A nomogram for in-hospital mortality in the intensive care unit: development and validation with a type B aortic...

The early recognition of high-risk cases with type B aortic dissection (TBAD) in the intensive care unit (ICU) represents a critical component of effective clinical management. But a specific, practical tool for this purpose is lacking. Our research aimed to fill this need by constructing and validating a specific, validated instrument to predict in-hospital mortality among these individuals.

We obtained patient data consistent with the diagnosis of TBAD from Medical Information Mart for Intensive Care IV (MIMIC-IV) and eICU Collaborative Research Database (eICU-CRD). Missing data in the MIMIC-IV cohort were handled using multiple imputation. In the primary model-development analysis, the Synthetic Minority Over-Sampling Technique (SMOTE) was applied before splitting the data into training and internal validation sets. A nomogram was constructed using the least absolute shrinkage and selection operator (LASSO) and multivariable logistic regression. Model performance was assessed by the area under the receiver operating characteristic curve (AUC), calibration analyses, and decision curve analysis (DCA).

For descriptive cohort characteristics, the original non-oversampled cohorts included 341 patients from MIMIC and 76 patients from eICU. For the primary model-development analysis, after multiple imputation and SMOTE, the MIMIC cohort was divided into a training set (n=407) and an internal validation set (n=174), while the eICU cohort served as the external validation set (n=76). Seven predictor variables were determined and used to construct a nomogram. The nomogram demonstrated good discrimination, with AUCs of 0.789 [95% confidence interval (CI): 0.745-0.835] in the training cohort, 0.801 (95% CI: 0.733-0.868) in the internal validation cohort, and 0.792 (95% CI: 0.600-0.955) in the external cohort. In a non-oversampled sensitivity calibration analysis, calibration was good in the original training cohort, acceptable but attenuated in the original internal validation cohort, and less stable in the small complete-case external subset.

We constructed and validated a nomogram that can individually predict in-hospital mortality for ICU patients with TBAD. This practical tool may help clinicians conduct early prognostic assessments and develop treatment plans.