The C-statistic was "used to compare the performance of each coding algorithm in predicting in-hospital mortality. The C-statistic is a measure of a model's ability to discriminate those who die of those who do not die in hospital." p.1132

"The 95% confidence intervals that we derived by bootstrapping for each C-statistic estimate do not overlap, a finding that indicated that the C-statistic differences favoring the new models are obvious and beyond what one would expect to see due to chance alone." p.1139