Predicting length of hospitalization of sick neonates from their initial status

It has been suggested that the estimated date of confinement (EDC) may be used to predict the length of hospital stay for sick newborns. We have found this method unreliable and designed the following study to develop a mathematical model to predict length of stay (LOS). We reviewed the records of 393 neonates. Statistical analysis was performed using multiple linear regression. The factors that reached statistical significance were birth weight (BW), gestational age (GA), and a combined respiratory score (RES). The RES was developed to quantify the degree of initial respiratory illness. Through this model we developed the following formula: log(e) (LOS) = 4.395-0.023 (GA) -0.00054(BW) + 0.0274 (RES). The R value is 0.78. The model predicts an LOS +/- 10 days in 73% of cases. Overall, this model yields a 29% improvement in predictability of LOS compared with a model which used EDC only. This formula may provide useful information for parents and caregivers of hospitalized neonates.