Validating Navya Earthshot: An AI-enabled point-of-care solution for guideline-adherent treatment planning in a decentralized cancer care model.

Abstract

1626 Background: Adherence to guidelines increases overall survival, globally. In resource-constrained settings, ~ 30% of patients receive undertreatment or overtreatment. Despite significant investment in decentralized cancer care—with tertiary hub centers providing support to non-specialized spoke centers—shortage of oncologists creates a knowledge gap, which may be partially addressed by clinically validated AI solutions. Navya is a clinically validated AI solution for cancer patients in use since 2014 which matches patient-specific data to evidence and generates treatment recommendations vetted via asynchronous expert review. Navya Earthshot is a new, provider facing solution for point of care cancer treatment planning for non-specialized providers, and is built as an AI driven search interface on Navya’s validated domain model supporting subspecialized expert opinions in oncology. Methods: This multicenter, prospective validation took place at 25 hospitals across India participating in a decentralized cancer care model. All patients with breast, oral and lung cancerbetween January and June 2024 with all decisions (curative and palliative; local and systemic therapies) were included. Navya Earthshot matched input patient data available in the patient medical record with National Cancer Grid (NCG) guidelines, and output evidence based treatment plans at the point of care. The output was shared at each center, and concordance was scored by the tumor board/treating oncologists, as well as by a group of domain experts experienced in analyzing NCG guidelines. Results: Navya Earthshot processed 1787 decisions in a decentralized cancer care system, pertaining to 40% (725) breast, 20% (351) oral, 40% (711) lung cancer diagnoses respectively. Patients were well represented with respect to age ( < 45 years (23%) and > 45years (77%), and early stage (24%); advanced stage (58%) and incomplete diagnostic workup (18%)). Of the 1787 decisions, Navya Earthshot output referred 27% (478) to hub center tumor boards due to presence of uncommon histologies or scenarios not covered by the NCG guidelines (3rd line therapy etc.). In the remainder 73% (1309) decisions, Navya Earthshot output diagnostic or treatment plans. Of these, 85% (1114/1309) were scored concordant with NCG guidelines, and adopted by the local treating oncologist. The remaining 15% (195) decisions were referred to a hub center for treatment planning. Conclusions: Navya Earthshot can improve capacity of oncologists in resource-constrained settings and enhance adherence to guideline-driven care in a decentralized cancer care model. In a majority of cases, this point-of-care solution can improve access to care locally, reduce reliance on tertiary hub centers, and improve patient outcomes, globally.