Models to predict total hospital length of stay (LOS) among babies admitted to the neonatal intensive care unit (NICU) can be employed to facilitate discharge and follow-up planning (e.g., specialty referrals). It is particularly desirable to define robust statistical models that employ objective predictors (i.e., not based on diagnosis). We have developed a predictive model suitable for use by community NICU's. We employed a dataset based on 1238 surviving preterm/low birth weight babies (<2500 grams and/or<37 weeks gestation) admitted to 4 Kaiser Permanente level III NICU's in 1993-1994. We tested objective predictor variables (e.g., birth weight) singly and in combination. Used alone, birth weight predicted 46% of total LOS, while gestational age predicted 59% of total LOS. The best model employed time required to reach 38 completed weeks; whether a baby was in the lowest birth weight quartile for its gestational age; and whether a baby was in the highest Score for Neonatal Acute Physiology (SNAP) quartile for its gestational age. This model predicted 71% of LOS. The model predicted 67% of LOS when applied to a seperate (validation) dataset based on 986 surviving preterm/low birth weight babies admitted to 6 NICU's in 1995. Among babies of 24-32 weeks gestation in the validation dataset, 39.7% of the babies' actual LOS were within 7 days of that predicted by the model, and 50.9% were within 10 days of that predicted by the model. The model performed better among babies of 33-36 weeks gestation: 84.2% of the babies'actual LOS were within 7 days and 90.5% were within 10 days of that predicted by the model. We also developed a model for babies whose LOS was ≥ 4 days. This model included time required to reach 38 completed weeks, whether a baby was in the lowest birth weight quartile, and whether or not assisted ventilation was in progress at the begginning of day 4. This second model predicted 75% of LOS among babies who stayed for ≥ 4 days. Using simple models early in the nursery course, one can provide parents and caregivers with accurate estimates of how long a baby is likely to stay in the NICU.