Healthcare recommender systems often prioritize clinical relevance while underrepresenting socio-economic determinants of health, such as affordability, insurance coverage, and geographical accessibility. This limitation can produce clinically plausible but practically inaccessible physician recommendations. To address this problem, this study proposes a Neuro-Symbolic Path Reasoning (NSPR) framework for explainable physician recommendation. NSPR integrates semantic entity linking, knowledge graph path reasoning, and explicit constraint satisfaction modeling within a Multi-Constraint Knowledge Graph (MC-KG). The MC-KG was constructed from anonymized BioVisit telemedicine data, including 5000 patient queries, 500 physician profiles, and approximately 28,000 medical and constraint-related nodes. Patient queries are mapped to symptom entities using a fine-tuned biomedical language model, while candidate physicians are ranked through path-based semantic relevance and socio-economic feasibility functions. The final ranking score combines TransE-based path plausibility with constraint satisfaction terms for cost, location, and insurance compatibility. Experimental results indicate that NSPR achieves competitive ranking performance, with an NDCG@10 of 0.82, while reducing Cost Alignment Error (CAE) by 45% relative to Neural Collaborative Filtering. The framework also improves provider exposure fairness by reducing the Gini Index to 0.41 and achieves an explanation fidelity score of 4.2/5 in clinician evaluation. Ablation analysis indicates that constraint modeling is essential for reducing financial toxicity, whereas knowledge graph reasoning is necessary for maintaining clinical relevance. These findings indicate that NSPR provides a transparent and constraint-aware recommendation framework that balances medical relevance, accessibility, and fairness. They further suggest that neuro-symbolic reasoning can support more trustworthy and equitable physician recommendation in telemedicine settings.
