Most inventory models address demand uncertainty but give limited attention to lead time uncertainty, particularly endogenous lead time uncertainty, and often ignore stockpile policies and large-scale disruptions. We propose a two-layer, demand-driven optimization framework that jointly models exogenous demand uncertainty and decision-dependent lead time uncertainty under partially backlogged demand. We developed a stochastic and robust framework that uses data-driven multiple uncertainty sets and a rolling horizon to control conservatism. We reformulate the resulting model into a tractable mixed-integer linear program and evaluate inventory policies using real hospital data (NYU Langone Health), demonstrating improved cost efficiency and system resilience.
