Mitigating machine learning bias between high income and low–middle income countries for enhanced model fairness and generalizability | Synapse