Figure 3
From: Does adding risk-trends to survival models improve in-hospital mortality predictions? A cohort study
Integrated discrimination improvement (IDI) with 95% confidence intervals (CI) comparing the existing model with and without the trend indicators. On each of the first 32 admission days (the 95th percentile of observed length of stay), we calculated the IDI with 95% CIs [12] using patients' estimated hazard of death from each model and vital status for that day. An IDI above 0 suggests that the addition of the trend indicators to the existing model improved overall model discrimination.