Predictive Inequity: How Risk-Scoring Algorithms Perpetuate Socioeconomic Disparities in Public Health Resource Allocation

Automated risk-scoring tools are increasingly deployed by public health authorities to allocate scarce resources such as preventive interventions and early-detection screenings. This paper demonstrates, through analysis of three national deployment cases, that such tools systematically under-refer individuals from lower socioeconomic backgrounds due to training data biases and proxy variable selection. We propose a three-stage audit framework that health ministries should apply before and during deployment of any population-level risk model.