Background: Pneumonia is the leading cause of childhood morbidity and mortality' in
developing countries with hvpoxaemia as the most common and fatal complication.
Oxygen therapy is an important intervention for children with hypoxaemia. In many
settings in Kenya, clinical signs are used to identify children who require oxygen. The
Government of Kenya (GoK) has provided criteria for oxygen therapy. It states that
oxygen should be administered to a child with any of these signs: cyanosis, inability to
drink/breastfeed, impaired consciousness, grunting or head nodding. While there is data
exploring the utility of clinical signs to identify hypoxaemic children, this GoK ‘decision
rule' has never been evaluated. There is paucity of information on some of the signs
included in the GoK criteria and little local information on the prevalence of hypoxaemia
among children with severe forms of pneumonia.
Objectives: To determine the prevalence of hypoxaemia and evaluate the sensitivity and
specificity of the GoK criteria for oxygen therapy for children with severe or very severe
pneumonia admitted at Kenyatta National Hospital, Nairobi, Kenya, to determine
whether Human Immunodeficiency Virus (HIV) infection was a risk factor for
hypoxaemia and to evaluate the association between hypoxaemia and short term inpatient
mortality.
Methodology: This was a hospital based short longitudinal survey. We enrolled 343
children aged two to 59 months, assessed them for presence of clinical signs associated
with hypoxaemia, measured their arterial oxygen saturation using a portable hand held
10
pulse oximeter and had them tested for HIV infection. We followed up the children for
five days to determine mortality outcome.
Results: Prevalence of hypoxaemia was 50.7% in the study population. Stratified by severity,
39.7% and 59.4% of children with severe and very severe pneumonia respectively were
hypoxaemic. Cyanosis and grunting were found to be independent predictors of hypoxaemia. The
GoK criteria had a sensitivity of 65.5% and a specificity of 53.8% for detecting children who
required oxygen therapy. Thirty one children (9.0%) were HIV infected. Oxygen saturation of
<85% was associated with increased mortality (OR 3.3. 95% CI= 1.5 to 7.1, P=0.005).
Conclusions: Hypoxaemia is frequent, occurring in 50.7% of children hospitalized with
severe or very severe pneumonia at Kenyatta National Hospital. The GoK criteria for
oxygen therapy have a low sensitivity (65.5%) and specificity (53.8%) for predicting
hypoxaemia. Severe hypoxaemia (Sp02 <85%) is associated with a 3.3 fold increased
mortality.
Recommendations: The Government of Kenya should consider promoting the use of
pulse oximetry in all public hospitals to detect hypoxaemia. A cost-benefit study on the
use of pulse oximeters vis-a-vis continued use of clinical signs to determine which
children require oxygen therapy should be carried out.