Development and validation of a comprehensive prognostic model for 28-day ICU mortality in non-traumatic subarachnoid hemorrhage: an analysis based on the MIMIC-IV database.

Admission Laboratory-Based Prediction of Six-Month Mortality After COVID-19 Hospitalization: Model Development, Bootstrap Internal Validation, and Single-Center Temporal Validation

Background and Objectives: Early identification of patients at high risk of death after COVID-19 hospitalization may support monitoring, follow-up planning and resource allocation. We aimed to develop and internally validate a parsimonious admission laboratory-based model for six-month all-cause mortality and derive a simplified risk score. Materials and Methods: This retrospective single-center cohort included 1827 consecutive adults hospitalized with COVID-19 between 3 March and 6 November 2020. The primary model was developed in 1559 patients with complete predictor data after data-quality review. Prespecified predictors were age, albumin, neutrophil-to-lymphocyte ratio, blood urea nitrogen, lactate dehydrogenase, troponin I and sodium. Model performance was assessed using discrimination and calibration measures, 500-sample bootstrap internal validation and a chronological split-sample assessment in later admissions from the same center; this was internal–temporal rather than external validation. The simplified score was derived exclusively in the chronological development cohort and applied unchanged to the later cohort. Results: Among 1827 eligible patients, 437 died within six months (23.9%). The primary complete-case modeling cohort included 1559 patients, of whom 386 died. Older age, lower albumin and higher neutrophil-to-lymphocyte ratio, blood urea nitrogen, lactate dehydrogenase, troponin I and sodium were independently associated with mortality. The model achieved an apparent AUC of 0.936 (95% CI 0.919–0.954) and an optimism-corrected AUC of 0.933. The later same-center cohort yielded an AUC of 0.908 (95% CI 0.871–0.939). The development-derived simplified score had an AUC of 0.927 (95% CI 0.909–0.943) in the development cohort and 0.916 (95% CI 0.885–0.944) when applied unchanged to the later cohort. Conclusions: A model combining age with six routinely available admission laboratory variables showed strong discrimination for six-month all-cause mortality after COVID-19 hospitalization and retained good discrimination in single-center temporal assessment. However, calibration drift was observed, and the model and simplified score should currently be considered supportive risk-stratification frameworks rather than deployable stand-alone clinical calculators. External validation, contemporary recalibration and prospective assessment of clinical utility are required before implementation.

Prognostic Score Model for 30-Day Mortality in Patients with Acute Pulmonary Embolism Presenting to the Emergency Department

Background/Objectives: Early risk stratification is essential in pulmonary embolism (PE), but a simple tool integrating routinely available clinical and laboratory variables is lacking. We aimed to develop a simple score for predicting 30-day mortality in patients presenting to the emergency department (ED) with PE. Methods: We conducted a multicenter study in three hospitals in Korea. The score was derived at the largest hospital using least absolute shrinkage and selection operator logistic regression with bootstrap stability selection, and validated in the pooled cohort from the remaining two hospitals. Discrimination was assessed using the area under the receiver operating characteristic curve (AUROC) and compared with PESI and sPESI. Calibration was assessed using the Brier score and calibration parameters. Results: Among 2446 patients, 1753 were included in the derivation cohort and 693 in the validation cohort. The final score (3C score) assigned one point each for history of cancer, international normalized ratio ≥1.15, and C-reactive protein ≥50 mg/L. In the validation cohort, the AUROC was 0.767 (95% CI, 0.703–0.822) for 30-day mortality, compared with 0.748 for PESI and 0.725 for sPESI. 30-day mortality increased from 2.4% (score 0) to 34.4% (score 3). A score of 0 identified 42.7% of patients as low risk, with a negative predictive value of 97.6%. Calibration was acceptable (Brier score, 0.077; calibration slope, 1.098). Conclusions: The 3C score showed discrimination comparable to PESI and sPESI and identified a substantial subgroup with low risk. Its simplicity may facilitate ED risk assessment, although further validation is required before clinical implementation.

Prognostic utility of the inflammatory burden index for early mortality prediction in heart failure: a retrospective cohort study using the MIMIC-IV database

Background: Systemic inflammation a key factor in the progression of heart failure (HF). The inflammatory burden index (IBI) has prognostic value in different conditions; however, its impact on short-term mortality in patients with HF remains uncertain. This study aimed to assess the association between IBI and mortality risk in patients with HF. Methods: In this retrospective study, 600 patients with HF from the Medical Information Mart for Intensive Care IV database (2008–2022) were examined and divided into three groups according to the log-transformed IBI (LnIBI). The main outcome measured was 28-day mortality in the intensive care unit (ICU). We used multivariable Cox and logistic regression to assess the independent effect of LnIBI on mortality, after adjustment for confounders. The dose-response relationship was modeled using restricted cubic splines. Predictive performance was compared using receiver operating characteristic curves, and Kaplan-Meier analysis was used to assess survival differences;subgroup analyses were conducted. Results: The cohort included 342 males (57.0%), with a median age of 71.0 years. The 28-day ICU mortality rates were 17.2% (103/600). In adjusted models, higher LnIBI independently predicted an elevated risk (for example, 28-day mortality after ICU admission: Hazard ratio (HR): 1.24, 95% confidence interval (CI): 1.09–1.41, p = 0.001; highest versus lowest tertile: HR: 1.99, 95% CI: 1.20–3.29, p = 0.008). Restricted cubic splines confirmed the linear dose-response relationship. LnIBI showed a superior area under the curve compared with C-reactive protein (CRP) for most endpoints (for example, 0.643 versus 0.595 for 28-day mortality after ICU admission, p = 0.032). Kaplan-Meier curves indicated poorer survival in the higher LnIBI tertiles (p < 0.05). Conclusions: Elevated IBI is independently associated with increased short-term mortality risk in critically ill patients with HF, outperforming CRP in predictive accuracy.

Construction and validation of a sepsis prediction model using inflammatory and hemodynamic markers in the emergency department

Objective To develop an early diagnosis prediction model for sepsis in the emergency department by integrating inflammatory, hemodynamic, and nursing assessment indicators. Methods A retrospective cohort of 320 patients with suspected infection admitted to our hospital was enrolled. Participants were randomly allocated into a training set and a validation set at a 7:3 ratio. Key variables were selected using least absolute shrinkage and selection operator (LASSO) regression. Subsequently, logistic regression, random forest, and support vector machine models were constructed. Model performance and interpretability were evaluated using the receiver operating characteristic curve, calibration curve, decision curve analysis, and SHapley Additive exPlanations (SHAP) values. Results LASSO regression identified three core variables: procalcitonin (PCT), neutrophil-to-lymphocyte ratio (NLR), and Modified Early Warning Score (MEWS). Multivariate analysis revealed that all three were independent risk factors (p < 0.05). The random forest model demonstrated an area under the curve (AUC) of 0.737 in the training set and 0.714 in the internal random-split validation set. At the optimal cutoff, the validation performance yielded a sensitivity of 72.7%, specificity of 76.2%, PPV of 61.5%, and NPV of 84.2%. It showed good calibration and provided clinical net benefit. SHAP analysis indicated that MEWS contributed the most to the model’s predictions, followed by NLR and PCT. Conclusion The integrated model developed in this study exhibits moderate to satisfactory predictive performance and clinical utility. It may act as a candidate bedside risk-stratification tool pending further external multicenter validation for emergency nursing sepsis screening.

Frailty combined with nutritional risk for predicting stroke-associated pneumonia: a cohort study based on a nomogram model

Introduction Stroke-associated pneumonia (SAP) is a common and serious complication in patients with acute severe stroke, and existing risk assessment tools have limited predictive accuracy in critically ill populations. This study innovatively incorporated frailty and nutritional risk, which reflect stress tolerance and overall physiological reserve, into an early SAP prediction model. Methods A retrospective cohort study was conducted on 293 critically ill stroke patients admitted to the Neurocritical Care Unit of the First Affiliated Hospital of Chongqing Medical University between 2013 and 2024. Collect clinical characteristics and laboratory indicators of patients, assess their frailty status and nutritional risk, and analyze the additive interaction effect between the two on the occurrence of SAP. Independent predictors were identified through multivariate logistic regression and incorporated into a visual nomogram. Model performance was evaluated using the area under the receiver operating characteristic curve (AUC), calibration plots, decision curve analysis, and 10-fold cross-validation. Results A total of 293 patients with severe stroke were included in this study, among whom 126 (43%) developed SAP. The results of the additive interaction analysis showed a positive additive interaction between frailty and nutritional risk in the development of SAP, with an attributable proportion (AP) of 0.711 (95% CI = 0.358 ~ 1.065) and a synergy index (SI) of 3.694 (95% CI = 1.200 ~ 11.364). A SAP risk prediction model incorporating age, nasogastric tube use, neutrophil-to-lymphocyte ratio (NLR), frailty status, and nutritional risk demonstrated good discriminative performance, with an area under the curve (AUC) of 0.848, which was significantly higher than that of the conventional SAP prediction score (ISAN score: AUC = 0.589). Internal validation showed that the model achieved an accuracy of 73.93%, sensitivity of 77.30%, and specificity of 71.95%, indicating good stability. The calibration curve demonstrated good agreement between predicted and observed outcomes. Decision curve analysis (DCA) indicated that the model provided substantially greater clinical net benefit than the ISAN score. Discussion This study is the first to integrate frailty and nutritional risk into an SAP prediction model, significantly improving early risk identification and providing an innovative, practical tool for precision prevention and targeted intervention in critically ill stroke patients.

[Development and validation of an explainable machine learning-based model for predicting mortality risk in polytrauma patients].