[Construction and validation of a predictive model for early occurrence of lower extremity deep venous thrombosis in ICU patients with sepsis].

Objective: To investigate the risk factors of lower extremity deep venous thrombosis (LEDVT) in patients with sepsis during hospitalization in intensive care unit (ICU), and to construct a nomogram prediction model of LEDVT in sepsis patients in the ICU based on the critical care scores combined with inflammatory markers, and to validate its effectiveness in early prediction.

Methods: 726 sepsis patients admitted to the ICU of the Affiliated Hospital of Jining Medical University from January 2015 to December 2021 were retrospectively included as the training set to construct the prediction model. In addition, 213 sepsis patients admitted to the ICU of the Affiliated Hospital of Jining Medical University from January 2022 to June 2023 were retrospectively included as the validation set to verify the performance of the prediction model. Clinical data of patients were collected, such as demographic information, vital signs at the time of admission to the ICU, underlying diseases, past history, various types of scores within 24 hours of admission to the ICU, the first laboratory indexes of admission to the ICU, lower extremity venous ultrasound results, treatment, and prognostic indexes. Lasso regression analysis was used to screen the influencing factors for the occurrence of LEDVT in sepsis patients, and the results of Logistic regression analysis were synthesized to construct a nomogram model. The nomogram model was evaluated by receiver operator characteristic curve (ROC curve), calibration curve, clinical impact curve (CIC) and decision curve analysis (DCA).

Results: The incidence of LEDVT after ICU admission was 21.5% (156/726) in the training set of sepsis patients and 21.6% (46/213) in the validation set of sepsis patients. The baseline data of patients in both training and validation sets were comparable. Lasso regression analysis showed that seven independent variables were screened from 67 parameters to be associated with the occurrence of LEDVT in patients with sepsis. Logistic regression analysis showed that the age [odds ratio (OR) = 1.03, 95% confidence interval (95%CI) was 1.01 to 1.04, P < 0.001], body mass index (BMI: OR = 1.05, 95%CI was 1.01 to 1.09, P = 0.009), venous thromboembolism (VTE) score (OR = 1.20, 95%CI was 1.11 to 1.29, P < 0.001), activated partial thromboplastin time (APTT: OR = 0.98, 95%CI was 0.97 to 0.99, P = 0.009), D-dimer (OR = 1.03, 95%CI was 1.01 to 1.04, P < 0.001), skin or soft-tissue infection (OR = 2.53, 95%CI was 1.29 to 4.98, P = 0.007), and femoral venous cannulation (OR = 3.72, 95%CI was 2.50 to 5.54, P < 0.001) were the independent influences on the occurrence of LEDVT in patients with sepsis. The nomogram model was constructed by combining the above variables, and the ROC curve analysis showed that the area under the curve (AUC) of the nomogram model for predicting the occurrence of LEDVT in patients with sepsis was 0.793 (95%CI was 0.746 to 0.841), and the AUC in the validation set was 0.844 (95%CI was 0.786 to 0.901). The calibration curve showed that its predicted probability was in good agreement with the actual probabilities were in good agreement, and both CIC and DCA curves suggested a favorable net clinical benefit.

Conclusions: The nomogram model based on the critical illness scores combined with inflammatory markers can be used for early prediction of LEDVT in ICU sepsis patients, which helps clinicians to identify the risk factors for LEDVT in sepsis patients earlier, so as to achieve early treatment.