Prediction of Death for Extremely Premature Infants in a Population-Based Cohort

Abstract

OBJECTIVE: Although gestational age (GA) is often used as the primary basis for counseling and decision-making for extremely premature infants, a study of tertiary care centers showed that additional factors could improve prediction of outcomes. Our objective was to determine how such a model could improve predictions for a population-based cohort. METHODS: From 2005 to 2008, data were collected prospectively for the California Perinatal Quality Care Collaborative, which encompasses 90% of NICUs in California. For infants born at GAs of 22 to 25 weeks, we assessed the ability of the Eunice Kennedy Shriver National Institute of Child Health and Human Development 5-factor model to predict survival rates, compared with a model using GA alone. RESULTS: In the study cohort of 4527 infants, 3647 received intensive care. Survival rates were 53% for the whole cohort and 66% for infants who received intensive care. In multivariate analyses of data for infants who received intensive care, prenatal steroid exposure, female sex, singleton birth, and higher birth weight (per 100-g increment) were each associated with a reduction in the risk of death before discharge similar to that for a 1-week increase in GA. The multivariate model increased the ability to group infants in the highest and lowest risk categories (mortality rates of >80% and <20%, respectively). CONCLUSIONS: In a population-based cohort, the addition of prenatal steroid exposure, sex, singleton or multiple birth, and birth weight to GA allowed for improved prediction of rates of survival to discharge for extremely premature infants.