Development and validation of a predictive risk model for post-stroke deep vein thrombosis

Deep vein thrombosis (DVT) represents a common and potentially fatal complication following acute ischemic stroke (AIS). However, early risk stratification is often difficult due to the asymptomatic nature of post-stroke DVT and the scarcity of AIS-specific predictive instruments. This study aimed to develop a nomogram utilizing baseline patient characteristics to forecast DVT risk after AIS. In this retrospective study, 848 AIS patients were analyzed, with 79 subsequently developing DVT. Least Absolute Shrinkage and Selection Operator (LASSO) regression and multivariable logistic regression identified five independent predictors: age (odds ratio [OR] = 1.074), fibrinogen (OR = 1.700), D-dimer (OR = 1.123), platelet count (OR = 1.001), and sex (OR = 0.475). A predictive nomogram was then constructed based on these variables. The model achieved an initial area under the receiver operating characteristic curve (AUC) of 0.809 and a Brier score of 0.074. Internal validation, conducted via 1,000-resample bootstrapping, yielded a stable optimism-corrected AUC of 0.800. Calibration plots demonstrated excellent concordance between predicted probabilities and observed outcomes. Decision curve analysis (DCA) indicated a positive net clinical benefit across a threshold probability range of 1%-91%, encompassing the clinically relevant 5%-20% decision window. In conclusion, this study successfully developed an exploratory nomogram for the early estimation of DVT risk in patients with AIS.