医学
回顾性队列研究
外科
血栓形成
深静脉
放射科
试验预测值
静脉血栓形成
血管通路
风险评估
血管疾病
梅德林
肱动脉
文本挖掘
风险因素
置信区间
超声科
疾病严重程度
摘要
Objective: Postoperative patients with indwelling vascular access devices (VADs) face a high risk of lower extremity deep vein thrombosis (DVT). This study sought to develop a VAD-specific DVT risk prediction model and verify the efficacy of a targeted three-phase VAD management system. Methods: A retrospective cohort study enrolled 156 postoperative VAD users (August 2023-August 2024), equally assigned to a conventional nursing group and an access management group. Clinical data were collected, and variables were screened via clinical relevance and univariate logistic regression (P<0.05), followed by multivariate logistic regression to build the prediction model. The model's performance was validated using ROC curves, Hosmer-Lemeshow test, and Brier score, with 5-fold cross-validation for stability. The three-phase system included pre-infusion VAD optimization, intra-infusion targeted monitoring, and post-infusion risk stratification with DVT screening. Results: Central venous catheter (CVC) use (OR=2.92, P=0.005) and prior thrombosis history (OR=3.61, P=0.002) were identified as independent predictors. The model exhibited good discriminative ability (AUC=0.83, 95% CI: 0.76-0.90) and stability (average AUC=0.81±0.03). DVT incidence showed a significant risk gradient (1.85% vs.6.94% vs.16.67%, P=0.013). The access management group had significantly lower VAD-related complications and DVT incidence, with higher DVT prevention effectiveness and cognitive scores (all P<0.05). Conclusion: The VAD-specific DVT risk prediction model achieves effective risk stratification for postoperative VAD users, and the three-phase management system markedly reduces VAD-related complications and lower extremity DVT incidence. This integrated strategy bridges the gap between VAD care and DVT prevention, providing actionable clinical guidance to enhance postoperative VAD use safety with evidence-based support.
科研通智能强力驱动
Strongly Powered by AbleSci AI