Development and Validation of the Intensive Documentation Index for ICU Mortality Prediction

ABSTRACT Objective Nursing documentation patterns may reflect patient acuity and clinical deterioration, yet their prognostic value remains underexplored. We developed the Intensive Documentation Index (IDI), a novel framework quantifying temporal documentation rhythms, and evaluated its ability to enhance ICU in-hospital mortality prediction. Materials and Methods We analyzed 26,153 ICU admissions of heart failure (HF) patients from the MIMIC-IV database (2008-2019). Nine IDI features capturing documentation rhythm, volume, and surveillance gaps were extracted from electronic health record timestamps during the first 24 hours of ICU stay. Logistic regression models with and without IDI features were compared using temporal validation (training: 2008-2018; test: 2019). Results In-hospital mortality was 15.99% (n = 4,181/26,153). The baseline model (age, sex, ICU length of stay) achieved an AUROC of 0.658 (95% CI 0.609-0.710). Addition of nine IDI features significantly improved discrimination to AUROC 0.683 (95% CI 0.631-0.732), an absolute increase of 0.025 (p = 0.015, DeLong test). The coefficient of variation of inter-event intervals (idi_cv_interevent) was the strongest predictor (OR = 1.53 per SD; 95% CI = 1.35-1.74; p < 0.001). ICU length of stay was retained in the baseline model as a validated severity proxy; its potential as a post-baseline variable was examined in sensitivity analysis (see Sensitivity Analyses). Discussion IDI features, particularly documentation rhythm irregularity, improved mortality prediction overall (ΔAUC +0.025; p = 0.015) and in the White subgroup (ΔAUC +0.035; p = 0.014). However, IDI did not demonstrate statistically significant or clinically meaningful benefit in Black (ΔAUC +0.013; p = 0.694), Hispanic (ΔAUC −0.114; p = 0.466), or Asian (ΔAUC 0.000; p = 1.000) subgroups, with the latter two groups also exhibiting poor calibration. Small subgroup sample sizes (Hispanic n = 26; Asian n = 24) limit interpretation. Retrospective EHR timestamp latency may further attenuate real-time signal. Prospective validation with adequate minority representation is required. Conclusion Documentation rhythm patterns captured via the IDI modestly but reliably improve ICU mortality prediction beyond traditional clinical variables. IDI is a readily available prognostic signal derived from passively recorded timestamps and warrants prospective validation.