: [1.57-2.38]; p < 0.01). No collinearity was observed between coma and impaired consciousness (VIF = 1.0; no co-occurrence). Table 1General characteristics of the study population according to ageVariablesTotal (N=480)< 5 years≥5 yearsp-valuen (%)n (%)n (%) Sex 0.4 Female233 (48.5)118 (46,8)115 (50.4) Male247 (51.5)134 (53.2)113 (49.6) ITN use Yes108 (22.5)74 (29.3)34 (14.9) <0.01 Self-medication with ATM* Yes52 (10,8)23 (9.1)29 (12.7) 0.2 Delay in consultation 0.2 0-3 days203 (44.5)112 (46.5)91 (42.3) 4-7 days219 (48.0)116 (48.1)103 (47.9) >7 days34 (7.5)13 (5.4)21 (9.8) WHO-criteria Prostration236 (49.2)96 (38.1)140 (64.4) <0.01 Convulsions122 (25.4)63 (25.0)59 (25.9) 0.8 Vomiting111 (23.1)56 (22.2)55 (24.1) 0.6 Impaired consciousness52 (10.8)19 (7.5)33 (14.5) 0.01 Coma20 (4.2)5 (2.0)15 (6.6) 0.01 Jaundice24 (5.0)13 (5.2)11 (4.8) 0.02 Respiratory distress14 (2.9)11 (4.4)3 (1.3) 0.04 Hemoglobinuria10 (2.1)3 (1.2)7 (3.1) 0.1 Severe anemia141 (29.4)120 (47.6)21 (9.2) <0.01 Hyperparasitemia109 (22.7)83 (32.9)26 (11.4) 0.02 AKI20 (5.8)5 (2.0)15 (6.6) 0.01 Hypoglycemia16 (3.3)10 (28.3)6 (15.4) 0.2 ATM*: antimalarials; IC**: impaired consciousness; data available for self-medication: N = 456; WHO criteria version 2014 Table 1General characteristics of the study population according to ageVariablesTotal (N=480)< 5 years≥5 yearsp-valuen (%)n (%)n (%) Sex 0.4 Female233 (48.5)118 (46,8)115 (50.4) Male247 (51.5)134 (53.2)113 (49.6) ITN use Yes108 (22.5)74 (29.3)34 (14.9) <0.01 Self-medication with ATM* Yes52 (10,8)23 (9.1)29 (12.7) 0.2 Delay in consultation 0.2 0-3 days203 (44.5)112 (46.5)91 (42.3) 4-7 days219 (48.0)116 (48.1)103 (47.9) >7 days34 (7.5)13 (5.4)21 (9.8) WHO-criteria Prostration236 (49.2)96 (38.1)140 (64.4) <0.01 Convulsions122 (25.4)63 (25.0)59 (25.9) 0.8 Vomiting111 (23.1)56 (22.2)55 (24.1) 0.6 Impaired consciousness52 (10.8)19 (7.5)33 (14.5) 0.01 Coma20 (4.2)5 (2.0)15 (6.6) 0.01 Jaundice24 (5.0)13 (5.2)11 (4.8) 0.02 Respiratory distress14 (2.9)11 (4.4)3 (1.3) 0.04 Hemoglobinuria10 (2.1)3 (1.2)7 (3.1) 0.1 Severe anemia141 (29.4)120 (47.6)21 (9.2) <0.01 Hyperparasitemia109 (22.7)83 (32.9)26 (11.4) 0.02 AKI20 (5.8)5 (2.0)15 (6.6) 0.01 Hypoglycemia16 (3.3)10 (28.3)6 (15.4) 0.2 ATM*: antimalarials; IC**: impaired consciousness; data available for self-medication: N = 456; WHO criteria version 2014 General characteristics of the study population according to age ATM*: antimalarials; IC**: impaired consciousness; data available for self-medication: N = 456; WHO criteria version 2014 Clinical outcomes and ICU transfer Among the 480 children hospitalised for severe malaria, 90.8% (n = 436) had a favorable outcome, while 8.5% (n = 41) were transferred to the intensive care unit. Three deaths (0.6%) were recorded. They were six (6) months, three (3) and four (4) years old. Each one presented with a single WHO severe malaria criterion at admission and one (4.2%) had sickle cell. Two children had severe anaemia as the primary severity criterion, while one presented with convulsions. They all had concomitant comorbidity, one child with severe anaemia had concomitant meningitis, the second one had septicaemia, and the child presenting with convulsions developed convulsive status epilepticus. Factors associated with ICU transfer In univariate analysis, ICU transfer was significantly more frequent in children aged ≥ 5 years compared with those under 5 years (p = 0.03). An increasing trend in ICU transfer rates was observed with longer consultation delays, rising from 6.0% for delays ≤ 3 days to 20.6% for delays > 7 days (p = 0.01). Neurological manifestations were strongly associated with ICU transfer, namely convulsions, impaired consciousness, and coma (p < 0.01). AKI was also more frequent among ICU-transferred children (p < 0.01), while respiratory distress showed a borderline association (p = 0.08). In contrast, severe anaemia, hyperparasitaemia, vomiting, hypoglycaemia, jaundice and hemoglobinuria were not risk factors for ICU transfer (Table 2: Risk factors of ICU transfer). Biological parameters according to outcome and coinfection status Parasitaemia, haemoglobin, and platelet counts according to clinical outcome and coinfection status are illustrated in Fig. 2. Parasitaemia differed significantly across outcome groups (p = 0.021). Children with a favourable outcome had a median parasitaemia of 40,810 [7,700 - 187,880] parasites/µL, compared with 6,930 [1,350 - 79,310] parasites/µL among those transferred to the ICU and 83,584 [42,215 - 124,952] parasites/µL in fatal cases. In contrast, haemoglobin (p = 0.16) and platelet counts did not differ significantly between outcome groups (p = 0.47). Overall, thrombocytopenia was observed in 49.2% (n = 215/437) of patients. Median haemoglobin levels were 7.3 [4.5-9.9] g/dL in children with a favourable outcome, 8.6 [6.8-9.8] g/dL in those transferred to the ICU, and 4.7 [4.6-6.9] g/dL in fatal cases. Median platelet counts were 102.5 × 10³/µL [58.0-184.2], 84.0 × 10³/µL [38.5-174.5], and 134.0 × 10³/µL [70.5–137.0] in the favorable, ICU transfer, and death groups, respectively. Parasitaemia was higher among children with coinfection (77,000 [5,390 - 244,660] parasites/µL compared with those without coinfection (36,575 [6,160 - 154,000] parasites/µL), although this difference was not statistically significant (p = 0.30) (Fig. 2). Among anaemic patients, 3.4% (n=1/29) had haemoglobin levels below 3 g/dL, and 6.3% (n = 7/111) had levels between 3 and 5 g/dL. Unexpectedly, ICU transfer was more frequent among children with haemoglobin levels above 11 g/dL (11.8% vs. 8.0%; p = 0.04). Creatinine values among patients with elevated levels showed limited variability, precluding further stratification according to severity. Fig. 2Distribution of parasitaemia, haemoglobin, and platelet counts according to clinical outcome and coinfection status in children with severe malaria. (A) Parasitaemia (log₁₀ parasites/µL) according to clinical outcome (favourable outcome, ICU transfer, death). (B) Haemoglobin levels (g/dL) according to clinical outcome. (C) Platelet count (×10³/µL) according to clinical outcome. (D) Parasitaemia (log₁₀ parasites/µL) according to coinfection status (coinfection vs. no coinfection) Distribution of parasitaemia, haemoglobin, and platelet counts according to clinical outcome and coinfection status in children with severe malaria. (A) Parasitaemia (log₁₀ parasites/µL) according to clinical outcome (favourable outcome, ICU transfer, death). (B) Haemoglobin levels (g/dL) according to clinical outcome. (C) Platelet count (×10³/µL) according to clinical outcome. (D) Parasitaemia (log₁₀ parasites/µL) according to coinfection status (coinfection vs. no coinfection) Multivariable analysis of predictors of ICU transfer In multivariate analysis, delayed consultation of more than one week and neurological manifestations remained the main independent predictors of ICU transfer (Table 2: Risk factors of ICU transfer). Children with neurological manifestations showed a strong, severity-dependent increase in the risk of ICU admission, with adjusted odds ratios indicating an approximately sixfold increase for convulsions, a fifteenfold increase for impaired consciousness, and an extremely elevated risk in the presence of coma (Table 2: Risk factors of ICU transfer). In contrast, after adjustment, age, severe anaemia, hyperparasitaemia, vomiting, and AKI were no longer independently associated with ICU transfer, suggesting that escalation of care was primarily driven by neurological severity rather than isolated biological abnormalities. (Table 2: Risk factors of ICU transfer) Table 2Risk factors of ICU transferVariablesFavorable (N=436)ICU transfer (N=41)cOR[95%IC] p aOR[95%IC] p Age < 5 years234 (94.0)15 (6.0)11 ≥5 years202 (88.6)26 (11.4)2.01[1.03-3.90] 0.03 0.95[0.36-2.5] 0.05 Sex Female217 (93.5)15 (6.5)1.72[0.89-3.33] 0.10 1.97[0.80-4.8] 0.10 Male219 (89.4)26 (10.6)1 Delay in consultation 0-3 days189 (94.0)12 (6.0)11 4-7 days197 (90.4)21 (9.6)1.68[0.80-3.51] 0.17 2.50[0.90-7.38] 0.20 >7 days27 (79.4)7 (20.6)4.08[1.48-11.27] 0.01 5.30[1.01-2.54] 0.01 WHO-Criteria Prostration221 (94.0)14 (6.0)0.50[0.25-0.97] 0.04 Convulsions99 (22.7)22 (53.7)3.94[2.05-7.58] <0.01 6.00[2.25-17.34] <0.01 Vomiting103 (24.5)3 (7.3)0.24[0.07-0.80] 0.01 - Impaired consciousness38 (8.7)14 (34.1)5.43[2.58 -11.41] <0.01 14.86[5.58-42.40] <0.01 Coma2 (0.5)18 (43.9)15.6[3.32 -72.89] <0.01 53.3[10.9-178.1] <0.01 Jaundice24 (100.0)0 (0.0)NA Respiratory distress11 (2.5)3 (7.3)3.05[0.82-11.38] 0.08 4.40[0.67-28.8] 0.12 Hemoglobinuria10 (100.0)0 (0.0)NA Severe anemia134 (96.4)5 (3.6)0.31[0.12-0.82] 0.01 0.57[0.15-2.16] 0.40 Hyperparasitemia103 (23.6)6 (14.6)0.56[0.23-1.38] 0.21 AKI15 (3.4)5 (12.2)3.90[1.28-11.90] 0.07 3.11[0.49-15.12] 0.37 Hypoglycemia13 (3.0)3 (7.3)2.57[0.68-9.75] 0.20 - Table 2Risk factors of ICU transferVariablesFavorable (N=436)ICU transfer (N=41)cOR[95%IC] p aOR[95%IC] p Age < 5 years234 (94.0)15 (6.0)11 ≥5 years202 (88.6)26 (11.4)2.01[1.03-3.90] 0.03 0.95[0.36-2.5] 0.05 Sex Female217 (93.5)15 (6.5)1.72[0.89-3.33] 0.10 1.97[0.80-4.8] 0.10 Male219 (89.4)26 (10.6)1 Delay in consultation 0-3 days189 (94.0)12 (6.0)11 4-7 days197 (90.4)21 (9.6)1.68[0.80-3.51] 0.17 2.50[0.90-7.38] 0.20 >7 days27 (79.4)7 (20.6)4.08[1.48-11.27] 0.01 5.30[1.01-2.54] 0.01 WHO-Criteria Prostration221 (94.0)14 (6.0)0.50[0.25-0.97] 0.04 Convulsions99 (22.7)22 (53.7)3.94[2.05-7.58] <0.01 6.00[2.25-17.34] <0.01 Vomiting103 (24.5)3 (7.3)0.24[0.07-0.80] 0.01 - Impaired consciousness38 (8.7)14 (34.1)5.43[2.58 -11.41] <0.01 14.86[5.58-42.40] <0.01 Coma2 (0.5)18 (43.9)15.6[3.32 -72.89] <0.01 53.3[10.9-178.1] <0.01 Jaundice24 (100.0)0 (0.0)NA Respiratory distress11 (2.5)3 (7.3)3.05[0.82-11.38] 0.08 4.40[0.67-28.8] 0.12 Hemoglobinuria10 (100.0)0 (0.0)NA Severe anemia134 (96.4)5 (3.6)0.31[0.12-0.82] 0.01 0.57[0.15-2.16] 0.40 Hyperparasitemia103 (23.6)6 (14.6)0.56[0.23-1.38] 0.21 AKI15 (3.4)5 (12.2)3.90[1.28-11.90] 0.07 3.11[0.49-15.12] 0.37 Hypoglycemia13 (3.0)3 (7.3)2.57[0.68-9.75] 0.20 - Risk factors of ICU transfer Co-occurrence patterns of severe malaria criteria in the overall population The analysis of co-occurring WHO SM criteria in the overall population revealed a higher rate of combination involving two SM criteria (Fig. 3). The most prevalent two-criterion combination was vomiting associated with prostration, observed in 63 (13.1%) children, followed by convulsions with prostration in 48 children (10.0%) (Fig. 3). The association of SMA with hyperparasitaemia was identified in 39 (8.1%) cases, while prostration combined with hyperparasitaemia accounted for 32 (6.7%) cases. Other frequent pairings included vomiting with hyperparasitaemia (n = 28/480; 5.8%), convulsions with hyperparasitaemia (n = 27/480; 5.6%), vomiting with SMA (n = 23/480; 4.8%) and convulsions with SMA (n = 19/480; 4.0%) (Fig. 3). Several three-criterion combinations were also observed, notably prostration with vomiting and convulsions (n = 18/480; 3.8%) and SMA with hyperparasitaemia and prostration (n = 16/480; 3.3%) (Fig. 3). Combinations involving at least one non-frequent severe malaria criterion were less common but remained consistently represented across the population. These included convulsions with impaired consciousness (n = 24/480; 5.0%), impaired consciousness with prostration (n = 14/480; 2.9%), and associations combining haematological and visceral involvement such as SMA with jaundice (n = 16/480; 3.3%) and prostration with AKI (n = 10/480; 2.1%). More complex combinations involving three criteria or more were uncommon, each affecting fewer than 3% of children. The most frequent triad was convulsions with impaired consciousness and prostration, observed in 11 (2.3%) children. Other multi-criteria profiles included SMA combined with hyperparasitaemia and jaundice (n = 9/480; 1.9%), convulsions with SMA and hyperparasitaemia (n = 8/480; 1.7%), prostration with hyperparasitaemia and AKI (n = 7/480; 1.5%), and SMA associated with hyperparasitaemia and dark urine (n = 6/480; 1.3%) (Fig. 3). Fig. 3Heatmap of pairwise associations between WHO severe malaria criteria in the overall paediatric population (N = 480). (Legend: Cell colors indicate the strength of association between criteria, as specified, with warmer colors reflecting stronger positive associations) Heatmap of pairwise associations between WHO severe malaria criteria in the overall paediatric population (N = 480). (Legend: Cell colors indicate the strength of association between criteria, as specified, with warmer colors reflecting stronger positive associations) Co-occurrence of the most frequent WHO severe malaria criteria according to age group Age-stratified analyses revealed distinct patterns of co-occurrence. Among children aged < 5 years (n = 252), the most frequent combination was SMA with hyperparasitaemia, affecting 35 (13.9%) children, followed by prostration with convulsions (n = 27/252; 10.7%), incoercible vomiting with hyperparasitaemia (n = 24/252; 9.5%) and prostration with incoercible vomiting (6.0%) (Fig. 4a). In contrast, among children aged ≥ 5 years (n = 228), prostration with vomiting was the predominant combination, observed in 44 (19.3%) cases, followed by prostration with convulsions (n = 21/228; 9.2%) (Fig. 3b). Combinations involving altered neurological status were more frequently identified in this older age group, including convulsions with impaired consciousness in 13 (5.7%) children (Fig. 4b). Fig. 4aCo-occurrence of severe malaria criteria in children under five years of age Co-occurrence of severe malaria criteria in children under five years of age Fig. 4bCo-occurrence of severe malaria criteria in children aged five years and older Co-occurrence of severe malaria criteria in children aged five years and older Risk of ICU transfer according to the number of severe malaria criteria Among ICU transfers, combinations were strongly dominated by neurological clustering. The most frequent ICU patterns were convulsions+impaired consciousness (n = 10/41, 24.4%), convulsions+ prostration (n = 6/41, 14.6%), and convulsions+coma (n = 5/41, 12.2%), compared with 3.2% (n = 14/436), 9.4%; (n = 41/436) and 0.5% (n = 2/436) among non-ICU children, respectively. Multi-criteria neurological profiles were uncommon but concentrated in ICU transfers, including convulsions+coma+prostration (n = 2/41, 4.9%) and convulsions+prostration+vomiting (n = 1/41, 2.4%). Neuro-haematological and multiorgan combinations were less frequent but still in ICU transfers. Children requiring ICU admission presented a higher mean number of severe malaria criteria (2.15 ± 0.96) compared with those managed outside ICU (1.75 ± 0.79) (p < 0.01). The difference in distributions remained statistically significant when assessed using a Welch t-test (p = 0.01). Compared with children presenting with a single WHO severity criterion, those with two criteria did not show a significant increase in ICU transfer risk after adjustment (Table 3: Risk of ICU transfer according to the number of severe malaria criteria). The presence of three severe malaria criteria was associated with a threefold risk of ICU transfer, after in multivariable analysis. The highest risk was observed among children presenting with four or more severity criteria, who were tenfold at risk to be transferred to ICU (Table 3: Risk of ICU transfer according to the number of severe malaria criteria). Table 3Risk of ICU transfer according to the number of severe malaria criteriaNumber of severity criteriaNo-ICU transfer, n (%)ICU transfer n, (%)aOR[95% CI]p-value 1 188 (42.8)12 (29.3)1-- 2 188 (42.8)15 (36.6) 1.32 [0.59-3.01] 0.49 3 50 (11.4)10 (24.4) 3.41 [1.27-8.96] 0.01 ≥ 4 13 (3.0)4 (9.7) 10.59 [2.38-42.87] < 0.01 Table 3Risk of ICU transfer according to the number of severe malaria criteriaNumber of severity criteriaNo-ICU transfer, n (%)ICU transfer n, (%)aOR[95% CI]p-value 1 188 (42.8)12 (29.3)1-- 2 188 (42.8)15 (36.6) 1.32 [0.59-3.01] 0.49 3 50 (11.4)10 (24.4) 3.41 [1.27-8.96] 0.01 ≥ 4 13 (3.0)4 (9.7) 10.59 [2.38-42.87] < 0.01 Risk of ICU transfer according to the number of severe malaria criteria Discussion This study updates the clinical profile of paediatric severe malaria in an urban setting of Gabon, during a period that included the COVID-19 era and provides a quantification of severity through a cumulative count of WHO severe malaria criteria. Importantly, the last detailed urban paediatric severe malaria descriptions from Libreville, the capital city of Gabon were published more than a decade ago, notably by Dzeing-Ella and colleagues, who reported a clinical spectrum dominated by severe anaemia and respiratory distress, within a context of high transmission and distinct healthcare pathways [24]. Since then, several reports have documented an epidemiological shift in Gabon, with changes in malaria burden and a progressive age shift in risk toward children > 5 years, consistent with evolving transmission intensity and control measures [17, 19]. Our focus on the COVID-period was motivated by concerns that restrictions and health-system disruptions could alter care-seeking, delays to consultation, and clinical severity at presentation, mechanisms widely discussed for malaria services in sub-Saharan Africa [25, 26]. Unlike most studies conducted in sub-Saharan Africa that focus primarily on mortality or case fatality rates, our work highlights ICU transfer as an early, operational, and clinically meaningful marker of disease progression and severity [27–30]. Although the overall in-hospital mortality was low (0.7%), nearly one in thirteen children with severe malaria required ICU-level care. This finding highlights an important epidemiological and clinical reality: a declining malaria mortality does not equate to a reduced burden of severe disease. Instead, it may reflect improvements in early diagnosis, access to effective antimalarial treatment, and hospital-based supportive care, while a substantial burden of severe morbidity persists. In resource-limited settings, ICU transfer represents a critical decision reflecting acute clinical deterioration and the need for advanced supportive care. It often precedes fatal outcomes and constitutes a more frequent and statistically robust endpoint than mortality, particularly in tertiary hospitals where early interventions reduce case fatality [31]. By focusing on ICU transfer, this study would provide actionable information for frontline clinicians and hospital systems, enabling earlier identification of children at high risk of deterioration and more efficient allocation of limited intensive care resources. Children requiring ICU transfer presented markedly different severity profiles compared with those managed outside ICU. ICU admission was predominantly associated with combinations of neurological criteria, particularly those involving repeated convulsions, impaired consciousness, and coma, either alone or in combination with prostration. In contrast, isolated biological abnormalities such as severe anaemia or hyperparasitaemia without neurological involvement were less predictive of ICU transfer [20]. These observations are consistent with paediatric studies from Central and West Africa, which identify altered consciousness, recurrent seizures, and coma as key drivers of clinical deterioration and escalation of care [20, 24, 32, 33]. These findings support the integration of neurological assessment for patient triage as priority criteria for referral and escalation of care in paediatric malaria management protocols. The use of neurological scores as a continuous severity criterion is also suggested, although categorical WHO definitions remain operationally relevant. Children older than five years, particularly those over ten years of age, were proportionally more likely to require ICU transfer. This age-related vulnerability contrasts with the traditional perception that severe malaria predominantly affects younger children and may reflect evolving transmission dynamics, partial immunity acquisition, and behavioural factors influencing healthcare-seeking practices [2]. Age-related patterns further contextualise these findings. In urban Gabon, older children are now the most affected by uncomplicated malaria, reflecting shifting transmission dynamics [16, 19]. Our data suggest that this group may also be more susceptible to severe neurological forms, possibly due to waning anti-disease immunity, whereas younger children continue to benefit from sustained vector control and free access to care. Delayed consultation was more frequent among older children and was associated with increased odds of ICU transfer in bivariate analyses. This delay may be linked to lower prioritisation of older children in malaria prevention strategies, greater reliance on self-medication, and delayed recognition of danger signs by caregivers [16, 34]. These findings highlight the need to expand malaria awareness campaigns beyond the under-five population and to emphasise early consultation for febrile illness across all paediatric age groups. Taken together, the differential impact of severe malaria criteria on ICU transfer should be interpreted according to their clinical reversibility. In this urban setting, hypoglycaemia and severe anaemia, although frequent, are rapidly identified and effectively managed with glucose infusion and blood transfusion, which may explain their limited association with ICU admission. Notably, parasitaemia was lower in ICU-transferred children than in those with a favourable outcome, reflecting that ICU transfer was driven by neurological criteria, coma and impaired consciousness, rather than by parasite burden per se, consistent with the pathophysiology of cerebral malaria. Indeed, neurological manifestations (convulsions, impaired consciousness, coma) reflect acute cerebral dysfunction, for which therapeutic options are limited, making them the principal drivers of ICU transfer. Markers of advanced organ dysfunction, such as haemoglobinuria and acute kidney injury, were rarely observed, consistent with previous urban data from Gabon and Central Africa, where such complications are uncommon [24]. A key contribution of this study is the demonstration of an increasing relationship between the cumulative burden of severe malaria criteria and the risk of ICU transfer. Children transferred to ICU presented with a higher number of concurrent severity criteria, and the likelihood of ICU admission increased markedly beyond a threshold of three criteria. Thus, severe malaria severity is better conceptualised as a cumulative process rather than a dichotomous condition defined by any single WHO criterion. Additionally, a transition towards clinical instability and multisystem involvement could also be suggested. This dose-response relationship, which persisted after adjustment, highlights the added value of considering the overall severity burden when assessing risk of deterioration. Therefore, in resource-constrained hospital settings, where ICU capacity is limited, incorporating cumulative severity assessment may help identify children at highest risk of deterioration and prioritise early escalation of care. Finally, mortality rate was markedly lower than rates reported in many African settings [17, 29, 35]. This favourable outcome would reflect improved access to parenteral artesunate, earlier referral to a tertiary care centre, and the expertise of paediatric clinicians in managing severe malaria cases before ICU transfer becomes necessary. Furthermore, in Libreville, comparatively better access to healthcare and greater availability of medical resources than in other country settings may have contributed to improved outcomes. The availability of free malaria diagnosis in sentinel surveillance sites of Libreville may also promote earlier healthcare-seeking and timely treatment, thereby reducing disease severity and associated mortality Importantly, three of the four deaths occurred at the emergency unit before ICU transfer could be arranged, illustrating the rapid progression of severe malaria and reinforcing the need for early identification of high-risk patients. These findings emphasise the time-critical nature of severe malaria, particularly when neurological signs are present, as deterioration may occur rapidly and unpredictably. This highlights the need for immediate clinical recognition, prompt initiation of parenteral artesunate, and early triage to higher levels of care. Several factors may have influenced the observed associations. Delayed consultation, which is a well-established determinant of disease severity, may have contributed to both the accumulation of severity criteria and the likelihood of ICU transfer. Age-related differences in acquired immunity in urban settings could also confound the relationship between age, severity patterns, and clinical outcomes. In addition, pre-hospital treatment practices, including self-medication, may have modified clinical presentation at admission. Biological factors such as comorbidities or coinfections, as well as referral bias inherent to hospital-based studies, may further have influenced the distribution of severity criteria. Finally, ICU transfer reflects clinical decision-making and resource availability, and may therefore be subject to confounding by indication, particularly favouring neurologically severe presentation. The present finding implications for both clinical practice and public health. From a clinical perspective, these results support the use of cumulative severity assessment at admission as a pragmatic tool to identify children at increased risk of clinical worsening. In particular, the presence of three or more WHO severe malaria criteria, especially when neurological manifestations are involved, should prompt early escalation of care, closer monitoring, and anticipation of ICU transfer, even in the absence of coma at presentation. From a public health point, a shift in vulnerability towards older children in urban and semi-urban settings is now observed, likely reflecting changing transmission dynamics and waning anti-disease immunity. There is a need to adapt malaria control and triage strategies beyond the under-five age group, including targeted prevention, timely care-seeking, and strengthened referral pathways. Incorporating cumulative severity into routine assessment may help optimise resource allocation and improve outcomes in settings with limited ICU capacity. The relevance of these findings to other settings should be interpreted in light of local healthcare context. While improved access to care, availability of ICU services, and free malaria diagnosis in urban Libreville may have contributed to lower mortality and earlier management, the identified predictors of clinical deterioration, particularly neurological manifestations and cumulative severity burden, are likely to be generalisable across malaria-endemic regions. However, in rural or resource-limited settings where delays in care and limited ICU access are more pronounced, these factors may be associated with higher mortality rather than ICU transfer. These considerations highlight the need to adapt triage and management strategies to local health system capacity. Some limitations should nonetheless be acknowledged. The study was conducted at a single tertiary hospital, which may limit generalisability in terms of healthcare infrastructure and clinical practices. However, CHUMEFJE serves a diverse paediatric population from both urban and peri-urban areas of Libreville. Given the known micro-heterogeneity of malaria transmission in the capital, where peripheral areas often act as high-transmission pockets, our cohort likely represents a wide range of epidemiological profiles and levels of acquired immunity. While this diversity strengthens the relevance of our findings, external validation in different geographical settings and across primary and secondary healthcare facilities is needed to confirm the robustness and generalisability of these predictors.Its retrospective design may have led to under-reporting or misclassification of some criteria, particularly those requiring repeated assessment or specialised laboratory tests. Indeed, not all WHO severity markers were available, particularly those related to metabolic acidosis, such as blood lactate levels or acide-base status, which were not routinely measured at CHUMEFJE. Consequently, the cumulative severity variable was analysed as an unweighted count, although individual criteria may differ in prognostic importance. Metabolic acidosis is widely recognised as one of the strongest predictors of fatal outcomes in severe malaria. However, its absence in our analysis should be interpreted in light of the very low mortality rate observed. The rare fatal cases were primarily associated with severe comorbidities, such as meningitis or septicaemia, rather than isolated metabolic failure. Furthermore, all children effectively transferred to the ICU had favourable outcomes, suggesting that the clinical markers and cumulative severity feature identified served as efficient triggers for escalation of care, even in the absence of advanced metabolic monitoring.This indicates that the clinical markers and cumulative severity feature we identified served as efficient triggers for escalation of care, ensuring survival even in the absence of advanced metabolic monitoring. Furthermore, nutritional status was not assessed using standardised Z-scores in this study. Malnutrition is a known contributor to malaria-related morbidity and mortality; therefore, its absence from our analysis is a limitation. However, given the primary focus on clinical predictors of ICU transfer and the exceptionally low mortality rate in our series, the impact of nutritional status as a confounder, while possible, was likely limited in this specific context. Finally, the small number of deaths precluded inferential mortality analyses. The absence of formal, protocol-based ICU admission criteria at CHUMEFJE during the study period means that transfer decisions were based on clinical judgement, introducing potential for inter-physician variability. However, the strong and consistent association observed between ICU transfer and the most severe neurological manifestations, coma and impaired consciousness, supports the clinical validity of the outcome definition and argues against systematic misclassification. Nevertheless, this study provides a contemporary, detailed description of paediatric severe malaria in an urban Central African setting, more than a decade after the last comparable hospital-based reports. The large sample size allowed robust age-stratified analyses and an exploration of co-occurring WHO severe malaria criteria. Importantly, the use of ICU transfer among survivors as a clinically meaningful outcome enabled the investigation of severity gradients beyond mortality alone. The introduction of a cumulative count of severe malaria criteria offers a pragmatic approach to capture overall severity burden, closely aligned with routine clinical practice. Conclusion In this cohort of children hospitalised with severe malaria in Libreville, transfer to the ICU was frequent despite low mortality and was primarily driven by neurological manifestations, particularly coma. ICU transfer represents a critical and clinically relevant marker of disease severity that complements mortality-based assessments. These findings emphasise the importance of early recognition of neurological deterioration and support the integration of the number of WHO SM criteria ICU transfer predictors into paediatric malaria triage and management protocols in endemic settings.