Validity and reliability of clinical signs in the diagnosis of dehydration in children.

To determine the validity and reliability of various clinical findings in the diagnosis of dehydration in children. Prospective cohort study. An urban pediatric hospital emergency department. One hundred eighty-six children ranging in age from 1 month to 5 years old with diarrhea, vomiting, or poor oral fluid intake, either admitted or followed as outpatients. Exclusion criteria included malnutrition, recent prior therapy at another facility, symptoms for longer than 5 days' duration, and hyponatremia or hypernatremia. All children were evaluated for 10 clinical signs before treatment. The diagnostic standard for dehydration was fluid deficit as determined from serial weight gain after treatment. Sixty-three children (34%) had dehydration, defined as a deficit of 5% or more of body weight. At this deficit, clinical signs were already apparent (median = 5). Individual findings had generally low sensitivity and high specificity, although parent report of decreased urine output was sensitive but not specific. The presence of any three or more signs had a sensitivity of 87% and specificity of 82% for detecting a deficit of 5% or more. A subset of four factors-capillary refill >2 seconds, absent tears, dry mucous membranes, and ill general appearance-predicted dehydration as well as the entire set, with the presence of any two or more of these signs indicating a deficit of at least 5%. Interobserver reliability was good to excellent for all but one of the findings studied (quality of respirations). Conventionally used clinical signs of dehydration are valid and reliable; however, individual findings lack sensitivity. Diagnosis of clinically important dehydration should be based on the presence of at least three clinical findings.