Inter-observer variation in the assessment of clinical signs in sick Tanzanian children.

We assessed the inter-observer agreement in identification of a range of 24 clinical signs associated with disease presentation in 327 children aged < 5 years admitted to hospital in January-June 1999 in Ifakara, southern Tanzania. Children with diagnoses of malaria, pneumonia, diarrhoea, anaemia and malnutrition were examined independently by 2 clinical officers. Findings were recorded on a standard proforma. The Kappa-statistic was used to assess inter-observer agreement for each sign. Physical signs were more likely to be agreed upon by clinicians if they involved inspection than if they involved auscultation. The signs included in the Integrated Management of Childhood Illness (IMCI) algorithm were found to be largely appropriate (Kappa-scores > 0.41) although there was only fair agreement (Kappa-score 0.21-0.40) in the detection of neck stiffness and chest indrawing and slight agreement in the detection of dehydration (Kappa-score 0.199). All objective neurological signs were less reliably assessed in infants than in older children. The difficulties surrounding the diagnosis of impaired consciousness in young children should increase vigilance in the diagnosis and management of neurological complications of illnesses in infancy.