Development of Interactive Nomograms for Predicting Short-Term Survival in ICU Patients with Aplastic Anemia

TL;DR

The paper addresses the challenge of predicting short-term mortality for ICU patients with aplastic anemia by developing interpretable, logistic regression–based nomograms using a two-step feature selection that narrowed over 400 variables to seven clinically meaningful predictors. It employs a novel continuous-target SMOTE and evaluates multiple models, with logistic regression achieving strong internal AUROCs (0.823, 0.831, 0.830) and reasonable external validation (0.739–0.709), corroborated by SHAP and permutation analyses for interpretability. External validation on the eICU database demonstrates generalizability, and an interactive Dash-based nomogram (with a public GitHub link) translates the model into real-time bedside risk estimates. These tools enable targeted risk stratification and informed decision-making in critical care for a rare hematologic disorder, potentially guiding resource allocation and therapeutic planning at the point of care.

Abstract

Aplastic anemia is a rare, life-threatening hematologic disorder characterized by pancytopenia and bone marrow failure. ICU admission in these patients often signals critical complications or disease progression, making early risk assessment crucial for clinical decision-making and resource allocation. In this study, we used the MIMIC-IV database to identify ICU patients diagnosed with aplastic anemia and extracted clinical features from five domains: demographics, synthetic indicators, laboratory results, comorbidities, and medications. Over 400 variables were reduced to seven key predictors through machine learning-based feature selection. Logistic regression and Cox regression models were constructed to predict 7-, 14-, and 28-day mortality, and their performance was evaluated using AUROC. External validation was conducted using the eICU Collaborative Research Database to assess model generalizability. Among 1,662 included patients, the logistic regression model demonstrated superior performance, with AUROC values of 0.8227, 0.8311, and 0.8298 for 7-, 14-, and 28-day mortality, respectively, compared to the Cox model. External validation yielded AUROCs of 0.7391, 0.7119, and 0.7093. Interactive nomograms were developed based on the logistic regression model to visually estimate individual patient risk. In conclusion, we identified a concise set of seven predictors, led by APS III, to build validated and generalizable nomograms that accurately estimate short-term mortality in ICU patients with aplastic anemia. These tools may aid clinicians in personalized risk stratification and decision-making at the point of care.

Figures (9)

• Figure 1: Criterion of study population extraction
• Figure 2: Variance inflation factor of selected features.
• Figure 3: AUROC-curves for test set of LR Models
• Figure 4: (SHAP summary plot for the 7-day mortality prediction model.
• Figure 5: SHAP summary plot for the 14-day mortality prediction model.