JMIR Publications

ABSTRACT

Background:

Individuals living in rural areas often face significant barriers to accessing specialized cancer care, including subspecialty oncology expertise (e.g., disease-specific specialists), multidisciplinary tumor boards, complex surgical planning, and supportive care services. Delivering cancer care via telehealth (i.e., tele-oncology) is a promising strategy for improving access to such care. Despite its proliferation during the COVID-19 pandemic, implementation barriers to tele-oncology persist, including gaps in patients’ technological literacy, limitations in rural broadband availability, and integrating tele-oncology into in-person clinics. Artificial intelligence (AI)-powered predictive analytics offers potential solutions to these implementation barriers by identifying patients at risk for technology access issues, optimizing resource allocation, and personalizing implementation approaches to address specific rural contexts.

Objective:

The objective of this scoping review was to examine the potential of artificial intelligence (AI)-powered predictive analytics to optimize tele-oncology implementation.

Methods:

We conducted a literature search across PubMed/MEDLINE, Embase, CINAHL, Web of Science, and IEEE Xplore for studies published between January 2015 and April 2025 addressing tele-oncology implementation, healthcare barriers, and AI applications in healthcare delivery. We identified 827 unique articles, screened 68 for full-text review, and ultimately included 14 articles that met our inclusion criteria. We extracted data related to implementation factors, telehealth modalities, AI applications, and outcomes, then synthesized this information thematically to identify ways in which AI-powered predictive analytics can improve tele-oncology implementation. We followed the PRISMA-ScR guidelines for reporting.

Results:

Our findings suggest that AI-powered predictive analytics could transform tele-oncology implementation through three key applications: (1) predicting patient-level implementation barriers before they arise, (2) optimizing resource allocation decisions across healthcare systems, and (3) enabling personalized implementation strategies tailored to specific rural contexts. However, implementing AI systems themselves presents challenges, including those related to technical infrastructure, workflow integration, provider and patient acceptance, and sustainability.

Conclusions:

Taken together, our results suggest that AI-powered predictive analytics have the potential to enhance tele-oncology implementation by predicting barriers before they arise, optimizing resource allocation, and enabling personalized implementation strategies, though important technical and practical implementation considerations must be addressed. We propose a three-phase research agenda to guide future work: developing robust data infrastructure and predictive models, conducting implementation science research on integrating AI-driven insights into clinical workflows, and examining approaches for scaling and sustaining successful AI-enhanced implementation models. Future research in this area should work to address these areas, tacking both the technical and practical implementation challenges of AI-powered tele-oncology care, while emphasizing stakeholder engagement, rigorous evaluation, and ensuring healthcare access for all throughout all phases.