Establishment of prediction model of deep venous thrombosis formation of lower extremity after general surgery

Objective To investigate the value of thrombelastography (TEG) and conventional coagulation indexes in predicting the formation of deep vein thrombosis (DVT) of lower extremities after general surgery, and to establish a prediction model.

Methods A total of 272 patients undergoing general surgery from January 2018 to January 2021 were selected as research objects. According to the postoperative Doppler ultrasound results, they were divided into DVT group (249 cases) and non-DVT group (23 cases). The prothrombin time (PT), activated partial prothrombin time (APTT), thrombin time (TT), fibrinogen (Fib), D-dimer, coagulation reaction time (R), blood clotting time (K), α-angle and maximum amplitude (MA) were compared between the two groups. The factors related to thrombosis were screened, and Logistic regression analysis was used to establish a prediction model.

Results The age of patients in the DVT group was significantly higher than that in the non-DVT group (P < 0.05); the operation time of the DVT group was significantly longer than that of the non-DVT group (P < 0.05). The level of D-dimer in the DVT group was significantly higher than that in the non-DVT group (P < 0.05); the preoperative PT, APTT and TT of the DVT group were significantly shorter than those of the non-DVT group (P < 0.05). The preoperative K, α-angle and MA of the DVT group were significantly longer or greater than those of the non-DVT group (P < 0.05); the preoperative R of the DVT group was significantly shorter than that of the non-DVT group (P < 0.05). Preoperative α-angle, MA, K, D-dimer and PT had moderate predictive value for DVTarea under the curve (AUC) was 0.7 to 0.9. Nine preoperative variables including age, operation time, PT, APTT, TT, D-dimer, R, α-angle (K was not included because of the similar meaning of with α-angle) and MA were included in binary Logistic regression analysis to obtain the prediction model of postoperative thrombosis. The AUC of receiver operating characteristic (ROC) curve of the model was 0.964 (95%CI, 0.934 to 0.983, P < 0.05). When the Youden index was the maximum, the corresponding optimal cut-off value was 0.174, the sensitivity was 91.30%, and the specificity was 95.18%.

Conclusion The prediction model based on 9 preoperative variables including age, operation time, PT, APTT, TT, D-dimer as well as R, α-angle and MA in TEG can better screen high-risk patients with DVT.