Methodologies for determining levels of U.S. physician requirement are as complex as they are controversial. One long-standing controversy surrounds the advantages of an epidemiologic need-based forecasting model over an economic demand-based model. This paper examines the need-based requirement approach as recently developed by the Graduate Medical Education National Advisory Committee (GMENAC). This approach is assessed for the pediatric specialty by replicating the original model using data derived from three large HMOs. These data were empirically obtained from the computerized visit records of more than 10,000 children at each of the three plans and normatively from Delphi panels consisting of pediatric practitioners at those same sites. Results indicate that if U.S. pediatrician requirement was estimated on the basis of HMO practice data, rather than GMENAC's national ideals, fewer physicians would be needed. The pediatric requirement based on local Delphi panel judgments was lower still, due in great part to the suggestion of increased delegation rates to nonphysician providers. Implications of this comparative analysis for the GMENAC need-based methodology and future physician requirement modeling efforts are discussed.