Are bedside features of shock reproducible between different observers?

Shock is often under-reported in children attending hospitals in developing countries. Readily obtainable features of shock (capillary refill time, temperature gradient, pulse volume, and signs of dehydration) are widely used to help prioritise management in the emergency assessment of critically ill or injured children. However, data are lacking on their validity, including, importantly, reproducibility between observers. Agreement of these signs was examined in 100 consecutive children admitted to a paediatric ward on the coast of Kenya. After an initial training of clinical sign recognition, there was moderate agreement for most features of cardiovascular compromise (delayed capillary refill > or =4 s, kappa = 0.49; and weak pulse volume, kappa = 0.4) and only substantial agreement for temperature gradient (kappa = 0.62). For hydration status, only in the assessment of skin turgor was there a moderate level of agreement (kappa = 0.55). Capillary refill times and assessment of pulse volume recommended by the recent American consensus guidelines achieved only a "low" moderate to poor interrater agreement, questioning the reliability of such parameters.