Injury burden and mixed/proximal deep vein thrombosis after patellar fracture: a multicenter retrospective cohort study of DVT-positive patients

Deep vein thrombosis (DVT) is a common complication after patellar fracture, but most patella-specific studies have focused on DVT occurrence rather than anatomical thrombus extent. This study examined factors associated with mixed/proximal DVT among patients who had already developed in-hospital DVT. This multicenter retrospective cohort study screened adult patients with patellar fractures treated between November 2013 and January 2023 at two tertiary-care level I trauma centers. Descriptive information from the overall eligible cohort was used for context, whereas the primary analysis was restricted to DVT-positive patients. DVT was classified as distal, mixed, or proximal according to thrombus location. The primary outcome was mixed/proximal versus distal DVT. Baseline characteristics were compared between groups, stability selection was performed using bootstrap resampling with lasso regularization, and final associations were estimated using Firth penalized logistic regression. Among 3319 eligible patients with patellar fractures, 1021 developed in-hospital DVT. Of these, 846 had distal DVT, 159 had mixed DVT, and 16 had proximal DVT; therefore, 175 patients were classified as having mixed/proximal DVT, accounting for 17.1% of DVT-positive patients. Compared with distal DVT, mixed/proximal DVT was associated with a higher prevalence of multiple fractures, longer time from injury to hospital treatment, lower albumin levels, and higher high-sensitivity C-reactive protein levels. Stability selection identified multiple fractures as the only predictor with meaningful selection frequency. In the final Firth model, multiple fractures were independently associated with mixed/proximal DVT (OR 2.50, 95% CI 1.80–3.50; P < 0.001). The clinical and full models showed modest apparent discrimination, with AUCs of 0.651 and 0.654, respectively. Bootstrap internal validation demonstrated reduced model performance after optimism correction, with an optimism-corrected AUC of 0.596. Among DVT-positive patients after patellar fracture, distal DVT was predominant, whereas mixed/proximal DVT represented a smaller but clinically relevant anatomical phenotype. Multiple fractures were the most stable factor associated with mixed/proximal DVT, suggesting that thrombus extent is linked to injury burden. Because model performance was modest and external validation was not performed, these findings should be interpreted as phenotype-oriented and hypothesis-generating rather than as a basis for immediate prediction-model use.