OUP user menu

Severe P. falciparum malaria in Kenyan children: evidence for hypovolaemia

K. Maitland, M. Levin, M. English, S. Mithwani, N. Peshu, K. Marsh, C.R.J.C. Newton
DOI: http://dx.doi.org/10.1093/qjmed/hcg077 427-434 First published online: 1 June 2003


Background: The role of volume resuscitation in severe Plasmodium falciparum malaria is controversial.

Aim: To examine the role of hypovolaemia in severe childhood malaria.

Study design: Retrospective review.

Methods: We studied 515 children admitted with severe malaria to a high-dependency unit (HDU) in Kilifi, Kenya. On admission to the HDU, children underwent a further assessment of vital signs and a standard clinical examination.

Results: Factors associated with a fatal outcome included deep breathing or acidosis (base excess below –8), hypotension (systolic blood pressure <80 mmHg), raised plasma creatinine (>80 μmol/l), low oxygen saturation (<90%), dehydration and hypoglycaemia (<2.5 mmol/l). Shock was present in 212/372 (57%) children, of whom 37 (17.5%) died, and was absent in 160, of who only 7 (4.4%) died (χ2 = 14.9; p = 0.001).

Discussion: Impaired tissue perfusion may play a role in the mortality of severe malaria. Moreover, volume resuscitation, an important life-saving intervention in children with hypovolaemia, should be considered in severe malaria with evidence of impaired tissue perfusion.


The current case fatality of severe malaria in African children presenting to hospital is between 10 and 30%, some 50% of these deaths occurring within 12 h of admission.1,,2 Thus, a significant reduction in case fatality of severe malaria can only be expected from early treatment. Ancillary therapies evaluated to date have failed to improve survival.3 In severe malaria, acidosis is the best independent predictor of death, in both adults and children.4–,8 Thus, management strategies targeted at factors contributing to acidosis may improve outcome.

In sick children, the commonest cause of acidosis worldwide is volume depletion (hypovolaemia). Hypotension, the cardinal feature of decompensated shock, has not been reported commonly in case series of children with severe malaria.6,9,,10 However, in children hypovolaemia is frequently occult, and unlike in adults, homeostatic mechanisms usually compensate adequately to maintain blood pressure.11 For example, hypotension is rare in the early stages of septic shock. The features suggestive of hypovolaemia are summarized in Figure 1 (reviewed by Saez-Llorens and McCracken, 199311). These include cardiovascular signs of impaired tissue perfusion and major organ dysfunction such as hypoxia, reduced urine output, agitation, decreased consciousness and seizures, as well as metabolic derangement (including hypoglycaemia and electrolyte imbalance). Thus, severe malaria shares many characteristics in common with the sepsis syndrome. Consideration should therefore be given to therapies with similar physiological goals, in particular volume replacement. Paradoxically, in the absence of overt signs of dehydration, children with severe malaria have traditionally been relatively volume-restricted for fear of causing pulmonary oedema12 and aggravating raised intracranial pressure, a complication of severe malaria.13

Figure 1.

Terminology and definition of septic shock in paediatrics. Adapted with modification from references 11 and 17.

In this study, we looked retrospectively for evidence of hypovolaemia in Kenyan children admitted with severe malaria.


Study site and patients

A team from the Kenya Medical Research Institute (KEMRI) conducted the study at Kilifi District Hospital (KDH). The medical team works in collaboration with the hospital staff to run the paediatric in-patient service, which admits over 5000 children per annum. A high-dependency unit (HDU) is available for the management of children with severe, life-threatening disease. No ventilation facilities are available on the HDU. Medically qualified members of the KEMRI team completed a standard admission questionnaire and examination.

Children with severe malaria were routinely transferred to the HDU. Severe malaria was defined as blood-film-positive P. falciparum plus one or more of the following: prostration (inability to sit or breast feed), coma, prolonged or recurrent seizures, respiratory distress (deep breathing or intercostal recession), circulatory collapse, anaemia (haemoglobin <5 g/dl plus respiratory distress) or jaundice.1 In addition, a small number of children were transferred with hyperparasitaemia >20%, or less severe anaemia (haemoglobin 5–6 g/dl plus respiratory distress). On admission to the HDU, children underwent a further assessment of vital signs and a standard clinical examination. Oxygen saturation and pulse rates were recorded using a pulse oximeter (Nellcor). Other vital signs included a visual assessment of respiratory rate, non-invasive blood pressure (Dynamap, Johnson), core (rectal) and axillary temperatures, assessment of conscious level: prostration (inability to sit or breast feed) and Blantyre Coma Score,14 and pupillary reaction to light. Details of the clinical assessment and standard treatment have been reported elsewhere.7,,15 Children with an oxygen saturation of <95% were given oxygen by nasal prong or mask. All children received maintenance fluids, with 4 ml/kg/h of 4% dextrose/0.18% saline, but in children with signs of moderate dehydration, fluids were increased to account for their deficit. During the period of study, children with deep breathing and/or moderate to severe acidosis were routinely treated with a bolus of normal saline infused over 1 h. In general, volumes of 20 ml/kg were given to those with a base excess exceeding –15 mmol, and 10 ml/kg were given to those with a base excess of –8 to −15 mmol. Children with anaemia and respiratory distress were treated with 20 ml/kg of whole blood.16


Double data entry and range checking were used to verify all data collected. Z-scores for the anthropometric parameters weight-for-age (WA), weight-for-height (WH) and height-for-age (HA) were calculated for each individual using Epi Info v2000 (CDC, Atlanta), which compares data to the National Center for Health Statistics (NCHS) reference values. Dichotomous and categorical variables were created from continuous variables. Derived variables were created from clinical factors defined by Paediatric Advanced Life Support (PALS) guidelines17 as indicating a definitive need for urgent therapeutic intervention and for laboratory variables outside the 95%CIs for normality.18 The following dichotomous variables were created: extreme tachycardia (>180 bpm in children aged 0–8 years or >160 in those aged >8 years); bradycardia (<80 bpm in children aged 0–8 years or <60 in those aged >8 years); hypoxia (oxygen saturation <90% or unable to record by pulse oximeter); tachypnoea (respiratory rate >60 breaths/min at any age); hypothermia (rectal temperature <36°C); capillary refill ≥3 s; hypotension (systolic blood pressure <80 mmHg in children >1 year and <70 mmHg in those <1 year or unrecordable); acidosis (base excess <−8)19; elevated creatinine (>80 μmol/l if >1 year or >110 μmol/l if <1 year); hypoglycaemia (<2.5 mmol/l); hyperglycaemia (>10 mmol/l); hypokalaemia (<3 mmol/l); hyperkalaemia (>5.5 mmol/l) or hyponatraemia (<125 mmol/l).18 Categorical variables were created for WHZ scores (>0, >−2, −2 to −3, <−3). Analysis used SPSS V8 (SPSS, UK) and STATA V6 (Timberlake Ltd, UK). Univariate odds ratios (95%CIs) were produced for binary variables using death as the outcome measure. Multivariate logistic regression was used to estimate the relative contribution of each feature to fatal outcome. The receiver operator curve (ROC) was used to assess the ability of the shock scoring system to predict a fatal outcome.



Between 1 September 1999 and 31 December 2000, 515 parasitaemic children were transferred to the HDU with clinical evidence of severe malaria (Table 1). Seventy percent of malaria admissions were aged <36 months. The overall mortality rate was 12.8% (66/515). Mortality was higher in children aged <1 year (16%), who accounted for 29% of all the deaths, and in children aged 48–59 months (21%), although this latter group only accounted for 11% of the total deaths. A high proportion of children had evidence of malnutrition: 24% had a WHZ score <−2. Moreover, both under-nutrition (WHZ score –2 to –3) and severe malnutrition (WHZ score <−3) were associated with a higher mortality (16% and 19%, respectively). Severe malarial anaemia (haemoglobin <5 g/dl was more common in those aged <2 years, while impaired consciousness was more frequent in those aged >3 years (Figure 2). The frequency of acidosis did not vary significantly with age.

View this table:
Table 1

General characteristics of the children transferred to the HDU with suspected malaria

Factorn% of subgroupMortality n (%)
Overall group51566 (12.8%)
Clinical definition
Cerebralb40077.752 (12.9%)
Acidosisc28956.152 (18.0%)
Severe anaemiad12925.023 (18.3%)
Visible jaundice173.32 (11.9%)
Hypoglycaemiae58/47812.116 (27.6%)
Age group (months)a
< 3112.12 (18.2%)
3–1110320.016 (15.5%)
12–2311822.912 (10.2%)
24–3510420.214 (13.5%)
36–475510.78 (14.5%)
48–59346.67 (20.7%)
60+5210.13 (5.8%)
Male26551.534 (12.8%)
Female25048.532 (12.8%)
WHZ score474
≥05110.72 (3.9%)
< 0 SD28459.925 (8.8%)
< –2 SD11323.818 (15.9%)
< –3 SD265.55 (19.2%)
  • aAges were unknown in 38 (7.4%) children. bImpaired consciousness defined either as prostrate or Blantyre Coma Score ≤2. cAcidosis defined as a base excess of −8 or more and/or deep breathing. dSevere anaemia: haemoglobin < 5 g/dl. eHypoglycaemia: blood glucose < 2.5 mmol/l.

Figure 2.

Features of severe malaria by age group. The numbers per syndrome is greater than the overall number of individuals since most children fit more than one diagnostic group.

Mortality and clinical syndrome

The presence of acidosis (base excess <−8, with or without deep breathing) was by itself associated with a 15% case fatality. Among the acidotic children, mortality increased in association with severe anaemia (21%) but not with impaired consciousness (14%) (Figure 3). Mortality was lowest amongst children with only a single feature of the traditionally regarded groups of severe malaria: impaired consciousness (prostrate or Blantyre Coma Score ≤2) 9/134 (6.7%) or severe malarial anaemia 0/7 (0%). However, not all children with severe malarial anaemia were transferred to the HDU for blood transfusion; in those admitted to the general paediatric ward during the same period with malaria and a haemoglobin <5 g/dl (n = 442) without respiratory distress, the mortality rate was 5.2%. As described previously,20 the aetiology of such deaths is probably multi-factorial. Deep breathing was a good clinical indicator of acidosis (91% positive predictive value), but its sensitivity was poor: only 88 children (32%) with a base deficit <–8 had clinical evidence of deep breathing. Hypoglycaemia was present in 51/410 children (12.4%). Of these, only five (9.8%) had impaired consciousness alone (case fatality 40%). The majority (43, 84.3%) had hypoglycaemia with impaired consciousness and acidosis (base excess <−8). Mortality was highest (7/15; 47%) in those with acidosis, impaired consciousness, anaemia and hypoglycaemia.

Figure 3.

Syndromic distribution of 515 children transferred to HDU with severe malaria (mortality shown in boxes).

Clinical and biochemical correlates with fatal outcome

The vital signs, clinical characteristics and laboratory variables at admission were examined for features designated as requiring urgent intervention (Table 2). Taken individually, tachypnoea, hypoxia (recorded by a pulse oximeter), and deep breathing were present in 16–20% of children, and were associated with considerable mortality (20%, 30% and 31%, respectively). Similarly, children with features suggestive of decompensated shock (hypotension, and elevated creatinine) or intracellular fluid depletion (sunken eyes or decreased skin turgor), all had an associated case fatality of >25%. At admission, 31 children had an unrecordable oxygen saturation (by pulse oximetry) and 61 an unrecordable blood pressure. Of these, 16/31 (51.6%) and 29/61 (47.5%) died, respectively. Other markers of impaired tissue perfusion (capillary refill time ≥3 s and extreme tachycardia) were common, but in isolation were not strongly predictive of a fatal outcome. Severely deranged biochemical values that were associated with a significantly inflated case fatality included hypoglycaemia (<2.5 mmol/l), severe acidaemia (pH <7.2) and hyperkalaemia (potassium >5.5 mmol/l).

Figure 4

Receiver Operator Curve for shock scoring system.

Shock and dehydration

We devised an objective scoring system for ‘shock’, adapted from published criteria,11 and applied this retrospectively to the data (Table 3). Toe-core temperature gap (>5°C) and pulse volume were excluded from this index, since these data were not collected routinely. Hypoglycaemia and impaired consciousness were also excluded, since they are cardinal features of severe malaria. Mortality increased linearly with shock score. In 372 children with complete data, shock (score ≥2) was present in 212 (57%), of whom 37 (17.5%) died, and was absent in 160 (score <2), of whom seven (4%) died (χ2 for trend 23.6, p <0.001). Thus, a shock score of ≥2 identified 37/44 (84%) deaths: ROC (95%CI) 0.714 (0.634–0.794); p <0.0001. Numerically, the largest contribution to the score was from biochemical or clinical features of acidosis (67% of the group scoring 2, 83% of those with scores of 3–5 and 90% of children scoring 6 or more). The second largest contributions came from delayed capillary refill time (33%, 55% and 75% of children with scores of 2, 3–5 and ≥6, respectively) and creatinine >80 μmol/l (28%, 37%, 38% of those with scores of 2, 3–5 and ≥6, respectively). The contributions of severe tachypnoea, tachycardia and hypothermia to the shock score were mainly limited to those with higher scores. In children with a shock score of 6 or more, 86% had decompensated shock (systolic BP <80 mmHg or unrecordable) 69% had oxygen saturation <90%, and 48% were assessed to have dehydration.

View this table:
Table 2

Triage parameters and electrolyte derangement suggesting urgent need for intervention in the critically ill child

FeatureValid (%)Fatality rate (%)
Airway and respiration
Hypoxia  (O2 saturation < 90%)a 86/501 (17.1)30
Respiratory rate  >60 breaths/min 83/501 (16.6)20
Deep breathing104/515 (20.2)31
Cardiovascular and hydration
Tachycardia (>160 beats/min) 81/503 (16.1)17
Bradycardia (< 80 beats/min) 5/503 (1)40
Hypotension (< 80 mmHg) 66/507 (13)26
Capillary refill ≥3 s165/496 (33.3)15
Sunken eyes 36/514 (7)28
Decreased skin turgor 29/514 (5.6)28
Convulsions303/515 (58.8)8.3
Prostration353/514 (68.7)13
Blantyre Coma Score ≤2266/509 (52.3)16
Severe acidaemia (pH < 7.2) 96/436 (22)36
Elevated creatinine  (>80 mmol/l) 96/469 (20.4)26
Hypoglycaemia (< 2.5 mmol/l) 58/478 (12.1)28
Hyperglycaemia (>10 mmol/l) 47/478 (9.8)19
Hyperkalaemia (>5.5 mmol/l) 61/493 (12.3)28
Hyponatraemia (< 125 mmol/l) 21/493 (4.3)14
Hypokalaemia (< 3 mmol/l) 16/493 (3.2)6
  • aMeasured by pulse oximeter.

View this table:
Table 3

Clinical index of shock in children with severe malaria

No. of factors...n0123–56+χ2 for trendp
Shock (419)37257103869729
Died n (%) 44 2 (3.5)  5 (4.9)10 (11.6)16 (16.5)11 (37.9)23.6< 0.0001
  • Shock score determined as presence of hypotension or two or more of the following features: hypoxia, tachypnoea, delayed capillary refill, hypothermia, acidosis (and/or deep breathing), elevated creatinine or dehydration (sunken fontenelle and/or sunken eyes). Hypotension was scored as two points; the remaining features as one point each.

Predictors of fatal outcome

The major factors associated with a fatal outcome on univariate analysis were acidosis (or its clinical correlate, deep breathing), and those variables considered part of the hypovolaemia complex (Table 4). By contrast, the following variables were not significantly associated with a fatal outcome: age group; sex; impaired consciousness; jaundice; extreme tachycardia; bradycardia; tachypnoea; hypothermia; capillary refill time; hypokalaemia; hyponatraemia and hyperglycaemia. Four factors were associated with a fatal outcome on multiple logistic regression: oxygen saturation <90%; deep breathing; raised creatinine (>80 μmol/l) and hypoglycaemia (<2.5 mmol/l) (Table 4). Interactive terms between these four factors were not significant and did not materially alter the findings, except for an interactive term between raised creatinine and deep breathing: OR 0.18 (0.05–0.70), p = 0.013. Adjusted odds ratios are shown in Table 4.

View this table:
Table 4

Variables associated with a fatal outcome

FeatureUnivariate OR (95%CI)pAdjusted MLR* (95%CI)p
Deep breathing 5.1 (2.9–9.0) < 0.0001 6.65 (2.93-15.05) < 0.0001
O2 saturation < 90% 4.7 (2.6–9.0) < 0.0001 4.49 (2.19-9.18) < 0.0001
Hypoglycaemia < 2.5 mmol/l 3.4 (1.7–6.7) < 0.0001 2.25 (1.01-5.0) 0.05
Creatinine >80 μmol/l 3.0 (1.7–5.4) < 0.0001 5.76 (2.31-14.33) 0.0002
Potassium >5.5 mmol/l 3.0 (1.6–5.7) < 0.0001 –
Sunken eyes 3.0 (1.3–6.5) 0.005 –
Hypotension < 80 mmHg 2.8 (1.5–5.2) 0.001 –
WAZ < –3 (malnutrition) 2.3 (1.2–4.1) 0.006 –
Acidosis (BE<–8) 2.3 (1.2–4.4) 0.01 –
Anaemia (Hb <5 g/dl) 1.7 (1.0–3.1) 0.036 –
  • *Multivariate logistic regression.


This study provides considerable support for the hypothesis that hypovolaemia is an important, treatable complication of severe malaria. Many of the clinical and laboratory features associated with a poor outcome in the current study are recognizable consequences of impaired tissue perfusion. These include acidosis (base excess <–8 and/or its clinical correlate, deep breathing), hypotension, dehydration and elevated creatinine. Ancillary features regarded as part of the clinical picture of the sepsis syndrome,11 including hypoxia, hypoglycaemia and electrolyte disturbances, were also important factors in predicting a fatal outcome. Overall, in children with complete data, shock (hypotension or two or more other features; Table 3) was present in 84% (37/44) of deaths. Elevated creatinine, biochemical or clinical features of acidosis and delayed capillary refill time were numerically the three major contributors to the shock score of ≥2. In the context of critical illness management, the presence of any one of these three features, irrespective of aetiology, would be sufficient evidence to consider a carefully administered fluid challenge.

Acidosis was present in 90% of the children who died. Furthermore, severe acidaemia (blood pH <7.2) was present in 96/436 (22%) of cases, of whom 35/96 (36%) died. At a pH <7.2, basic enzymic processes and physiological mechanisms falter and, if left uncorrected, have profound effects upon myocardial contractility and arterial and venous tone, which further aggravate impaired perfusion.21 Hyperkalaemia was one of the risk factors for fatal outcome, but it is likely that acidosis contributed to the increased serum potassium through compensatory cellular exchange of hydrogen and potassium ions, and through reduced renal excretion.22 Much of the work on the aetiology of acidosis in severe malaria has focused on lactic acidaemia.2,5,,6 However, previous studies conducted in Kilifi suggest that the acidosis is only partially attributable to lactic acid.23,,24 This accords with observations in Vietnamese adults with severe malaria, in whom acidosis was attributable to a number of factors, including lactic acidosis, impaired renal function and reduced hepatic metabolism.8

Hypoxia (pulse oximetry saturations <90% or unrecordable) and raised creatinine were both found to affect prognosis adversely in the current study. Hypoxia was not confirmed by measurement of arterial blood pO2. It is unclear, therefore, how many children had genuine hypoxia. However, given the evidence for hypovolaemia in many of these patients, it is likely that inadequate tissue perfusion was largely responsible for the observed ‘hypoxia’, and it is noteworthy that mortality in those with an unrecordable saturation was >50%. Although alternative explanations for hypoxia include intrapulmonary ventilation-perfusion mismatch (e.g. pneumonia, pulmonary oedema or adult-type respiratory distress syndrome), hypoventilation (due to impaired consciousness and/or intercurrent convulsions), or profound anaemia, we believe that reduced tissue perfusion is the most likely explanation for a number of reasons. First, both pulmonary oedema and ARDS are thought to be rare in children with severe malaria.6,9,,10 Second, while hypoxia was more common in children with clinical evidence of convulsions (59/328, 18.6%) than in those without (21/171, 12.3%; χ2 = 2.7, p = 0.099), hypoxia did not alter the prognosis in these patients. Similarly, although mortality was significantly more common amongst children in whom impaired consciousness was associated with hypoxia (65/259, 25.1%) than in those without hypoxia (15/237, 6.3%; χ2  =  32.2, p < 0.0001), hypercapnia (venous pCO2 >6) (a feature of hypoventilation) was not more common in the hypoxic group. Finally, although severe malarial anaemia leads to impaired tissue oxygenation and acidosis, decreased oxygen saturation (measured by pulse oximetry) is not a recognized feature.23 Studies currently underway point to transient hypoxia that resolves with simple strategies such as establishing a clear airway, controlling convulsions and treatment of hypovolaemia (Maitland, unpublished data).

A raised plasma creatinine concentration was an important, independent predictor of death. Creatinine is a good marker of renal perfusion and function. In critically ill children, renal impairment is often multifactorial in origin, but is usually reversible if treated by volume resuscitation early in the course of the disease. In the current case series, elevated creatinine levels normalized in surviving children with standard management, which included fluid administration, suggesting that hypovolaemia (leading to reversible impairment of renal perfusion) was involved. Our current findings contrast with two previous studies that examined renal function in African children with mild25,,26 and cerebral malaria,26 and found no significant association between signs of renal impairment and death. Elevated urea and creatinine were attributed to dehydration in a previous study in Kilifi, but prognosis was not examined.27 However, raised plasma creatinine was an independent predictor of mortality in children in Papua New Guinea,6 and studies in Vietnamese adults have shown that base deficit was the best correlate with a fatal outcome, with serum creatinine and lactate accounting for most of the variation in serum base deficit.8 The main mechanisms of malaria-induced renal damage in adults appear to be tubular. Haemodynamic and oxygen transport studies have shown evidence of reduced renal blood flow and oxygen delivery,28 and features of acute renal tubular necrosis have been identified in a histopathological study of malaria-associated acute renal failure.29 Thus, although renal impairment (increased creatinine) was associated with mortality in our study, this may reflect hypovolaemia, since renal failure has not been demonstrated in other descriptive studies of severe malaria in African children,4,9,,10 and in survivors the elevated creatinine resolved with routine management.

During the course of this study, our routine management of children admitted with severe malaria and acidosis included a bolus infusion of normal saline. This contrasts with standard practice in many centres across Africa, where fluids are restricted to 80 ml/kg/day,4 a volume that would, in a febrile child, represent modest fluid restriction. The judicious fluid management practiced in many centres principally reflects concerns relating to cardiogenic shock and cerebral oedema. We suggest that the more generous volumes of fluid administered in our series may account for our low overall mortality (12.8%). The most appropriate way of establishing the role of volume replacement is through a carefully conducted randomized controlled trial. However, before any such studies can be undertaken, it is essential that pilot studies first be conducted to establish the safety of the intervention.

In summary, we present evidence that hypovolaemia is a feature of severe childhood malaria, and that it is likely to be a major factor in the aetiology of acidosis. We suggest that severe malaria shares many features in common with the sepsis syndrome, and that consideration should therefore be given to therapies with similar physiological goals, in particular volume replacement, as potentially life-saving interventions.


The authors would like to thank the staff of Kilifi District Hospital and KEMRI/Wellcome Trust, Kilifi, particularly the medical and nursing staff on the KEMRI ‘high-dependency’ Unit. We thank Dr Tom Williams for reviewing the manuscript. K. Maitland, M. English, K. Marsh and C.R.J.C. Newton (No. 050533) are supported by the Wellcome Trust, UK. This manuscript is published with the permission of the Director of KEMRI.


View Abstract