The Problem
Using data analysis techniques, disparate and dirty data from multiple hospital sites across Australia were cleaned and merged to prepare a profile of the products purchased and used across the group, so as to inform the creation of a ranked list of products of significance in terms of volume of use and expenditure.
Pharmaceutical suppliers were invited to offer pricing for the listed items utilising a custom eSourcing portal. A non-linear optimization process was then performed over the resulting data, and the optimum purchasing strategy identified, resulting in significant cost savings for the group.
Hospitals use many different software systems to support, manage and audit the procurement, pricing, distribution and usage of medicines. Many of the processes are human-driven and rely on hard copy documentation and manual data entry. As a consequence the data within the systems is inaccurate and rarely contains enough information alone to identify overall cost.
Additionally, many products have multiple brand names despite having a common, generic name, and there are inconsistencies between nomenclature describing presentation (e.g. form, pack size) and strength. Finally, there is no standard interface for exporting the data.
The Solution
Vokke worked with PharmConsult to develop a canonical data format that’s common between all the systems present in the hospitals and wrote software to convert each of the data sets into this common format, automatically correcting and cleaning the data in the process.
By the end of this step, hundreds of thousands of entries from sites across Australia were combined into a single, high quality data set ready for analysis. The analysis produced a condensed list of medicines used by the hospital group to inform a request for pricing (RFP) from pharmaceutical suppliers.
Requesting pricing proposals from pharmaceutical suppliers located around Australia on hundreds of products is a complex process. Suppliers will often propose bundled or bulk offers so that specific products are cheaper when purchased alongside others or in specific quantities. Aggregating pricing proposals into a single, definitive data source is challenging due to the presence of multiple variables for each product and non-linear pricing models, such that ‘comparing apples with apples’ becomes a costly and error-prone process when performed manually.
To assist with the pricing offer collection process, Vokke worked with PharmConsult to develop a secure online portal that allowed pharmaceutical suppliers to submit individualised pricing proposals. Capturing data via a structured, online portal dramatically reduced collection errors and defects within the data, and significantly streamlined the development of a unified data source for further analysis.
Even a small range of products with offers from multiple suppliers proposing different purchasing plans, such as bulk discounts and bundled offers can lead to a massive explosion in possible purchasing outcomes. In order to determine which offers, from each supplier, deliver the best value, all outcomes need to be explored.
When you have hundreds of products, the number of offer combinations can easily exceed that which can be reasonably or accurately evaluated manually Vokke developed a custom nonlinear optimization algorithm that leveraged the cloud to rapidly compute the most optimum purchasing strategy, optimizing for cost savings and preferences among suppliers.
The results of the analysis, with all offer combinations for each product ranked in order of cost benefit were presented to the hospital group for consideration.
