OPTIMIZING REPACK SIZES FOR OUTPATIENT PHARMACY AUTOMATED DISPENSING SYSTEMS

Monday, January 6, 2014
Nassim (The Regent Hotel)
Poster Board # P1-14

Hong Yee Lim1, Eric Yng Xixin Yang1, Hui Hui Wang2, Wee Chuan Hing3, Kiok-Liang Teow, MSc2, Zhecheng Zhu2 and Angeline Chiam3, (1)Tan Tock Seng Hospital, Singapore, Singapore, (2)National Healthcare Group, Singapore, Singapore, (3)National University Hospital, Singapore, Singapore
Purpose:

Three public healthcare institutions’ outpatient pharmacies are deploying automated drug dispensing systems for better safety and improved efficiency. For the machines to pick the drug items, they have to be in boxes of pre-determined quantities, rather than in loose tablets. The purpose of the study is to design and implement a method that will optimise the re-pack configurations for greatest efficiency and lowest cost.

Method:

This is the first time that such machines are to be deployed where we need to determine the repacked box sizes to meet each patient’s prescription order. A team of pharmacists and operations researchers came together to study the trade-offs between the multiple objectives, such as extent of complete automation and number of boxes, Historical prescription patterns were analysed to and formed the basis of the quantities needed by the patients. Mathematical programming (mixed integer programming) was used as the framework to model and solve the problem at drug item level. The key objectives were maximising the extent of automation, i.e., prescriptions orders completely fulfilled by automation machines, and minimising the number of boxes used. The implicit constraint was to fulfil the quantity needed and within a given number of box configurations. This model was repeatedly used to find the optimal repacked sizes for each of the hundreds of drug items, each based on its prescription patterns and characteristics.

Result:

In total, there were about 800 drug items being studied equivalent to more than 5 million orders annually. A few options per drug item were given further deliberation. Of these items studied, we could achieve overall 95% automation and 2 boxes per line item. These results became the basis for inventory planning and workflow design.

Conclusion:

The problem was combinatorial in nature and no simple “greedy” algorithm would guarantee an optimal solution. The mathematical programming framework provided a scientific method to breakdown the problem and ensured the solution search would be efficient and complete. The results were customised to individual drug item’s usage. In view of growing health care demand, this model helped us to tap the full potential of the automated dispensing machines. Also, the approach allowed the flexibility for one to fine tune the model if there were changes in constraints and optimisation criteria.