Individualized predictions of early isolated distal deep vein thrombosis in patients with acute ischemic stroke: a retrospective study

BackgroundAlthough isolated distal deep vein thrombosis (IDDVT) is a clinical complication for acute ischemic stroke (AIS) patients, very few clinicians value it and few methods can predict early IDDVT. This study aimed to establish and validate an individualized predictive nomogram for the risk of early IDDVT in AIS patients.MethodsThis study enrolled 647 consecutive AIS patients who were randomly divided into a training cohort (n=431) and a validation cohort (n=216). Based on logistic analyses in training cohort, a nomogram was constructed to predict early IDDVT. The nomogram was then validated using area under the receiver operating characteristic curve (AUROC) and calibration plots.ResultsThe multivariate logistic regression analysis revealed that age, gender, lower limb paralysis, current pneumonia, atrial fibrillation and malignant tumor were independent risk factors of early IDDVT; these variables were integrated to construct the nomogram. Calibration plots revealed acceptable agreement between the predicted and actual IDDVT probabilities in both the training and validation cohorts. The nomogram had AUROC values of 0.767 (95% CI: 0.742-0.806) and 0.820 (95% CI: 0.762-0.869) in the training and validation cohorts, respectively. Additionally, in the validation cohort, the AUROC of the nomogram was higher than those of the other scores for predicting IDDVT.ConclusionsThe present nomogram provides clinicians with a novel and easy-to-use tool for the prediction of the individualized risk of IDDVT in the early stages of AIS, which would be helpful to initiate imaging examination and interventions timely.

Automated summary

This study established and validated a nomogram for predicting the risk of early IDDVT in AIS patients, based on independent risk factors including age, gender, lower limb paralysis, current pneumonia, atrial fibrillation, and malignant tumor. The nomogram showed high predictive accuracy in both training and validation cohorts, with AUROC values of 0.767 and 0.820 respectively.