(Management Case Study) Managing Inventory: Using Predictive Models to Control Inventory Costs

Purpose: The Mayo Clinic Cancer Center recently moved to a perpetual inventory system. The new system brought several opportunities but also many challenges with aligning the amount of drug ordered compared to what was utilized over a defined period of time. This resulted in a need for the cancer center practice to seek alternative strategies to optimize and predict the amount of drug needed to cover patient treatments during a specified time period for a subset of expensive oncology treatment agents. This project aimed to look at different predictive models to reduce carrying costs and algin inventory purchases with actual utilization.

Methods: The cancer center pharmacy selected a subset of high-cost medications with varying degrees of usage patterns over a twelve-week period. During this time period, all administration and dispense data was reviewed in addition to deviations to the original scheduled treatment cycle. The project team reviewed drug indication, approved treatment schedule, and the most common contributing factors for dose delays/reductions found in the pivotal clinical trials. This information was then used to evaluate if similar trends and rates reported in the pivotal clinical trials were observed in the identified data set. This information was recorded and stored in a secure protected file. The project group then took each of the high-cost medications in question and utilized EHR reports to predict anticipated patient needs based upon pre-ordered and singed treatments for scheduled patients. The project group used the usage data over the twelve-week period and created formulas to predict inventory needs. The results were shared with our purchasing team for consideration to change daily purchasing practices and compared to actual utilization of the target high-cost medications.

Results: The alternative strategy used to predict inventory within the Mayo Clinic Cancer Center Pharmacy resulted in a deviation from the historical ordering practice. Three of the high-cost drugs needed to be ordered more frequently as our model showed minimal risk of dose delays, reductions, interruptions, and period of time before patients stopped or discontinued therapy. This resulted in fewer drop ship orders for this subset of treatments and reduced strain and workload on our purchasing team. Two of the high-cost drugs needed to be ordered less frequently as our model showed a higher risk for dose delays, interruptions, and reductions. This resulted in reduced carrying costs for these treatments and more space within the pharmacy storage area to accommodate other treatments. The tool used to predict inventory reduced stockouts for those items that were understocked and prevented stockouts from occurring for those that were overstocked after adjustments were made.

Conclusion: Using the tool for predicting inventory of high-cost drugs in the cancer center pharmacy was effective in reducing costs, workload, and stockout risks. By making adjustment to ordering practices based on the model's predictions, the pharmacy was able to optimize its inventory management, reduce carrying costs, and minimize stockouts while controlling costs. This project demonstrates the feasibility and effectiveness of applying a predictive model to optimize inventory for high-cost cancer drugs.