Akademik Veri Yönetim Sistemi
NORTHERN CLINICS OF ISTANBUL = ISTANBUL KUZEY KLINIKLERI, cilt.13, sa.2, ss.230-242, 2026 (ESCI, Scopus, TRDizin)
OBJECTIVE: Informed consent is the cornerstone of modern medical ethics, but current documentation systems negatively impact patient autonomy and clinical quality due to deficiencies in readability, comprehensibility, and standardization. These is-sues hinder patient participation and require innovative solutions. This study introduces the AI-powered LuminaConsent system to address standard deficiencies, comprehensibility issues, and efficiency constraints in urological informed consent documents. METHODS: In a three-armed comparative study, LuminaConsent (artificial intelligence), Turkish Urological Surgery Asso-ciation standard forms, and expert-developed documents were evaluated in 10 urological procedures. The system is based on the RAG architecture, which uses OpenAI’s GPT-4o-mini model and a special knowledge base consisting of 12 clinical publications. Three independent urology specialists conducted a blind evaluation using a 100-point scale across five areas: scientific content accuracy, patient communication effectiveness, quality of risk-benefit information, perioperative guidance, and legal-ethical compliance.
RESULTS: LuminaConsent achieved higher performance with mean scores of 82.33 points (SD±4.2) versus 78.77 points (SD±6.1) for professional society standards and 57.43 points (SD±3.8) for specialist documentation, representing statisti-cally significant improvements of 43.3% over specialist practices (p<0.001) and 4.5% over professional society standards (p<0.05). The system demonstrated consistent high-quality output across all procedures while generating comprehensive documentation within 96-180 seconds compared to traditional processes requiring multiple days.
CONCLUSION: LuminaConsent offers a pioneering model for systematic AI integration in clinical practice with its evidence-based content generation and bilingual processing capabilities. The findings support the potential to empower patient auton-omy, reduce application variations, and improve ethical standards.