Intelligent inventory management approaches for perishable pharmaceutical products in a healthcare supply chain

This study develops intelligent inventory management (IIM) approaches for managing perishable pharmaceutical products in a healthcare supply chain consisting of multiple regional hospitals and a central warehouse. The purchase price of the product is assumed to be age dependent, and the age distribution of the product is assumed to be known and is shared among all the supply chain members. Each agent representing a regional hospital or the central warehouse in the supply chain replenishes its stocks from its upstream agent. In IIM approaches, reinforcement learning methods, specifically Q-learning and Deep Q-network, are used to construct inventory policies. The IIM policies provide hospitals with near-optimal order quantities and remaining life distributions for products they order. Through many test instances, the performance of the IIM policies is compared to that of periodic review (R, s, S) policies obtained through a stochastic mixed integer programming model, solved using CPLEX and a genetic algorithm. The computational results demonstrate that the IIM policies are more cost-effective than the (R, s, S) policies, although the genetic algorithm has a speed advantage. Moreover, as compared with the (R, s, S) polices, the IIM polices have a lower possibility of product shortage, and thereby a higher service level for the patients, and a lower risk of product expiration. By implementing the IIM approaches, a healthcare system may also save storage space, in addition to having a lower total inventory cost. Sensitivity analyses are performed to derive managerial insights on the performance of the policies when the values of the key cost parameters change.