In addition to the location optimization, the model also enables the assessment of a hospital in terms of the potential cost savings by adjusting the quantity of the supplies on the preference cards. To do so, we have first assessed the cost of the current config- uration setting of the hospital as a baseline and then evaluated the implementation of three alternatives to the current condition, including (1) optimized quantity of the items in each location, (2) optimized preference cards and (3) both optimized quantity of the items and preference cards. Our computational results show that by implementing each alternative, the hospital under consid- eration is able to save $291, $428 and $668 per day, respectively. Through the model analysis, it has been observed that un- optimized preference cards can lead to sending items via the case carts to an OR, therefore making them unavailable for other ORs. These items may eventually be placed back into storage without being used. This condition results in a temporary stock out and induces more cost on the system for emergency fulfillment, which could be avoided through optimizing preference cards. It has also been discussed that although the preference card optimization results in decreasing of unnecessary items on the case carts, the quantities of frequently used items would be increased to prevent some picking during the procedure. One should notice that we have estimated cost parameters based on our observations. For implementation purposes, a motion and time study may be conducted to help with more precise cost parameter calculation. Moreover, it is recommended that at the first step, a pilot project for a small group of frequently-used items be implemented to make sure that the patient outcomes and quality of care are not affected by the implementation. The result of the pilot project needs to be carefully monitored before enlarging the scope of the project. It is also worth noting that in our model, we did not directly account for holding inventory cost (or carrying inventory cost). The reason was that the holding inventory cost is associated with the physical space occupied by inventory, such as rent, utility costs, taxes, etc. However, in the problem of interest, the existing space in CSS, ORs, and Cores is already allocated for inventory. Therefore, any decision regarding not using a portion of the space will not result in cost savings. Thus, the model is more suitable for currently operating facilities, as opposed to those being designed. In that case, inventory holding cost needs to be estimated and added to the model along with the inventory level of each item in each location as a new decision variable. In addition, the model is not providing a recipe for each sur- geon to design a preference card. Rather, the model looks at the supplies' cumulative usage in all procedures performed in an OR by all surgeons on a daily basis. The supplies requested on the preference cards, which have been sent to the ORS via case carts, are also aggregated daily. Therefore, the model does not account for optimizing the preference cards at a detailed level, i.e., what should be listed on the preference card for a given surgeon and specific procedure. Surgeons play a key role in the design of preference cards and any optimization method should involve them in the decision-making process to assure surgeons that preference cards are designed in line with their expertise. However, the proposed method can be served to trigger a project for adjusting quantities on the preference cards based on historical data and surgeon preferences. Running alternative A discussed in the paper will The prime objective of this paper was to develop a robust stochastic mixed-integer model to provide a framework to answer the question of where and in what quantity each surgical supply should be stored, called location optimization, considering the fact that the actual usage of supplies in the ORS is uncertain. We have identified the costs involved in the process of using and restocking supplies through a real-life case study. However, the model is quite general and can be applied to any healthcare system.