Background: Transfusions to neonates convey both benefits and risks, and evidence is needed to guide wise use. Such evidence is accumulating, but more information is needed to generate sound evidence-based practices. Objective: We sought to analyze published information on nine aspects of transfusion practice in neonatal intensive care units. Methods: We assigned ‘categories of evidence' and ‘recommendations' using the format of the United States Preventive Services Task Force of the Agency for Healthcare Research and Quality. Results: The nine practices studied were: (1) delayed clamping or milking of the umbilical cord at preterm delivery - recommended, high/substantial A; (2) drawing the initial blood tests from cord/placental blood from very low birth weight (VLBW, <1,500 g) infants at delivery - recommended, moderate/moderate B; (3) limiting phlebotomy losses of VLBW infants - recommended, moderate/substantial B; (4) selected use of erythropoiesis-stimulating agents to prevent transfusions - recommended, moderate/moderate-moderate/small B, C; (5) using platelet mass, rather than platelet count, in platelet transfusion decisions - recommended, moderate/small C; (6) permitting the platelet count to fall to <20,000/µl in ‘stable' neonates before transfusing platelets - recommended, low/small I; (8) permitting the platelet count to fall to <50,000/µl in ‘unstable' neonates before transfusing platelets - recommended, moderate/small C, and (9) not performing routine coagulation test screening on every VLBW infant - recommended, moderate/small C. Conclusions: We view these recommendations as dynamic, to be revised as additional evidence becomes available. We predict this list will expand as new studies provide more information to guide best transfusion practices.