Survival prediction modelling in extreme prematurity: are days important?

To demonstrate that days are important in extreme prematurity when creating survival prediction models based on gestation.

Prospectively collected data were analysed for all admitted infants born 23 + 0 to 27 + 6 weeks gestation in the Australian and New Zealand Neonatal Network from 2009 to 2016. The effect of days on observed survival rates was assessed using a non-parametric test for trend. Prediction models created based on gestational age in completed weeks only or weeks plus days were compared.

Seven thousand eight hundred and thirty-six extreme preterm infants were studied. Observed survival increased with days for 23, 24, 25, and 27 weeks gestational age (P = 0.01; P < 0.001; P = 0.003; P = 0.003) but not for 26 weeks (P = 0.19). A survival prediction model based on weeks and days performed better than completed weeks only (AUC 0.722 vs 0.712; P < 0.001).

In extreme prematurity, survival estimate accuracy improves when survival prediction models include days in addition to weeks.