Risk prediction model for psoriatic arthritis: NHANES data and multi-algorithm approach.

Area of Science:

• Rheumatology
• Medical Informatics
• Epidemiology

Background:

• Psoriatic arthritis (PsA) diagnosis in psoriasis patients can be challenging.
• Predictive models are needed to identify individuals at high risk for PsA development.
• Existing models may have limitations in terms of data sources or variable selection.

Purpose of the Study:

• To develop and validate a simplified predictive model for identifying psoriatic arthritis (PsA) in patients with psoriasis.
• To identify key predictors for PsA risk using a large national health database.

Main Methods:

• Utilized data from the National Health and Nutrition Examination Survey (NHANES) database.
• Employed variable selection techniques including least absolute shrinkage and selection operator, Boruta algorithm, random forest, and stepwise regression.
• Constructed and evaluated logistic regression models using ROC curves, PR curves, calibration plots, Brier scores, and decision curve analysis (DCA).

Main Results:

• The Boruta algorithm identified key predictors: age, fasting glucose, education level, thyroid disease, hypertension, and chronic bronchitis.
• The developed Borutamodel achieved an AUC of 0.781 on the training set and 0.780 on the testing set for ROC analysis.
• The model demonstrated good fit with Brier scores of 0.186 (testing) and 0.191 (training) and showed net clinical benefit across decision thresholds.

Conclusions:

• The Borutamodel is a promising tool for early risk assessment of psoriatic arthritis in psoriasis patients.
• This model effectively integrates demographic, clinical, and comorbidity data for improved PsA risk prediction.
• Leveraging national data and comprehensive variable analysis, this study offers a novel approach beyond traditional symptom-based assessments.