Presented at the Neonatal Society 2004 Autumn Meeting.
Dorling J1, Field D1, Draper E2, Manktelow B2
1 Neonatal Unit, Leicester Royal Infirmary, Leicester LE1 5WW, UK
2 Department of Health Sciences, University of Leicester, 22–28 Princess Road West, Leicester LE1 6TP, UK
Introduction: The development of a model that predicts neurological outcome following premature birth would enable a number of epidemiological approaches to be pursued. These include identifying important organisational factors that cannot be tested by controlled trials and risk corrected comparisons of unit performance. We hypothesised that using data available at birth, a scoring system can be developed to accurately predict outcome in infants born before 33 completed weeks of gestation. Secondly we wished to determine whether changing the gestational age cut off altered the predictive ability of the score.
Methods: Data were obtained from the Trent Neonatal Survey for 311 infants born to mothers resident in Nottingham or Leicestershire in 1997. Neurodevelopmental outcomes were determined from parental questionnaire or extracted from routinely available community data (1). Statistical analysis was performed using SPSS v11. Factors associated with outcome at 2 years of age were identified by univariate analysis using the Chi-squared test, Fishers exact test or logistic regression analysis as appropriate. Only variables collectable at birth were analysed to minimise the potential for the quality of treatment received to influence the score. Predictive variables were combined into models using multivariable logistic regression analysis. Analyses were then repeated, for differing gestational age ranges. Receiver operator characteristic curve analysis (ROC) was used to assess the ability of each model to discriminate between outcomes.
Results: The models were derived from 6 variables: gestational age, birthweight, sex, Apgar score at 1 minute, and presence of a congenital malformation. The ROC curve values for the differing gestational age groups are listed in the table.
Conclusions: For a cohort of the most premature infants, it appears possible to reliably predict outcome at 2 years of age from data obtained at birth. For infants above 29 weeks gestational age discrimination drops is poor and risk correction for neurological outcome may not be possible in these in infants. Additional study of these models in larger cohorts is needed.
References
1. Field D et al, BMJ 2001;323:1276-81.