Machine learning-based prediction of the post-thrombotic syndrome: Model development and validation study.

Background: Prevention is highly involved in reducing the incidence of post-thrombotic syndrome (PTS). We aimed to develop accurate models with machine learning (ML) algorithms to predict whether PTS would occur within 24 months. Materials and methods: The clinical data used for model building were obtained from the Acute Venous Thrombosis: Thrombus Removal with Adjunctive Catheter-Directed Thrombolysis study and the external validation cohort was acquired from the Sun Yat-sen Memorial Hospital in China. The main outcome was defined as the occurrence of PTS events (Villalta score ≥5). Twenty-three clinical variables were included, and four ML algorithms were applied to build the models. For discrimination and calibration, F scores were used to evaluate the prediction ability of the models. The external validation cohort was divided into ten groups based on the risk estimate deciles to identify the hazard threshold. Results: In total, 555 patients with deep vein thrombosis (DVT) were included to build models using ML algorithms, and the models were further validated in a Chinese cohort comprising 117 patients. When predicting PTS within 2 years after acute DVT, logistic regression based on gradient descent and L1 regularization got the highest area under the curve (AUC) of 0.83 (95% CI:0.76-0.89) in external validation. When considering model performance in both the derivation and external validation cohorts, the eXtreme gradient boosting and gradient boosting decision tree models had similar results and presented better stability and generalization. The external validation cohort was divided into low, intermediate, and high-risk groups with the prediction probability of 0.3 and 0.4 as critical points. Conclusion: Machine learning models built for PTS had accurate prediction ability and stable generalization, which can further facilitate clinical decision-making, with potentially important implications for selecting patients who will benefit from endovascular surgery.

Peripapillary Microvasculature Predicts the Incidence and Development of Diabetic Retinopathy: An SS-OCTA Study.

PURPOSE: To examine the associations of peripapillary microvascular metrics with diabetic retinopathy (DR) incidence and development using swept-source optical coherence tomography angiography (SS-OCTA). DESIGN: Prospective cohort study. METHODS: A total of 1033 eyes from 1033 patients with type 2 diabetes mellitus (T2DM) were included, with 2-year follow-up. The peripapillary microvascular metrics at the superficial capillary plexus (SCP) were measured by SS-OCTA at the baseline, including peripapillary vascular density (pVD) and peripapillary vascular length density (pVLD). The DR incidence and progression were evaluated with 7 standard fields of stereoscopic color fundus photographs. The associations were tested with logistic regression models after adjusting for established risk factors and confounding factors. The prediction value of OCTA metrics was examined with the elevation of area under the receiver operating characteristic curve (AUROC). RESULTS: The 2-year incidence of diabetic retinopathy (DR) was 25.1% (n = 222) in non-DR (NDR) eyes, 7.4% DR progression (n = 11) in DR eyes, and 4.17% RDR eyes (n = 43) in all eyes. After adjusting for established factors, lower whole-image pVD (wi-pVD) (relative risk [RR] = 0.81; 95% CI = 0.68-0.96; P = .015), circular pVD (circ-pVD) (RR = 0.79; 95% CI = 0.66-0.95; P = .013), whole-image pVLD (wi-pVLD) (RR = 0.79; 95% CI = 0.67-0.94; P = .008), and circular pVLD (circ-pVLD) (RR = 0.76; 95% CI = 0.63-0.91; P = .003) were significantly associated with increased risk of DR incidence; wi-pVD (RR = 0.48; 95% CI = 0.35-0.67; P < .001), circ-pVD (RR = 0.65; 95% CI = 0.45-0.94; P = .023), and wi-pVLD (RR = 0.46; 95% CI = 0.33-0.66; P < .001) were associated with incident risk of RDR. Both pVD and pVLD of SCP were not significantly associated with DR progression. The AUROC for the DR incidence risk prediction model increased from 0.631 to 0.658 (4.28%; P = .041) by circ-pVLD; the AUC of the RDR incidence risk prediction model increased from 0.631 to 0.752 by wi-pVD (19.18%; P = .009), to 0.752 by circ-pVD (19.18%; P=.009), and to 0.752 by wi-pVLD (19.18%; P = .009). CONCLUSION: Lower pVD and pVLD of SCP are associated with 2-year incident DR and RDR among the T2DM population. The peripapillary metrics imaged by SS-OCTA can provide additional value to the prediction of DR incidence and development.

Development of SAB model for predicting mortality in intensive care unit after aortic aneurysm surgery.

BACKGROUND: Aortic aneurysm (AA) patients after vascular surgery are at high risk of death, some of them need intensive care. Our aim was to develop a simplified model with baseline data within 24 hours of intensive care unit (ICU) admission to early predict mortality. METHODS: Univariate analysis and least absolute shrinkage and selection operator were used to select important variables, which were then taken into logistic regression to fit the model. Discrimination and validation were used to evaluate the performance of the model. Bootstrap method was conducted to perform internal validation. Finally, decision clinical analysis curve was used to test the clinical usefulness of the model. RESULTS: We obtained baseline data of 482 AA patients from Medical Information Mart for Intensive Care III database, 33 (6.8%) of whom died in ICU. Our final model contained three variables and was called SAB model based on initials of three items [Sepsis, Anion gap, Bicarbonate (SAB)]. Area under the curve of SAB was 0.904 (95% CI: 0.841-0.967) while brier score was 0.043 (95% CI: 0.028-0.057). After internal validation, corrected area under the curve was 0.898 and brier score was 0.045, which showed good prediction ability of SAB model. The model can be assessed on https://vascularmodel.shinyapps.io/AorticAneurysm/. CONCLUSIONS: SAB model derived in this study can be easily used to predict in-ICU mortality of AA patients after surgery precisely.