The clinical presentations of cholestasis in infancy caused by neonatal hepatitis and biliary atresia are very similar. Diagnosis may be difficult on many occasions, but the surgical treatment of biliary atresia should be performed as early as possible. We established a 3-day workup protocol for the differential diagnosis of biliary atresia and neonatal hepatitis and compared the diagnostic accuracy, sensitivity, specificity, and predictive values of various methods. One hundred and twenty-six infants, including 84 with neonatal hepatitis (age, 65.1 +/- 24.1 days) and 42 with biliary atresia (age, 60.3 +/- 31.1 days), were studied prospectively from July 1982 to December 1990. The diagnostic accuracy of various methods was as follows: liver histology, 96.8%; color of duodenal juice, 91.6%; peak radioisotope count in duodenal juice, 84.2%; ultrasonographic examination of the hepatobiliary system, 80.2%; and persistence of clay-colored stool, 80.2%. After stepwise logistic regression, the diagnostic methods of significance were liver biopsy, color of duodenal juice, abdominal ultrasonography, and stool color. However, stool color and the onset of jaundice could not differentiate severe neonatal hepatitis from biliary atresia. The diagnostic methods of significance then were liver biopsy and duodenal juice color. With this 3-day protocol, no biliary atresia was missed although four cases of neonatal hepatitis were misdiagnosed, resulting in unnecessary laparotomy; we found an overall diagnostic accuracy of 96.8%. We conclude that this 3-day diagnostic protocol is very helpful in the differential diagnosis of neonatal hepatitis and biliary atresia. Liver histologic examination is the most reliable single test for the differential diagnosis.