What question did this study set out to answer?

The aim is to develop a validated model for predicting venous thromboembolism in patients with renal insufficiency.

April 15, 2026Open Access

Development and internal validation of a clinical prediction model for venous thromboembolism in patients with renal insufficiency

Puntos clave

The aim is to develop a validated model for predicting venous thromboembolism in patients with renal insufficiency.
Included 698 hospitalized patients with renal insufficiency, assessing VTE risk.
Used multiple imputation and Synthetic Minority Oversampling Technique for data handling.
Employed LASSO regression, random forest, and logistic regression for model development.
Identified independent predictors for VTE through feature selection techniques.
Conducted internal validation using Bootstrap resampling to ensure model stability.
Model achieved sensitivity of 0.79 and specificity of 0.74.
AUC of the model was 0.82, indicating good discrimination.
Identified several independent predictors, including age and previous VTE.
Model demonstrated clinical applicability confirmed by decision curve analysis.
Internal validation affirmed stability of the prediction model.

Resumen

Renal insufficiency (RI) patients have an increased risk of venous thromboembolism (VTE), but validated risk assessment models tailored to this population are scarce. We conducted this study to provide a personalized VTE risk assessment for patients with RI. A total of 698 patients with RI hospitalized at the First Affiliated Hospital of Kunming Medical University from January 2022 to December 2023 were included in the study, including 142 with VTE and 556 without. Multiple imputation techniques and the Synthetic Minority Oversampling Technique were used to address the missing data and class imbalance, respectively. We used LASSO regression, random forest, and bidirectional stepwise logistic regression for feature selection and developed a logistic regression prediction model for VTE in RI patients. We identified age, length of hospital stays, lower mean arterial pressure, previous VTE, fever, lower limb pain, higher RDW-CV, D-dimer, and central venous catheterization as independent predictors for VTE in RI. The model demonstrated good discrimination and calibration, with a sensitivity of 0.79, specificity of 0.74, AUC of 0.82 (95% CI: 0.80–0.85), F1 score of 0.77, and Brier score of 0.17. The Spiegelhalter Z-test showed a p-value of 0.556, and decision curve analysis confirmed clinical applicability. Internal validation with Bootstrap resamples affirmed its stability. Our model helps identify high-risk individuals who may benefit from thromboprophylaxis. Large-scale, multicenter studies are needed to validate these findings.

Me gusta

Guardar

Ver artículo completo