Applying Big Data Analytics to Optimize Pharmacy Operations

headBy Alecia Lashier, Director, Software Systems Engineering

We’ve been touting for some time now how Pharmacy Intelligence can change how pharmacy organizations assess and measure the success of their various fulfillment processes. A common theme we hear from pharmacies is they want to streamline their workflow, unburden staff workloads, improve in-store efficiency and dispensing quality, and free up their staff to better deliver patient-facing services. They also want to innovate using smart data analysis. Pharmacy Intelligence enables them to do just that.

As proof of this, Innovation recently collaborated on a thesis paper with Binghamton University Watson Institute for Systems Excellence (BU WISE) on optimizing pharmacy operations. The paper, “Pharmacy Robotic Dispensing and Planogram Analysis Using Association Rule Mining with Prescription Data, details an optimized approach for determining prescribed medication associations within a high-volume pharmacy environment. The paper has been published in Expert Systems With Applications (ESWA), a leading artificial intelligence journal.

In general, we work closely with BU WISE to ensure that the system design principles we use and software algorithms we develop across our product line, make sense and can be proven analytically. One of our main goals is to continuously reevaluate and improve on these practices, so we can push the envelope in regard to system performance and utilization without the need for added capital investment. Through this work, we believe we can achieve a higher-reaching goal of reducing the cost of processing prescriptions and healthcare expenses overall.

What Our Research Found

One of the largest cost areas within a high volume/central fill system is the cost of the Rx filling automation. We use in-depth tools for how to: determine how much automation is needed; synchronize the use of the various automation; and minimize the contention of inventory needed in various areas of a high volume pharmacy. These tools do a great job of balancing Rx workload based on a pre-defined data set. While the tools handle some level of variability, the ability to handle a wider range of tolerances would provide for even greater system throughput with the same set of automation. Although it’s easier to handle greater degrees of variance by adding additional automation, this drives up the overall cost, which negates our overall goal. Instead, we must recognize these correlations and variances, and dynamically adjust the system parameters using intelligent algorithms before the Rx order demand is presented to the system.

When we started on the thesis project, we wanted to analyze how frequent medications occur in combination across a very large data set. We believed that there is a large subset of medications that appear together, because it’s commonly understood that certain disease states typically require similar medication regimens, and even certain disease states would correlate with other related disease states. But the practice of applying this “understanding” had not yet been studied and applied in high-volume pharmacy fulfillment environments.

By analyzing the data, we would know for sure how frequent the correlations occur and the best algorithm for finding these correlations. We could then take that knowledge and apply the appropriate algorithms real time in the planogramming and scheduling algorithms available for live central fill systems.

The paper describes in detail the methods we used to determine the correlations. Because of the wide range of correlations and sheer number of combinations, the algorithm chosen is considerably complex. Run time for such a complex algorithm is important, especially when applying such an algorithm real time in a production environment. As a result, multiple algorithms were tested for accuracy and efficiently until it was decided that FP-growth algorithm would be used based on its ability to compress the data into usable formats. The FP-growth algorithm was more efficient than other tested ARM algorithms.

Benefits of Our Research

The results of this study show that there is indeed room for further optimization in both the positioning of medications within a high volume/central fill system and the scheduling of the medications for fulfillment. This allows process control systems, like PharmASSIST Symphony, to be further enhanced by artificial intelligence algorithms that reduce the total fulfillment time for a patient’s order, allow for wider variability within an order set, and optimize the use of inventory across the board.

We Can Help

Engage our Professional Services group today and learn how we can help optimize your pharmacy operation.

1 comment. Leave new

Great post!. Keep it up.

Leave a Reply

Your email address will not be published. Required fields are marked *