Comparison of predictive performance of models as reported in the development studies.

Studies which reported the same outcome measure were considered for comparison. Since the performance is only comparable in similar populations, the study populations were compared on whether the population was VLGA/BW or ELGA/BW, from a developed or developing country, from the pre- or post- surfactant era, and the reported mortality rate (an absolute difference of ≤10% was considered comparable). Only three categories contained more than two studies. All performance measures for each prediction model are given in Table S3.*multiple models, using best performance of those reported.†performance assessed on development set, may be overestimated.