Development and external evaluation of a routine laboratory-based three-ratio score for in- hospital mortality in older critically ill patients with atherosclerotic cardiovascular disease: a MIMIC-IV and eICU study
Year: 2026
Type: preprint
Study design: Observational Study
Abstract: Background Older patients with atherosclerotic cardiovascular disease (ASCVD) often develop concurrent changes in renal handling, acid-base balance, and glucose metabolism after admission to the intensive care unit (ICU). A single laboratory measurement may not capture the combined extent of these early abnormalities. Methods We conducted a multidatabase retrospective cohort study using MIMIC-IV 3.1 and eICU 2.0. We included patients aged 65 years or older who had ASCVD, remained in the ICU for at least 24 h, and had the four laboratory measurements required to calculate the three-ratio score (TRS-3). TRS-3 was defined as the equally weighted mean of the blood urea nitrogen-to-creatinine, blood urea nitrogen-to-bicarbonate, and glucose-to-bicarbonate ratios after normalization to clinical reference values. The primary outcome was in-hospital all-cause mortality after the first 24 h of ICU admission. Associations were assessed using multivariable logistic regression, competing risk models, and restricted cubic splines. TabM models underwent patient-level cross-fitting in MIMIC-IV, after which the preprocessing parameters, model parameters, and classification threshold were applied unchanged to eICU. Results The MIMIC-IV cohort included 8037 patients, of whom 714 died in hospital. The eICU cohort included 9686 patients, of whom 1002 died in hospital. After full adjustment, the odds ratios (ORs) for in-hospital mortality per 1-standard-deviation increase in TRS-3 were 1.39 (95% CI: 1.28–1.51) in MIMIC-IV and 1.31 (95% CI: 1.22–1.41) in eICU. Compared with the lowest quartile, the highest quartile had ORs of 3.81 (95% CI: 2.80–5.20) and 2.70 (95% CI: 2.10–3.48), respectively. Both cohorts showed continuous nonlinear risk gradients. Adding TRS-3 to the core clinical TabM model increased the area under the receiver operating characteristic curve (AUROC) from 0.788 to 0.829 in MIMIC-IV and from 0.717 to 0.760 in eICU, with paired DeLong P < 0.001 in both cohorts. Net reclassification improvement and integrated discrimination improvement were positive in both cohorts. Conclusions A higher early TRS-3 was associated with a greater probability of in-hospital death among older critically ill patients with ASCVD and provided incremental discrimination beyond the core clinical TabM model. This fixed score can be calculated from 4 routine laboratory measurements, although its probability calibration and classification threshold require evaluation in prospective cohorts.
Source: Research Square
Language: English
Field: Medicine
Domain: Health Sciences
Open Access status: green
OpenAlex ID: W7223104968