| Abstrakt: |
In this work, we simulate a pre-sampled demand of blood units for the blood bank of a tertiary care super specialty hospital specializing in cardiac care and neuro care to mimic the present system based on last year's data. Simulation is a useful tool to address various problems related to healthcare as it is timesaving and can behave close to the actual system. In this study, we simulate the blood bank for 2000 days. This work also considers separate inventory systems for fresh and normal demand types of blood units based upon the age of blood units demanded for different patients. We consider three inventory policies (s, S) policy, (s, Q) policy and (s, S, q, Q) policy for the blood bank. To find the respective best re-order level (s), order-up-to level (S) and fixed-order quantity (q and/or Q) of blood units, we initially develop an Integer Linear Programming (ILP) model for the (s, S), (s, Q) and (s, S, q, Q) policies without considering the compatibility across blood groups. To overcome scaling issues encountered with the ILP model, we have proposed an approach based on the Genetic Algorithm (GA) to find high-quality solutions for large problem instances within a reasonable time duration, thereby reducing the total wastage or amount of outdated blood units while satisfying the random demand. We have compared the random and tournament selection techniques used for GA and found the latter to perform better. After comparing the three policies, we find that both the (s, Q) and the (s, S, q, Q) policies are competitive, and they outperform the (s, S) policy. The need for fewer parameters and the relative ease of implementation makes the (s, Q) policy preferable to the (s, S, q, Q) policy. [ABSTRACT FROM AUTHOR] |
