The goal of this observational study is to develop and validate artificial intelligence (AI)-driven models for improving the diagnosis of Primary Sjögren's Syndrome (PSS) using routine laboratory test data. The main question it aims to answer is:

Can AI-based algorithms accurately diagnose Primary Sjögren's Syndrome by analyzing laboratory test results, and do they outperform traditional diagnostic criteria in Chinese populations?

Researchers will retrospectively analyze anonymized clinical records and laboratory data (e.g., autoantibody levels, inflammatory markers) from patients with suspected or confirmed PSS across multiple medical centers in China. No new interventions will be administered, as the study utilizes existing historical data to train and validate the AI models. The performance of AI algorithms will be compared with current diagnostic standards (e.g., ACR/EULAR criteria) in terms of sensitivity, specificity, and clinical utility.

Inclusion Criteria:

• Patients with clinician-diagnosed primary Sjögren's syndrome (pSS) meeting the 2016 ACR/EULAR or 2002 ACEG classification criteria (objective oral/ocular dryness, positive anti-SSA/Ro antibodies, or focal lymphocytic sialadenitis on biopsy).
• Control groups: Individuals with non-pSS autoimmune diseases (e.g., rheumatoid arthritis, systemic lupus erythematosus) or non-autoimmune conditions (e.g., dry eye/sicca symptoms without systemic autoimmunity).

Exclusion Criteria:

• Pregnancy, breastfeeding, with a clear diagnosis of other autoimmune diseases, severe infection and malignant tumors.
• Not newly diagnosed in any of the hospitals.
• Without any available laboratory tests.

This study includes patients with primary Sjögren's syndrome (pSS; diagnosed via 2016 ACR/EULAR or 2002 ACEG criteria: oral/ocular dryness, anti-SSA/Ro+, or positive biopsy) and controls (non-pSS autoimmune diseases or non-autoimmune sicca symptoms). Eligible participants were newly diagnosed at participating hospitals with complete lab data. Exclusions: pregnancy/breastfeeding, other autoimmune diseases, malignancies, severe infections, or missing lab records. Retrospective data spans multicenter clinical records, ensuring real-world validity.