Development and validation of a prediction model for lower extremity deep vein thrombosis risk in elderly patients with intracerebral hemorrhage.

Background: Geriatric patients are simultaneously exposed to a high risk of both bleeding and thromboembolism, making antithrombotic management for elderly intracerebral hemorrhage patients a dilemma. This study aimed to develop and validate a prediction model for lower extremity deep vein thrombosis risk in elderly patients with intracerebral hemorrhage. Methods: Demographics variables, clinical features and laboratory parameters were retrospectively collected from elderly intracerebral hemorrhage patients at our hospital from August 2023 to July 2024. Multiple imputation was used to handle missing values. Potential predictors of lower extremity deep vein thrombosis (demographic, clinical features and laboratory parameters) were assessed using least absolute shrinkage and selection operator (LASSO) regression for feature selection followed by multivariable logistic regression, a prediction model and a nomogram were constructed. Performance of the model was assessed through the receiver operating characteristic curve (ROC), calibration curve, decision curve analysis (DCA), and clinical impact curve (CIC). Results: Two hundred and forty-seven elderly intracerebral hemorrhage patients were included in the analysis. Five potential predictors were identified after LASSO regression: logDFR, D-dimer, eGFR, GCS and infection. These predictors were subsequently entered into a multivariable logistic regression model. Highly correlated predictors were further assessed to ensure model stability and interpretability. The final model included four independent predictors: logDFR (OR 1.84, 95% CI 1.28-2.63), eGFR (OR 1.01, 95% CI 1.00-1.02), GCS (OR 0.87, 95% CI 0.79-0.96) and infection (OR 2.20, 95% CI 1.05-4.63). The model had an area under the curve (AUC) of 0.819 (95%CI 0.808-0.830), suggesting good discrimination. Internal validation was performed using bootstrap resampling, the estimated optimism was 0.028, representing expected performance. Calibration assessment indicated only moderate overestimation of risk. Decision curve analysis (DCA) and clinical impact curve (CIC) consistently demonstrated the model's applicability across diverse clinical settings. Conclusion: This study developed and validated an effective prediction model, incorporating four clinically accessible parameters, for estimating the risk of lower extremity deep vein thrombosis in elderly patients with intracerebral hemorrhage. This model may facilitate the early identification of high-risk patients, thereby guiding prophylactic interventions.