To evaluate the relative utility of clinical and laboratory parameters of dehydration in children for predicting the magnitude of percent less of body weight (PLBW), we studied 97 children who required intravenous fluids for acute dehydration. After a complete history and physical examination, the managing physician made a clinical estimation of dehydration for each child, based on a standard clinical scale. Serum electrolytes were obtained in all children prior to intravenous hydration therapy. PLBW was calculated after recovery from acute dehydration by comparing the weight on presentation to the emergency department with the weight measured at a follow-up visit when the child was judged well. Children were classified according to PLBW into three groups which reflect the categories in a standard clinical scale: mild = PLBW < or = 5 (n = 50), moderate = PLBW 6-10 (n = 30), and severe = PLBW > 10 (n = 17). The physician's clinical estimate of dehydration compared to PLBW had a sensitivity of 74% (95% confidence interval (CI): 60-85) for mild dehydration, 33% (95% CI: 17-53) for moderate dehydration, and 70% (95% CI: 44-89) for severe dehydration. There was a significant difference in the mean serum bicarbonate concentrations (HCO3) between the PLBW groups (P < 0.01). The sensitivity of the HCO3 < 17 mEq/L in predicting PLBW was 77% (95% CI: 58-90) for PLBW 6-10, and 94% (95% CI: 71-100) for PLBW > 10. The combination of the clinical scale and the serum bicarbonate identified all 17 children with PLBW > 10 and 90% (27 of 30) children with PLBW 6-10. Our data suggest that physicians should not rely solely on clinical assessment to rule out severe dehydration in children, and that obtaining a serum bicarbonate may improve the accuracy of predicting serious dehydration.