Decentralizing Healthcare Referrals: How Graph Neural Networks and Blockchain Can Bridge Gaps in LMICs for Equitable Care

Abstract

Healthcare referral systems are pivotal for equitable care delivery, yet inefficiencies persist
across low- and middle-income countries (LMICs), exacerbating disparities in access to specialist care.
This mixed-methods study integrates longitudinal data from Portugal’s National Health Service (12M
consultations, 2017–2022) and Indonesia’s National Health Insurance claims (1.7M beneficiaries) to
evaluate referral mechanisms. Graph neural networks (GNNs) reveal power-law distributed referrals
(?=2.3), where 20% of specialists handle 78% of cases, driven by professional affiliations (?=0.67,
p<0.001) that disproportionately marginalize rural providers. Logistic regression identifies urban
deprivation (OR=1.72, 95% CI: 1.45–2.04) and fragmented e-referral systems (OR=2.10, 95% CI: 1.88–
2.35) as key compliance barriers, particularly among youth (18–44 years) with 54% higher odds of nonattendance. A PRISMA-guided systematic review of 63 studies highlights 10 systemic gaps, including
inadequate patient tracking (58% of studies), provider workload (49%), and patient mistrust (37%).
While e-referral integration reduces median care delays by 67% (21 vs. 7 days, p=0.003) and duplicate
referrals by 41%, adoption challenges persist in 62% of Indonesian primary clinics due to fragmented
IT infrastructure. Policy recommendations emphasize (1) decentralizing referral networks through
GNN-driven analytics to prioritize underserved populations, (2) scaling interoperable e-referral
platforms with blockchain-backed tracking, and (3) implementing community-led digital literacy
programs to address youth and rural disparities. These reforms align with Sustainable Development Goal
(SDG) 3.8, offering a roadmap to mitigate inequities and optimize referral efficiency in LMIC health
systems.