A new method to analyse the pace of child development: Cox regression validated by a bootstrap resampling procedure

BACKGROUND Various perinatal factors influencing neuromotor development are known from cross sectional studies. Factors influencing the age at which distinct abilities are acquired are uncertain. We hypothesized that the Cox regression model might identify these factors. METHODS Neonates treated at Aachen University Hospital in 2000/2001 were identified retrospectively (n = 796). Outcome data, based on a structured interview, were available from 466 children, as were perinatal data. Factors possibly related to outcome were identified by bootstrap selection and then included into a multivariate Cox regression model. To evaluate if the parental assessment might change with the time elapsed since birth we studied five age cohorts of 163 normally developed children. RESULTS Birth weight, gestational age, congenital cardiac disease and periventricular leukomalacia were related to outcome in the multivariate analysis (p < 0.05). Analysis of the control cohorts revealed that the parents' assessment of the ability of bladder control is modified by the time elapsed since birth. CONCLUSIONS Combined application of the bootstrap resampling procedure and multivariate Cox regression analysis effectively identifies perinatal factors influencing the age at which distinct abilities are acquired. These were similar as known from previous cross sectional studies. Retrospective data acquisition may lead to a bias because the parental memories change with time. This recommends applying this statistical approach in larger prospective trials.