Should AI results be disclosed in mammography reports? A randomised survey study of patient responses to concordant and discordant interpretations.
Authors
Affiliations (8)
Affiliations (8)
- Breast Imaging Division, Radiology Department, IEO European Institute of Oncology IRCCS, Milan, Italy. [email protected].
- Breast Radiology Unit, Radiology Department, Fondazione IRCCS Istituto Nazionale dei Tumori, Milan, Italy.
- Breast Imaging Division, Radiology Department, IEO European Institute of Oncology IRCCS, Milan, Italy.
- Applied Research Division for Cognitive and Psychological Science, IEO European Institute of Oncology IRCCS, Milan, Italy.
- Department of Psychology, Educational Science and Human Movement (SPPEFF), University of Palermo, Palermo, Italy.
- Department of Oncology and Hematology-Oncology, University of Milan Milan, Milan, Italy.
- Department of Medical Surgical Sciences and Translational Medicine, Sapienza University of Rome, Radiology Unit, Sant'Andrea University Hospital, Rome, Italy.
- Breast Imaging Division, Radiology Department, Foundation IRCCS Ca' Granda, Ospedale Maggiore Policlinico, Milan, Italy.
Abstract
To assess how disclosing artificial intelligence (AI) results, particularly discordant findings, affects patient trust, anxiety, follow-up intentions, and attitudes toward AI in mammography. The study also assessed whether adding an explanatory note mitigates adverse reactions. A cross-sectional randomised experimental survey was conducted among 600 women (mean age 55.4 ± 6.8 years) undergoing mammography in two academic centres in Milan, Italy, between January 2023 and January 2024. Participants were randomised into four hypothetical BI-RADS 1 scenarios: Radiologist Only (control), AI No-Flag (AI concordant with radiologist), AI Flagged (AI discordant false-positive), and AI Flagged + Explanation (discordant AI with contextual information). Primary outcomes included trust (0-100 scale), worry, second-opinion intent, legal action intent, and AI approval. Analyses involved ANOVA, chi-square tests, and logistic regression with Bonferroni correction. Disclosure of a discordant AI result significantly reduced trust in the radiologist (73.0 vs 90.1; p < 0.001), and increased anxiety (58.0% vs 16.0%; OR = 15.4), second-opinion intent (50.0% vs 8.7%; OR = 10.2), and legal action consideration (60.7% vs 38.7%; OR = 2.49). Adding explanatory context significantly mitigated these effects (e.g., anxiety: 25.3%; OR = 0.26). Compared to the Radiologist Only scenario, the AI Flagged + explanation scenario showed only a modest increase in anxiety (p = 0.04) and no significant trust reduction (p = 0.42). AI approval remained high (> 85%) across all groups. Disclosing discordant AI results reduces trust and increases anxiety, second-opinion intent, and legal concerns. Contextualised disclosure of AI results mitigates adverse emotional and behavioural responses, supporting its use as a communication strategy in AI-integrated mammography. Question Current guidelines lack clear recommendations on disclosing AI-generated mammography findings, creating uncertainty about patient trust, anxiety, and medicolegal implications of discordant results. Findings Disclosing discordant AI mammography findings reduced patient trust, increased anxiety, second-opinion seeking, and litigation intent; adding contextual explanations significantly mitigated these adverse effects. Clinical relevance Providing clear context about AI limitations in mammography reports mitigates patient anxiety, enhances trust in radiologists, and reduces unnecessary follow-up and potential medicolegal actions, supporting optimal patient communication during clinical implementation of AI.