402P Real world adoption and utility of AI in oncology in India: A nationwide survey of medical oncologists

Abstract

Background: Artificial intelligence (AI) & publicly available LLMS are making their presence in Oncology at a rapid pace, yet real-world insights into its adoption, utility, and challenges amongst medical oncologists remain limited.These insights could help policy makers & oncology associations in framing academic agendas, curriculum and guidelines for clinical practise.Methods: A cross-sectional, electronic survey of medical oncologists across India was conducted in August.A 30-item questionnaire disseminated via oncology networks on Whatsapp groups and Social media assessed AI awareness, usage, perceived utility, and concerns.Responses were descriptively analyzed.Results: 135 medical oncologists participated; Most practiced in private tertiary hospitals (36.2%) or medical colleges (32.5%).77% reported active AI use.Social media (45.9%) was the commonest knowledge source.AI was perceived as most impactful in diagnostics, radiology, NGS interpretation, (75%); 58% noted it's potential for clinical trials and drug discovery.Only 11% used AI regularly for clinical decision support, 50% occasionally; 42% found it helpful in difficult cases, while 41% encountered hallucinations occasionally .Half used AI for research tasks, while only one-third used it for hospital related documentation work.Most (74%) stressed the need for country-specific AI models ; 56% preferring open-source access to deidentifed datasets.58.5% cited confidentiality concerns with public LLMs.55% percent reported limited understanding of commonly used AI terms.56% desired short workshops for training in AI. 48.1% expected AI to transform oncology significantly within five years.Conclusions: AI adoption among Indian oncologists is widespread, especially for it's perceived impact in diagnostics, imaging &report interpretation.Limited usage for clinical decision support on a regular basis underscores unmet needs.Enhancing its integration into clinical workflows &hospital documentation can help in efficient time management.Knowledge gaps in basic AI concepts remain significant, highlighting need for targeted workshops & structured training.Development of secure countryspecific open-source oncology datasets should be prioritized.

Ähnliche Arbeiten