Presentation (CTI)

OBJECTIVES: Evidence-based policymaking (EBPM) requires integrating scientific evidence with multistakeholder perspectives within complex institutional contexts. Traditional consensus methods face well-known limitations in anonymity, scalability, and efficiency. The present work represents the development and real-world application of Polisis®, a hybrid human-AI enhanced electronic Nominal Group Technique (e-NGT) designed to support rapid, transparent, and participatory health policy formulation.
METHODS: he Polisis® model adapts the classical NGT across four structured phases: (1) asynchronous online brainstorming via a secure digital platform, with full anonymization of inputs to reduce conformity bias; (2) independent aggregation and policy recommendations formulation by two methodologists; (3) moderated in-person discussion with anonymized voting (Likert scale 1-5; consensus threshold ≥75% scoring ≥4) and real-time AI-assisted refinement, in which a pipeline of domain-specialized large language model agents synthesize panelist comments and generate revised recommendation with explicit change logs; (4) digital prioritization and ranking.
RESULTS: The methodology was applied to generate consensus recommendations on the integration of contrast-enhanced mammography (CEM) into diagnostic pathways for women with dense breasts in Italy, involving an 11-member multidisciplinary expert panel. Seven draft policy recommendations were developed from the online brainstorming phase across three thematic areas: institutional barriers, organizational impact, and patient awareness. Following a moderated discussion and AI-assisted refinement, eight final consensus recommendations were produced. All eight achieved 100% agreement at final voting, covering CEM reimbursement integration into the Essential Levels of Care (LEA), risk-stratified diagnostic pathways, digital interoperability, infrastructure mapping, accreditation standards, professional training, and public awareness. The hybrid workflow reduced facilitator burden, improved traceability of all revisions, and enabled comprehensive inclusion of panelist feedback without extending session time.
CONCLUSIONS: The Polisis® method demonstrates that structured, AI-enhanced e-NGT is feasible, scalable, and methodologically rigorous for health policy consensus-building. It addresses key limitations of traditional approaches while aligning with ACCORD reporting standards.