Abstract
As machine learning models become increasingly integrated into healthcare,structural inequities and social biases embedded in clinical data can beperpetuated or even amplified by data-driven models. In survival analysis,censoring and time dynamics can further add complexity to fair modeldevelopment. Additionally, algorithmic fairness approaches often overlookdisparities in cross-group rankings, e.g., high-risk Black patients may beranked below lower-risk White patients who do not experience the event ofmortality. Such misranking can reinforce biological essentialism and undermineequitable care. We propose a Fairness-Aware Survival Modeling (FASM), designedto mitigate algorithmic bias regarding both intra-group and cross-group riskrankings over time. Using breast cancer prognosis as a representative case andapplying FASM to SEER breast cancer data, we show that FASM substantiallyimproves fairness while preserving discrimination performance comparable tofairness-unaware survival models. Time-stratified evaluations show that FASMmaintains stable fairness over a 10-year horizon, with the greatestimprovements observed during the mid-term of follow-up. Our approach enablesthe development of survival models that prioritize both accuracy and equity inclinical decision-making, advancing fairness as a core principle in clinicalcare.