Harnessing conversational generative artificial intelligence and natural language processing to address the colorectal cancer screening adherence gap: the Azores endeavour

This is the authors' abstract. We don't add key points for this paper.

Organised colorectal cancer (CRC) screening reduces mortality, yet participation routinely falls short of policy targets, a gap most acute in structurally disadvantaged, insular, and rapidly ageing populations. The Azores, an outermost European Union archipelago, exemplifies this challenge: CRC accounts for 10% of regional cancer diagnoses and is the second leading oncological cause of potential years of life lost, with 2024 mortality (24.9 per 100,000) exceeding mainland Portugal. Despite a mature organised screening programme (ROCCRA), archipelago-wide participation stands at just 30.3%, with roughly one in five residents who consent and receive a home test kit never returning it. This perspective article argues that conversational artificial intelligence based on large language models (LLMs), the current generation of natural language processing (NLP) technology, capable of understanding and responding to patient concerns in real time, represents a promising approach for bridging this gap. Because the barriers identified in the literature are well, inherently dialogic, an interactive, multi-channel conversational cascade is better suited to address cognitive and operational friction than static reminders. Specifically, this paper synthesises presumed drivers of non-adherence, inferred from the literature and considered likely to apply to the Azorean context, reviews the evidence supporting conversational and voice-enabled outreach in cancer screening, and uses regional data to model the potential impact of a well-targeted voice AI deployment. Beyond the Azores, this framework illustrates how NLP-enabled conversational AI could support more responsive, equitable, and scalable public health screening programmes in geographically dispersed and underserved populations.