This study aimed to develop a risk of psoriatic arthritis (PsA) predictive model for plaque psoriasis patients based on the available features.
Patients with plaque psoriasis or PsA were recruited. The characteristics, skin lesions, and nail clinical manifestations of the patients have been collected. The least absolute shrinkage was used to optimize feature selection, and logistic regression analysis was applied to further select features and build a PsA risk predictive model. Calibration, discrimination, and clinical utility of the prediction model were evaluated by using the calibration plot, C-index, the area under the curve (AUC), and decision curve analysis. Internal validation was performed using bootstrapping validation. The model was subjected to external validation with two separate cohorts.
Age at onset, duration, nail involvement, erythematous lunula, onychorrhexis, oil drop, and subungual hyperkeratosis were presented as predictors to perform the prediction nomogram. The predictive model showed good calibration and discrimination (C-index: 0.759; 95% CI: 0.707–0.811). The AUC of this prediction model was 0.7578092. Excellent performances of the C-index were reached in the internal validation and external cohort validation (0.741, 0.844, and 0.845). The decision curve indicated good effect of the PsA nomogram in guiding clinical practice.
This novel PsA nomogram could assess the risk of PsA in plaque psoriasis patients with good efficiency.