The Standards GapBy Marla Weigert | Posted 2008-10-30 Email Print
WEBINAR: On-demand webcast
Next-Generation Applications Require the Power and Performance of Next-Generation Workstations REGISTER >
Building an item master database that delivers standardization across disparate systems and data sources significantly reduced costs and improved services for Premier Purchasing Partners’ members.
The Standards Gap
In a typical week, we process 6.5 million item records from varied sources, including the unique systems of 2,000 hospitals, 1,200 suppliers, 50 distributors, 10 third-party data suppliers, and our e-commerce partners and data-cleansing companies. The product data, which represents many diverse product domains, includes nearly everything a hospital buys: from basic medical supplies, bed linens, cleaning supplies and cafeteria items to high-end surgical instruments.
But the complexity doesn’t end there. There are no standard product, supplier or organization ID numbers. Every data source (supplier, manufacturer and distributor) may apply different ID numbers to the same item. On the flip side, different manufacturers may use the same number to identify different products.
Even in cases where you expect standards to exist, they don’t. For example, our members purchase supplies from 3M. At one point, we identified more than 200 representations for this company in our members’ product records.
To complicate matters further, there aren’t any standards governing product descriptions or packaging, yet we must identify the smallest unit of use for our analytical computations. Is a bag of cotton balls a unit of one (because it is a single bag) or a unit of 200 (because there are 200 cotton balls in a bag)? Even this seemingly simple data conversion becomes a complex task when there are no standards.
Finally, our members’ information management systems are disconnected: Purchasing systems are not connected to billing systems or to charge systems, thereby compounding the data volume problems when it comes time to cleanse item master lists. By collecting data from each of these disparate systems and then standardizing the data to the Premier item master list, we can give our members the ability to perform the analytics necessary to identify cost-saving opportunities.
A Single Source of Truth
To overcome the challenge of not having universal product-data standards and to reduce avoidable processing costs, we developed a single source of truth: our Premier item master list. This “gold standard” drives efficiencies, eliminates errors and reduces the high costs in the health care supply chain.
When we began the initiative in 2001, we installed an off-the-shelf product information management (PIM) system and a data quality (DQ) tool to validate, cleanse and standardize our data. When this combination proved inadequate, we outsourced data cleansing to four DQ companies. This approach proved too expensive, so we outsourced to a single DQ company.
Our initial approaches met with some success, but they required more manual effort than we expected or wanted. Also, they were not scalable and could not keep up with our ever-increasing data volumes.
A core challenge is dealing with the infinite number of ways an item can be represented. To an expert, “sterile blue latex-free surgical glove” is easily equated to “Surg Glv, blu, ster, L-F,” but to automate this requires a huge effort. Traditional technologies typically fail to recognize changes in word order, punctuation or abbreviation, because they use a pattern-based, or syntactic, approach.
Earlier this year, we implemented a semantic-based approach, the DataLens System from Silver Creek Systems, to automate data standardization—our latest step in using leading-edge technologies to take our overall efficiency to the next level. By automating the recognition and standardization of even the most cryptic data, we can implement our own standards. And when industry standards are eventually established, the system will enable us to adopt them quickly.
In the meantime, we can scale up our operation while reducing costs and delivering improved services to our members. What’s more, because the new technology can “auto-learn” the rules for recognizing new data, we avoid major maintenance costs, such as having IT teams code and maintain hundreds of thousands of rules in several thousand product categories. In addition, these new capabilities will allow us to avoid making wholesale changes to our IT landscape in the future.
Seven years into our PIM journey, our vision remains the same, even though almost nothing from our original implementation is still in place, as we’ve upgraded all the components.