OBJECTIVE: To test the hypothesis that utilization of a previously described measure of acuity (ie, the score for neonatal acute physiology [SNAP]) during the first 7 postnatal days predicts which infants with a birth weight of 1500 g or less received erythrocyte transfusion during the initial hospitalization. DESIGN: Retrospective chart review. SETTING: A regional tertiary care newborn intensive care unit at the Arizona Health Sciences Center, University Medical Center, Tucson. MATERIALS: Medical records of premature infants (birth weight, < or = 1500 g) who were admitted from October 1993 to January 1995. MAIN OUTCOME MEASURES: Occurrence or nonoccurrence of erythrocyte transfusion was determined in 47 infants who were compared for demographic information, phlebotomy blood loss, diagnoses, medications, and the SNAP at 0, 1, 2, and 7 days of life. RESULTS: Infants with a birth weight of 1500 g or less received a mean +/- SD of 1.9 +/- 2.9 transfusions with 22 (47%) of the infants given transfusions Infants who were given transfusions vs those who were not given transfusions were of a lower mean +/- SD birth weight (971 +/- 238 g vs 1272 +/- 144 g; P < .001) and a lower gestational age (27.7 +/- 1.6 weeks vs 30.7 +/- 2.8 weeks; P < .001), and they had a greater mean phlebotomy blood loss (3.3 +/- 1.6 mL/kg per day vs 1.4 +/- 0.5 mL/kg per day; P < .001) during the first postnatal week. The SNAP indexes in those who received transfusions were higher at 1, 2, and 7 days of life (P = .03, P = .001, and P < .001, respectively). Using stepwise logistic regression, phlebotomy blood loss and the SNAP at 7 days of life were significant predictors of the number of transfusions. The logistic model predicted which infants had been administered transfusions with 86% sensitivity and 88% specificity. CONCLUSIONS: The efficacy and cost-effectiveness of recombinant human erythropoietin therapy in premature infants remain under study. As earlier treatment with recombinant human erythropoietin may be more efficacious, early identification of which infants currently undergo transfusion may identify those who will receive the greatest benefit from recombinant human erythropoietin therapy. The SNAP distinguished those infants who were given transfusions from those who did not receive transfusions, even after adjusting for phlebotomy blood loss.