Clinical predictors and outcome of hypoxaemia among under-five diarrhoeal children with or without pneumonia in an urban hospital, Dhaka, Bangladesh


Corresponding Author Mohammod J. Chisti, CSD, ICDDR,B; 68 Shaheed Tajuddin Ahmed Sarani, Mohakhali, Dhaka 1212, Bangladesh. Tel.: +880 2 8860523 32 Extn. 2334; Fax: +880 2 8823116; E-mail:


Objective  To explore the predictors and outcome of hypoxaemia in children under 5 years of age who were hospitalized for the management of diarrhoea in Dhaka, where comorbidities are common.

Methods  In a prospective cohort study, we enrolled all children <5 years of age admitted to the special care ward (SCW) of the Dhaka Hospital of ICDDR,B from September to December 2007. Those who presented with hypoxaemia (SpO2 < 90%) constituted the study group, and those without hypoxaemia formed the comparison group.

Results  A total of 258 children were enrolled, all had diarrhoea. Of the total, 198 (77%) had pneumonia and 106 (41%) had severe malnutrition (<−3 Z-score of weight for age of the median of the National Centre for Health Statistics), 119 (46%) had hypoxaemia and 138 children did not have hypoxaemia at the time of admission. Children with hypoxaemia had a higher probability of a fatal outcome (21%vs. 4%; P < 0.001). Using logistic regression analysis, the independent predictors of hypoxaemia at the time of presentation were lower chest wall indrawing [OR 6.91, 95% confidence intervals (CI) 3.66–13.08, P < 0.001], nasal flaring (OR 3.22, 95% CI 1.45–7.17, P = 0.004) and severe sepsis (OR 4.48, 95% CI 1.62–12.42, P = 0.004).

Conclusion  In this seriously ill population of children with diarrhoea and comorbidities, hypoxaemia was associated with high case–fatality rates. Independent clinical predictors of hypoxaemia in this population, identifiable at the time of admission, were lower chest wall indrawing, nasal flaring and the clinical syndrome of severe sepsis.


Objectif:  Explorer les signes prédictifs et les résultats de l’hypoxémie chez les enfants de moins de cinq ans hospitalisés pour le traitement de la diarrhée à Dacca, où les co-morbidités sont fréquentes.

Méthodes:  Dans une étude de cohorte prospective, nous avons inscrit tous les enfants de moins de cinq ans admis à l’unité des soins spéciaux de l’hôpital de Dhaka de ICDDR, de septembre à décembre 2007. Ceux qui présentaient une hypoxémie (SpO< 90%) constituaient le groupe d’étude et ceux sans hypoxémie formaient le groupe de comparaison.

Résultats:  258 enfants, tous avec diarrhée ont été inscrits. 198 (77%) avaient une pneumonie et 106 (41%) avaient une malnutrition sévère (Z-score du poids pour l’âge de la médiane du NCHS <-3). 119 (46%) enfants avaient une hypoxémie et 138 n’en avaient pas au moment de l’admission. Les enfants avec une hypoxémie avaient une probabilité plus élevée d’une issue fatale (21% vs 4%, p < 0,001). En utilisant l’analyse de régression logistique, les signes prédictifs indépendants de l’hypoxémie au moment de l’admission étaient: la rétraction des muscles sous-costaux (OR: 6,91; IC95%: 3,66 à 13,08, p < 0,001), le battement des ailes du nez (OR: 3,22; IC95%: 1,45 à 7,17, p = 0,004) et un sepsis sévère (OR: 4,48; IC95%: 1,62 à 12,42, p = 0,004).

Conclusion:  Dans cette population d’enfants gravement malades souffrant de diarrhée et de co-morbidités, l’hypoxémie est associée à des taux élevés de létalité. Les signes prédictifs cliniques indépendants de l’hypoxémie dans cette population, identifiables au moment de l’admission étaient: la rétraction des muscles sous-costaux, le battement des ailes du nez et le syndrome clinique de sepsis sévère.


Objetivo:  Explorar los vaticinadores y los resultados de la hipoxemia en niños menores de cinco años hospitalizados con diarrea en Dhaka, en donde las comorbilidades son comunes.

Métodos:  En un estudio prospectivo de cohortes, incluimos a todos los niños menores de cinco años admitidos en la unidad de cuidados especiales (UCE) del hospital ICDDR,B de Dhaka, entre Septiembre y Diciembre del 2007. Aquellos que se presentaban con hipoxemia (SpO< 90%) constituían el grupo de estudio, y aquellos sin hipoxemia conformaban el grupo de comparación.

Resultados:  Se incluyeron 258 niños, todos tenían diarrea. 198 (77%) tenían neumonía y 106 (41%) estaban severamente desnutridos (<-3 Z-score para peso por edad según la media del NCHS). 119 (46%) tenían hipoxemia y 138 niños no tenían hipoxemia en el momento de ser admitidos. Los niños con hipoxemia tenían una mayor probabilidad de tener un resultado fatal (21% vs. 4%; p < 0.001). Utilizando un análisis de regresión logística, los vaticinadores independientes de hipoxemia en el momento de presentarse en el centro eran la retracción torácica inferior (OR 6.91, IC 95% 3.66-13.08, p < 0.001), el aleteo nasal (OR 3.22, IC 95% 1.45-7.17, p = 0.004), y la sepsis severa (OR 4.48, IC. 95% 1.62-12.42, p = 0.004).

Conclusión:  En esta población de niños seriamente enfermos con diarrea y comorbilidades, la hipoxemia estaba asociada con una alta tasa de letalidad. Los vaticinadores clínicos independientes de hipoxemia en esta población, identificables en el momento de la admisión, eran la retracción torácica inferior, el aleteo nasal y el síndrome de sepsis severa.


Hypoxaemia is a common and serious complication in severely ill children (Weber et al. 1997). Most severely ill children with hypoxaemia present with clinical signs of pneumonia (Weber et al. 1997; Lodha et al. 2004), but hypoxaemia can also occur with a number of other illnesses, particularly during the neonatal period (Duke et al. 2002; Subhi et al. 2009). Hypoxemia is one of the major risks of death from pneumonia, and much work has been carried out looking at clinical signs of hypoxaemia in patients with pneumonia (Onyango et al. 1993; Duke et al. 2001; Lodha et al. 2004; Subhi et al. 2009). Although diarrhoea is a common comorbidity in children with pneumonia and together account for 33% of global death in children <5 years of age (Black et al. 2010), little is known about the clinical signs of hypoxaemia in sick children in environments where these two common conditions overlap, particularly where malnutrition is common.

Pulse oximetry is the most reliable, non-invasive, accurate method of measuring arterial haemoglobin oxygen saturation (SpO2) in pneumonia (Duke et al. 2001) and also in other illnesses in children (Moller et al. 1993; Thoms et al. 2007). Nevertheless, many health facilities in developing countries, where the case fatality of sick children, including those with pneumonia, is high, do not have oximetry and experience limited availability and supply of oxygen (Onyango et al. 1993; Weber et al. 1997; Wandi et al. 2006; Matai et al. 2008). Such is generally the situation in most of the health centres and hospitals in Bangladesh.

In Dhaka hospital of ICDDR,B, as in other hospitals in Bangladesh, children often present with combinations of diarrhoea, dehydration, pneumonia, sepsis and malnutrition. However, there are no published data on the clinical predictors and outcome of hypoxaemia in children with diarrhoea who present with pneumonia or other co-morbidities. A better understanding of the predictors of hypoxaemia in this population may improve the therapeutic use of oxygen and reduce morbidity and mortality in a resource-poor setting like Bangladesh. The objectives of this study were to identify the prevalence, clinical predictors and outcomes of hypoxaemia in hospitalized children under the age of 5 years having diarrhoea with pneumonia, malnutrition and other co-morbidities in the Dhaka Hospital of ICDDR,B.


Study design

This was a prospective cohort study in which we enrolled all the children under the age of 5 years with diarrhoea who were admitted to the special care ward (SCW) of the Dhaka Hospital of ICDDR,B, from September to December 2007. The study was approved by the Ethical Review Committee of the centre, and informed verbal consent was obtained from parents or guardians of all participating children.


Each year, Dhaka Hospital of International Centre for Diarrhoeal Disease Research, Bangladesh (ICDDR,B), provides treatment for around 110 000 patients with diarrhoea, with or without associated complications or other health problems. Diarrhoea is the entry point for admission to ICDDR,B. If there are severe acute complications, such as severe dehydration, respiratory distress, cyanosis, apnoea, hypothermia, sepsis, shock, impaired consciousness or convulsion, the children are admitted to the SCW. The vast majority of the patients at ICDDR,B are from poor socio-economic backgrounds from urban and peri-urban Dhaka.


Eligible children were under 5 years of age with diarrhoea admitted to the SCW. Parents of children admitted to the SCW from September to December 2007 were approached for consent. A clinical diagnosis of pneumonia was made according to WHO criteria (Table 1) (WHO 2005). Children diagnosed with tuberculosis and those whose caregivers did not give consent were not included in the study. We defined hypoxaemia according to the definition recommended by WHO (SpO2 < 90%).

Table 1. Definition of important clinical study parameters
Mental statusNormal/irritable/lethargic/convulsing/comatose
Some/severe dehydrationDefined by ‘Dhaka methods’ of assessment of dehydration, which is almost similar to WHO method and approved by WHO (Alam & Ashraf 2003)
Severe stunting<−3-z score of height for age of the median value of the National Centre for Health Statistics (NCHS)
Severe undernutrition<−3-z score of weight for age of the median value of the NCHS
Severe wasting<−3-z score of weight for height of median value of the NCHS
HypoxaemiaIf SPO2 without O2 is <90%
Lower chest wall indrawingInward movement of the bony structures of the lower chest wall with inspiration
Grunting respirationIt is an expiratory sound occurs because of partial closure of glottis as an effort to maintain intra-alveolar pressure for the prevention of alveolar collapse
SepsisPresence of inflammation [abnormal WBC count (>12 × 109/l or, <4 × 109/l or, band and neutrophil ratio ≥ 0.1)] plus presence or presumed presence of infection with thermo-instability [hypo (≤35.0° C) or hyperthermia (≥38.5° C)], tachycardia and/or the indications of altered organ function (altered mental status and bounding pulse) (Goldstein et al. 2005; Dellinger et al. 2008) in the absence of clinical dehydration or after correction of dehydration
Severe sepsisSepsis plus signs of poor peripheral perfusion (absent peripheral pulses or capillary refilling time ≥2 s or hypotension) (Goldstein et al. 2005; Dellinger et al. 2008)
Clinical pneumoniaDefined by WHO guideline of acute respiratory infection (WHO, 2005)
Radiological pneumoniaDefined as the presence of lobar or patchy consolidation

Patient management

Children admitted to the SCW receive treatment, which may include antibiotics, supportive care including intravenous fluids, oxygen, frequent monitoring and nutritional support (breast milk, micronutrients, zinc). Mechanical ventilation was not available at the time of this study. All children in the study were seen by the attending SCW physician, a history was taken and clinical examination performed. Arterial oxygen saturation (SpO2) was measured using a portable pulse oximeter (OxiMax N-600; Nellcor, Boulder, CO, USA) and a blood glucose tester (Glucocheck; STADA, Bad Vilbel, Germany).

Children with hypoxaemia received O2 supplementation through nasal prongs (2 l/min) or mask (5 l/min). Antibiotics were given to children who presented with pneumonia, sepsis, severe cholera, dysentery, severe malnutrition and other bacterial infections. Children received breast milk feeding when it was safe to do so, or oral rehydration solution by nasogastric tube or IV fluids, if they had severe dehydration and severe respiratory distress. Enteral correction of dehydration was carried out by oral rehydration salt solution, and intravenous correction was carried out by either cholera saline (contents: sodium: 133 mm, potassium: 13 mm, chloride: 99 mm, acetate: 48 mm) or 0.9% sodium chloride (contents: sodium: 154 mm, chloride: 154 mm). Management of pneumonia was according to the WHO algorithm (WHO 1999), and management of protein-energy malnutrition followed the hospital’s guidelines (Ahmed et al. 1999, WHO 1999).


Case report forms were developed, pre-tested and finalized for data acquisition. Characteristics analyzed were demographic (age and gender), nutritional status (severe wasting, severe stunting, severe undernutrition), symptoms on history (history of convulsion), signs on examination (lower chest wall indrawing, nasal flaring, head nodding, grunting respiration, cyanosis, dehydrating diarrhoea, abnormal mental status and inability to drink), clinical diagnosis (severe sepsis), laboratory investigations (hypoglycaemia, hyponatraemia, hypernatraemia, temperature and Vibrio cholarae) and outcome. The important parameters are shown in Table 1.


All data were entered into SPSS for Windows (version 15.0; SPSS Inc, Chicago) and Epi Info (version 6.0; USD, Stone Mountain, GA). Differences in proportions were compared by Fisher’s exact test. In normally distributed data, differences in means were compared by Student’s t-test, and in non-normally distributed data, differences in median were compared by Mann–Whitney test. A probability of <0.05 was considered statistically significant. Strength of association was determined by calculating odds ratio (OR) and 95% confidence intervals (CI). In the analysis of predictors of hypoxaemia, variables were initially analysed in a univariate model, then covariates were adjusted for using logistic regression to identify independent predictors. We also calculated the sensitivity, specificity, positive predictive value and negative predictive value of the independent predictors of hypoxaemia.


Among 258 children admitted, 119 (46%) had hypoxaemia (hypoxaemia group) and 138 children did not have hypoxaemia (comparison group). Among the hypoxaemic children, 62 (52%) were male and their median (IQR) age was 5 (2.5, 10.0) months; 198 (77%) had pneumonia according to the WHO definition, of whom 108 (55%) had hypoxaemia. Twenty-seven percent of patients with non-severe pneumonia were hypoxaemic compared with 56% among those with severe and 62% with very severe pneumonia. Of the 192, 176 (92%) children who had chest X-rays had radiologically proven pneumonia. Among these children, hypoxaemia occurred most frequently with lobar consolidation [23 children of 27 with lobar consolidation were hypoxaemic (85%)]. Hypoxaemic children more often had a fatal outcome (21%vs. 3% Odds ratio for death 8.98). The clinical predictors and conditions associated with hypoxaemia were lower chest wall indrawing, nasal flaring and severe sepsis (Tables 2 and 3). The predictive precision of various clinical signs for hypoxaemia is shown in Table 4. All other parameters (severe wasting, severe stunting, severe undernutrition, head nodding, grunting respiration, abnormal mental status, dehydration, cyanosis, unable to drink, hypoglycaemia, hyponatraemia, temperature) were not significantly associated with hypoxaemia (Table 2).

Table 2. Characteristics of the hypoxaemic children under five admitted in the special care ward of the Dhaka Hospital of ICDDR,B
VariablesHypoxaemia (119)
n (%)
Normal oxygen saturation (138)
n (%)
OR (95% CI)
  1. OR, Odds ratio; CI, Confidence interval; RBS, Random blood sugar; SD, Standard deviation; IQR, Interquartile range.

Male62 (52)84 (60)
Age (months) (median, IQR)5.0 (2.5, 10.0)8.0 (3.0, 21.0)
History of convulsion104 (88)99 (71)2.98 (1.39–6.48)0.003
Lower chest wall indrawing100 (84)57 (41)7.48 (3.97–14.22)<0.001
Nasal flaring37 (31)11 (8)5.21 (2.40–11.55)<0.001
Head nodding5 (4)3 (2)1.99 (0.40–10.75)0.480
Grunting respiration7 (6)3 (2)2.83 (0.64–14.18)0.190
Cyanosis11 (9)10 (7)1.31 (0.50–3.49)0.710
Dehydrating diarrhea (some/severe)53 (45)72 (52)0.75 (0.44–1.26)0.30
Abnormal mental status (irritable/lethargic/convulsing/comatose)96 (81)98 (71)1.78 (0.95–3.36)0.070
Unable to drink53 (45)45 (33)1.64 (0.96–2.82)0.070
Severe sepsis21 (18)8 (6)3.43 (1.37–8.84)0.005
Severe wasting30 (25)31 (23)1.17 (0.64–2.17)0.688
Severe stunting36 (30)38 (28)1.15 (0.65–2.05)0.706
Severe undernutrition51 (43)55 (40)1.15 (0.68–1.94)0.683
Hypoglycaemia (RBS < 3 mm)11 (9)16 (12)0.77 (0.32–1.85)0.67
Hyponatraemia (serum sodium ≤ 130 mm)19/101 (19)38/127 (30)0.54 (0.28–1.04)0.08
Hypernatraemia (serum sodium ≥ 150 mm)18/101 (18)10 (8)2.54 (1.04–6.26)0.038
Temperature (°C) (mean ± SD)37.6 ± 1.437.8 ± 1.20.227
Vibrio cholarae in stools6 (5)21 (15)0.30 (0.11–0.84)0.017
Outcome (Died)25 (21)4 (3)8.98 (2.84–31.54)<0.001
Table 3. Results of logistic regression to explore the predictors of hypoxaemia among the admitted children under five in special care ward of the Dhaka Hospital of ICDDR,B
VariableAdjusted OR95% CI P-value
  1. CI, confidence interval.

History of convulsion1.350.58–3.150.481
Lower chest wall indrawing6.913.66–13.08<0.001
Nasal flaring3.221.45–7.170.004
Severe sepsis4.481.62–12.420.004
Table 4. Sensitivity, specificity, positive and negative predictive value of predictors of Hypoxaemia
VariableHypoxaemic (119)
n (%)
Non-hypoxaemic (138)
n (%)
Sensitivity (95% CI)Specificity (95% CI)Positive predictive value (95% CI)Negative predictive value (95% CI)
  1. CI, confidence interval.

History of convulsion104 (88)99 (71)88 (80–93)28 (21–37)51 (44–58)72 (58–83)
Lower chest wall indrawing100 (84)57 (41)84 (78–90)59 (50–67)64 (56–71)81 (72–88)
Nasal flaring37 (31)11 (8)31 (23–40)92 (86–98)77 (62–87)61 (54–67)
Severe sepsis21 (18)8 (6)18 (11–26)94 (89–97)72 (53–87)57 (50–63)
In combination
 Lower chest wall indrawing or nasal flaring103 (87)61 (44)87 (92–99)56 (47–64)63 (55–70)83 (73–90)
 Lower chest wall indrawing or history of convulsion116 (97)107 (76)97 (92–99)22 (16–31)52 (45–59)91 (75–98)
 Lower chest wall indrawing or severe sepsis101 (85)59 (43)85 (77–91)57 (49–66)63 (55–71)81 (72–88)
 Nasal flaring or history of convulsion111 (93)100 (72)94 (87–97)28 (20–36)53 (46–59)83 (68–92)
 Nasal flaring or severe sepsis53 (45)18 (13)45 (36–54)87 (80–92)75 (63–84)65 (57–71)


This study of children with diarrhoea and co-morbidities found that lower chest wall indrawing, nasal flaring and the clinical syndrome of severe sepsis are independent indicators of hypoxaemia. While lower chest wall indrawing and nasal flaring have been shown before to be predictors of hypoxaemia (Weber et al. 1997; Usen & Webert 2001; Lodha et al. 2004; Basnet et al. 2006), there has been no previous identification of severe sepsis.

In this population, 11% of children fulfilled the criteria for the clinical syndrome of severe sepsis. ‘Sepsis’ is defined as the presence of inflammation plus presence or presumed presence of infection with hyperthermia or hypothermia (rectal temperature > 38.5° C or < 35.0° C) (Chisti et al. 2010) and tachycardia (Dellinger et al. 2008). ‘Severe sepsis’ is defined as the presence of poor peripheral perfusion (absent peripheral pulses or capillary refilling time ≥ 2 s) plus sepsis (Dellinger et al. 2008). Severe sepsis is often associated with vasodilatation and capillary leakage, as a result of amplified cytokines or other inflammatory stimuli (Ebrahim 2011; Annane et al. 2005). This leads to disordered microcirculation and increased lactate production as a by-product of anaerobic cellular respiration. Seventy-two percent of children with severe sepsis in our study had hypoxaemia, and 14 (48%) of the 29 children with severe sepsis died, making up 48% of all deaths in the study. Of those with severe sepsis who died, 13 (93%) had hypoxaemia. The clinical syndrome of severe sepsis may be under recognized in developing countries, but denotes a subgroup of children with other common clinical syndromes (pneumonia, diarrhoea, dysentery, malaria, severe malnutrition), who are at highest risk of death.

We also observed that hypoxaemic children more often presented with history of convulsion within 24 h prior to admission. However, the number of children with parental reports of convulsions was very high. Caregivers or parents might misinterpret rigor or shivering as a convulsion, leading to an overestimation of the frequency of this feature on history.

Hypoxaemia was significantly related to deaths, which has been reported in a number of earlier studies (Onyango et al. 1993; Smyth et al. 1998; Lozano 2001). A significantly higher proportion of children with hypoxaemia had severe sepsis and hypernatraemia in our study population, and these two conditions are also often associated with fatal outcome (Gustot 2011, Samadi et al. 1983; Adrogue & Madias 2000; Chisti et al. 2009).

The prevalence of hypoxaemia may vary from region to region (Smyth et al. 1998; Duke et al. 2002; Lodha et al. 2004; Subhi et al. 2009). The overall high prevalence of hypoxaemia among children in our study population might be because of the selected population in the SCW of the hospital. Many were seriously ill with lobar pneumonia, severe malnutrition and/or severe sepsis. Higher prevalence of hypoxaemia, in such conditions, has also been reported earlier (Bansal et al. 2006; Garcia-Vidal et al. 2008).

In conclusion, hypoxaemia is associated with a high case fatality, even in a tertiary referral hospital. Children under 5 years of age with diarrhoea who present with lower chest wall indrawing, nasal flaring and severe sepsis are more likely to have hypoxaemia than children without these clinical features. However, these clinical features cannot be solely relied upon and are often not identified by health workers, which underscore the importance of the use of pulse oximetry. Thus, in unavailability of pulse oximetry to initiate early oxygen supplementation in combination with other appropriate management to reduce morbidity and deaths, the identification of these simple clinical predictors of hypoxaemia is critical.


We gratefully acknowledge the donors for their support and commitment to ICDDR,B’s research efforts. We express our sincere thanks to all physicians, clinical fellows, nurses, members of feeding team and cleaners of SCW for their invaluable support and contribution during patient enrolment and data collection. ICDDR,B acknowledges the following donors who currently provide unrestricted support to the Centre`s research efforts: Government of the People’s Republic of Bangladesh; Canadian International Development Agency (CIDA), Embassy of the Kingdom of the Netherlands (EKN), Swedish International Development Cooperation Agency (SIDA) and the Department for International Development, UK (DFID). This study was supported as part of AusAID’s Knowledge Hubs for Women and Children’s Health, through the Centre for International Child Health at the University of Melbourne.