Back

Multinational Validation of the Intensive Documentation Index for ICU Mortality Prediction: Temporal Resolution and ICU Mortality

Collier, A.; Shalhout, S. Z.

2026-03-23 health informatics
10.64898/2026.03.19.26348852 medRxiv
Show abstract

Clinical documentation timestamps generate a continuous, zero-burden behavioral signal in the electronic health record. We developed the Intensive Documentation Index (IDI) and validated it in two independent cohorts: MIMIC-IV (26,153 U.S. ICU heart failure patients, primary outcome in-hospital mortality) and HiRID (33,897 Swiss all-ICU patients, primary outcome ICU mortality). In MIMIC-IV, the IDI-enhanced logistic regression achieved an AUROC of 0.6491, compared with a baseline of 0.6242 (Brier score of 0.1299). In HiRID, where documentation latency is 1.2 minutes, compared with 15 hours in MIMIC-IV, AUROC was 0.9063, well above published APACHE IV and SAPS III benchmarks. The approximately 0.27 AUROC gap reflects the importance of temporal granularity in documentation-based risk stratification. IDI requires no physiologic measurements, making it complementary to established severity scores. Prospective validation in real-time EHR systems is required before clinical deployment.

Matching journals

The top 3 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.