Dehydration is a leading cause of morbidity and mortality in children worldwide. Each year, up to 300 children in the United States die from gastroenteritis-related dehydration. Most guidelines recommend determining whether the child is mildly (3 to 5 percent), moderately (6 to 9 percent), or severely dehydrated (10 percent or more). Proper diagnosis of dehydration is important because underestimation and overestimation of the degree of dehydration can cause unnecessary morbidity and increase health care costs. Steiner and colleagues reviewed the literature to determine the accuracy and precision of history, physical examination, and laboratory tests in identifying dehydration in children one month to five years of age.

The authors reviewed articles focusing on the diagnosis of dehydration, with none of the studies meeting criteria for high quality of evidence. Low urine output did not increase the likelihood of 5 percent dehydration, but normal urine output, as reported by the parents, did lower the likelihood. The best measures for detecting 5 percent dehydration were prolonged capillary refill, abnormal skin turgor, and abnormal respiratory pattern. Less helpful, either because of a low likelihood ratio (LR) or a wide confidence interval, were cool extremities, absence of tears, weak pulse, sunken eyes, and dry mucous membranes. Similarly, increased heart rate, sunken fontanelle, and overall poor appearance in infants contributed little to an accurate diagnosis of dehydration. Combinations of signs, using various prediction models, performed better.

A nine-point physical examination scale to classify children as mildly, moderately, or severely dehydrated had an LR of 3.4 for the presence of at least 5 percent dehydration and an LR of 4.3 for at least 10 percent dehydration. A lower and, therefore, less helpful LR was seen for moderate dehydration.

Laboratory tests included blood urea nitrogen (BUN) and BUN/serum creatinine ratio, acidosis, serum bicarbonate concentration, and urine-specific gravity. Of these, only the serum bicarbonate test was helpful; a concentration between 15 and 17 mEq per L reduced the likelihood that a child had greater than 5 percent dehydration in the presence of gastroenteritis.

The authors conclude that prolonged capillary refill time, abnormal skin turgor, and abnormal respiratory pattern are the most useful signs in diagnosing dehydration in children, with groups of signs improving the accuracy of categorization.