Preparedness of European pediatric oncologists to integrate AI in the clinical routine

Abstract

Background

Artificial intelligence (AI) holds promise in pediatric oncology, yet its full potential faces challenges. We undertook a survey aimed at assessing the viewpoints of European pediatric oncologists delving into their perceptions and expectations regarding the potential influence of AI in their clinical workflows.

Method

We conducted a survey by means of four hypothetical scenarios using AI and the Shinners Artificial Intelligence Perception (SHAIP) tool to assess healthcare professionals' perceptions of AI in pediatric oncology. We performed multinomial logistic regression to explore associations of responses to clinical scenarios with age and SHAIP scores.

Results

We obtained 140 responses and the analysis was performed on 108. The SHAIP questionnaire mean total score was 3.29 (SD 0.93) for the professional impact, and 2.37 (SD 0.61) for preparedness. Regarding the clinical scenarios, 34.9 % of respondents would ask for a procedure for confirming their diagnosis in case of discrepancy between AI decision support and human diagnosis; 55.8 % would be concerned about the generalizability an AI decision support system in case of lack of data from certain geographic areas during algorithm training; 47.6 % would feel uncomfortable in the informed consent process for an AI intervention; 10.2 % would no longer trust AI in case of a cyberattack affecting AI support for diagnosis.

Discussion

This survey underscores the importance of AI tools in pediatric oncology that incorporate human oversight in clinical decision-making and training AI algorithms with diverse and representative data. Our findings suggest that pediatric oncologists may not be adequately prepared for the seamless integration of AI in clinical practice.