Scaling prior authorization for philanthropic funding with artificial intelligence (AI): Five-year outcomes from Indian Cancer Society Cancer Cure Fund (ICS-CCF) NavyaAI implementation.

Abstract

1610 Background: Indian Cancer Society - Cancer Cure Fund (ICS-CCF) is one of the largest charitable program for directly reimbursable treatment expenses to needy cancer patients. The prior authorization process at ICS-CCF requires the Due Diligence Team (DDT) and the Governing Advisory Council (GAC) to review every application for compliance with standardized treatment guidelines, expected cure rates, and approved costs. To augment and enhance this process, ICS-CCF evaluated and implemented the use of Artificial Intelligence (AI) to support application review and recommendations over a five-year study period. This study is a five year report on operational outcomes to reduce dependence on expert oncologists in the prior authorization setting using AI for care delivery. Methods: The NavyaAI platform is a clinically validated AI model that matches clinical data from beneficiary applications (input) with available clinical evidence and National Cancer Grid (NCG) guidelines, adapted to ICS-CCF approval criteria (output). Applications in which the AI output matched the input were classified as “recommended” and forwarded directly to the GAC for approval. All remaining cases were referred to the DDT, comprising a minimum of two expert oncologists, who reviewed the applications and forwarded recommendations to the GAC as appropriate. Concordance between NavyaAI and GAC decisions was assessed. The time spent by the DDT reviewing referred applications was recorded. DDT reviews were conducted weekly, and GAC reviews were conducted bi-monthly. Results: From April 2020 to November 2025, a total of 12001 (malignant solid tumors 63.77% (7653/12001) and hemato-lymphoid malignancies 36.23% (4348/12001) applications were reviewed by NavyaAI. Of these, 82.58% (9,910/12001) were recommended, 1.71% (205/12001) were rejected, and 15.72% (1,886/12001) were referred to the DDT. Concordance between NavyaAI and GAC decisions was 99.57% (9,867/9,910) for authorizations and 92.68% (190/205) for rejections. The DDT spent an average of 3 minutes reviewing each application referred by NavyaAI. On average, NavyaAI forwarded 35 applications directly to the GAC per week, saving a minimum of 210 minutes of oncologist review time weekly. Conclusions: NavyaAI can independently review and assess the majority (~84%) of beneficiary applications for recommendation or rejection with high concordance to expert decision-making. Further implementation of AI-assisted authorization models is warranted to scale their long-term impact on quality, efficiency, and scalability of philanthropic funding decisions within governmental and non-governmental healthcare schemes.