A diagnostic prediction model for cardiovascular diseases (CVDs) in patients with psoriasis.

Abstract

Objective: Individuals with psoriasis are related to a significantly increased risk of cardiovascular diseases (CVDs), the major cause of death among psoriasis patients. Prompt diagnosis and intervention of CVDs can effectively retard the progression of the disease. This study developed and validated the CVDs diagnostic prediction model for psoriasis patients.

Methods: Medical records from psoriasis patients admitted to Beijing Hospital of Traditional Chinese Medicine between January 2009 and September 2024 were reviewed retrospectively. Patients were randomized as training and validation sets at the 7:3 ratio. We then selected variables through univariate logistic regression and least absolute shrinkage and selection operator (LASSO). The screened factors were subsequently incorporated in a multivariate logistic regression model for establishing the diagnostic nomogram. Moreover, this constructed model was validated internally and externally, and its performance was compared with a previous model.

Results: In this study, altogether 2,685 psoriasis patients were included. Five variables were finally selected for nomogram construction, which were age, hypertension, diabetes, dyslipidemia, and fasting blood glucose (FBG). According to our results, this model achieved favorable discrimination ability, and the area under the curve (AUC) values after 500 bootstrap resampling was 0.9355 (95% CI, 0.9219-0.9491) and 0.9118 (95% CI, 0.8899-0.9338) for training and validation sets, separately. Besides, calibration curves closely matched predicted and real values for both sets. Further, as indicated by DCA results, this model showed a high net benefit at predicted probabilities below 79% and 80% of training and validation sets, separately. In total, 188 psoriasis patients were enrolled in this work, with NHANES publicly available data being utilized for external validation. The corrected AUC was 0.8293 (95% CI, 0.7574-0.9012), and the calibration and DCA curves demonstrated good accuracy and clinical utility. Additionally, the model showed an increased AUC compared with a previously published diagnostic model. Its net reclassification index (NRI) and discrimination improvement index (IDI) were positive, showing that our model was superior to the previous model.

Conclusion: This study provides a cost-effective and practical tool that can assist physicians in identifying psoriasis patients at a higher CVDs risk. This may facilitate early disease diagnosis and intervention.