Predictors of neonatal sepsis in developing countries.
Schlüsselwörter
Abstrakt
BACKGROUND
Neonatal infections are a major cause of death worldwide. Simple procedures for identifying infants with infection that need referral for treatment are therefore of major public health importance.
METHODS
We investigated 3303 infants <2 months of age presenting with illness to health facilities in Ethiopia, The Gambia, Papua New Guinea and The Philippines, using a standardized approach. Historical factors and clinical signs predicting sepsis, meningitis, hypoxemia, deaths and an ordinal scale indicating severe disease were investigated by logistic regression, and the performance of simple combination rules was explored.
RESULTS
In multivariable analysis, reduced feeding ability, no spontaneous movement, temperature >38 degrees C, being drowsy/unconscious, a history of a feeding problem, history of change in activity, being agitated, the presence of lower chest wall indrawing, respiratory rate >60 breaths/min, grunting, cyanosis, a history of convulsions, a bulging fontanel and slow digital capillary refill were independent predictors of severe disease. The presence of any 1 of these 14 signs had a sensitivity for severe disease (defined as sepsis, meningitis, hypoxemia, or radiologically proven pneumonia) of 87% and a specificity of 54%. More stringent combinations, such as demanding 2 signs from the list, resulted in a considerable loss of sensitivity. By contrast only slight loss of sensitivity and considerable gain of specificity resulted from reducing the list to 9 signs. Requiring the presence of fever and any other sign produced a diagnostic rule with extremely low sensitivity (25%).
CONCLUSIONS
Physical signs can be used to identify young infants at risk of severe disease, however with limited specificity, resulting in large numbers of unnecessary referrals. Further studies are required to validate and refine the prediction of severe disease, especially in the first week of life, but there appear to be limits on the accuracy of prediction that is achievable.
