Pilot Study of AI-Powered Clinical Decision Support System to Predict Flares in Psoriasis

Background and Aims: Psoriasis is a chronic, immune-mediated skin disease characterized by unpredictable flare-ups. Timely prediction of flares can facilitate early interventions and reduce disease burden. This pilot study aimed to develop and evaluate an artificial intelligence (AI)-powered clinical decision support system (CDSS) to predict impending psoriasis flares using routinely collected clinical and patient-reported data. Methods: A retrospective dataset of ۱,۲۰۰ patients with moderate-to-severe plaque psoriasis was extracted from the dermatology clinics of three academic hospitals (۲۰۱۸–۲۰۲۳). Variables included demographics, Psoriasis Area and Severity Index (PASI) scores, comorbidities, treatment regimens, adherence rates, and longitudinal patient-reported outcomes (PROs). Data preprocessing involved imputation of missing values, normalization, and feature engineering, including time-series transformation and lag feature creation. An ensemble AI model combining long short-term memory (LSTM) networks and gradient-boosted decision trees (XGBoost) was trained to classify the likelihood of a flare occurring within the next ۳۰ days. Model performance was assessed using stratified ۵-fold cross-validation. Metrics included accuracy, sensitivity, specificity, F۱-score, and area under the receiver operating characteristic curve (AUC-ROC). Results: The hybrid AI model achieved an AUC-ROC of ۰.۸۷, with sensitivity and specificity of ۸۲.۴% and ۷۹.۱%, respectively. The CDSS interface successfully generated real-time alerts for high-risk patients in a simulated deployment, showing a ۲۰% reduction in predicted flare-related hospital visits during internal validation. Conclusion: This pilot study demonstrates the feasibility and clinical promise of AI-based flare prediction in psoriasis management.

کلیدواژه ها:

Psoriasis ، Artificial Intelligence ، Decision Support Systems