Routine database maintenance can lead to hospital treasure
by Rick Dana Barlow

Phil Pettigrew, director of materials management at Denver Health Medical Center, understands completely the power and influence that accurate data can grant someone of his caliber when he or she faces off with clinicians over a bottom-line issue. "A materials manager armed with the right data and who understands what that function means and how it applies to your facility is just as good as one doctor talking with another doctor," he said.

While that statement may not be rocket science to a seasoned materials manager like Pettigrew, the concept of maintaining accurate data has become needlessly complex, fraught with a host of misconceptions and unrealistic, if not unreasonable, expectations.

Studies show that poor data quality costs 1 percent to 4 percent of the average hospital’s expense stream, according to Carl Natenstedt, managing partner, Perigon LLC, which offers a data cleansing and maintenance tool, as well as an information portal for identifying savings opportunities based on spending history. That’s $4 million to $16 million per year for a hospital with a $400 million spend, he noted. The financial consequences can be huge.

Poor or missing contract information means a hospital may be paying a higher price for contract items, he contended. Duplicate items in the master item file may indicate a facility is purchasing the same product at multiple prices. Limited categorization schemes demonstrate an inability to standardize clinician preference items, and incomplete manufacturer data can lead to a limited ability to identify and negotiate new contracts.

Even though the problem of dirty data can be convoluted and messy, the solution to it may not be as simple as one would hope. "The most critical misconception is that data cleansing is a one-time fix and that software alone can solve the problem," said MJ Wylie, director of content services at Global Healthcare Exchange. "To address the problems caused by inaccurate product and pricing data, hospitals need to employ process improvements and systems to continually maintain the accuracy of data by synchronizing it with data used by their vendors and group purchasing organizations. The key is to build quality into the process, rather than try to fix problems after the fact.

"Unfortunately, the data cleansing processes currently used by many hospitals and healthcare organizations simply take purchasing history information and then compare it to what the data cleansing service provider says is the most accurate data and cleanse it once," Wylie continued. "The value of this process is limited by the accuracy of the data from the data cleansing service. More importantly, given the number of new products continually introduced and retired from the market, as well as considerable merger and acquisition activity on both the supplier and provider side, even the most accurate data will change over time – sometimes overnight."

Certainly education and training can help change data entry behavior but it’s not enough to solve the problem alone. "The way to achieve ongoing data cleansing, accuracy and synchronization requires a combination of easy access to the most accurate data, systems to automate the process of data cleansing and correction, and changing business processes, i.e., how people use the data and systems," she said.

Wylie noted that other industries, such as grocery, automotive and electronic, solved inaccurate data problems by creating a central repository with manufacturer-verified data that they can use to synchronize their data to that of suppliers. That’s why GHX, spearheaded by Wylie, developed its AllSource Content Repository, which contains data on more than 2 million medical/surgical products.

"GHX helps hospitals ‘get ready’ to use e-commerce by giving them tools to compare the data in their item masters to that in the AllSource repository and then cleanse and correct their databases with this information," she said. "During the ecommerce readiness process, we find an average 35 percent inconsistency between hospital data and the AllSource Content Repository." Meanwhile, Wylie also is helping build Content Intelligence, a patent-pending product that actually facilitates ongoing item master cleansing.

"Once the initial cleansing and correction is performed, GHX Content Intelligence helps maintain data accuracy by utilizing the AllSource data and business rules established between buyers and sellers to identify inaccuracies in purchase orders and correct them during the transaction process, and then notify buyers of updates that need to be made to their item masters on an ongoing basis," she continued. "A new tool called Source Update also enables hospitals to automate the process of making necessary changes to their item master. Source Update can also be used to more easily maintain the most accurate contract pricing in their systems, whether that contract price is direct from a manufacturer or distributor or from a GPO."

Dirty data danger signs
How do you recognize whether you’re working with data that needs cleansing? Priya Kamani, M.D., vice president, data management solutions at Neoforma, cites a number of potential indicators that should be compared to best practices.

First, the facility has a high number of total records in the item master and a high number of vendors represented in the item master. Best practice is 15,000 to 20,000, and less than 750, respectively, she said.

Second, the facility may struggle with data completeness. All items should include vendor name and catalog number, manufacturer name and number, product description and unit of measure, she noted. For best practice less than 1 percent of the item file should have this information missing, according to Kamani.

Third, the file should include categories associated for effective analysis and ease of searching. Problem indicators involve large numbers of items under the "no categorization," "miscellaneous" or "other" categories.

Fourth, the file may lack a distributor-manufacturer link, which means a large number of distributed products don’t have a manufacturer name and number associated with them, she said.

Natenstedt identified four major warning signs to pinpoint. They are:

• More than 1 percent of purchase orders have a price discrepancy

• More than 10 percent of PO line items are not sourced from the item file

• Product descriptions are not in a consistent format

• Item master does not contain contract numbers

Dirty data sources
Much like bodily infections can be traced back to human transmission through some bodily function or use of contaminated instruments, dirty data can be traced to human error.

Emily Cikovsky, senior product manager of content at Global Healthcare Exchange, attributes dirty data infection to a number of human-derived sources. They include manual data entry errors or mistakes in rekeying; use of inaccurate data sources, such as outdated paper catalogs from both distributors and manufacturers instead of manufacturer-verified and maintained data; multiple individuals updating item masters; and a lack of standardized processes for updating data, such as the use of abbreviations or naming conventions for product names and units of measure.

"There is no software that cleanses data by itself," Cikovsky said. "Software can facilitate the process but must be accompanied by improving business processes to cleanse and correct product data by comparing it against an accurate source of truth."

Hospitals that allow multiple employees to enter data or don’t have standardized processes for item master maintenance run the risk of generating the greatest data inaccuracies, according to Wylie. They may have multiple but inconsistent entries for the same product; missing data or insufficient descriptions, making it harder to identify the right product to purchase and ensure contract compliance; missing information, such as unit of measure, which can cause orders to fail in supplier systems; incorrect vendor or manufacturer information; and inconsistent data, such as using "cs" and "ca" to describe a "case" can result in order exceptions or shipments of the wrong quantity, Wylie noted.

Kamani agreed. "[Dirty data] usually stems from poor maintenance habits and inadequate access controls for editing and updating," she said, adding that access controls should be restricted. "Software or services can be used but the key is the ongoing maintenance of the data after the initial cleanse process to prevent degradation of the data."

Wylie also linked how well a hospital maintains its item master with the number of order exceptions. "For example, a hospital that has multiple people updating its item master (15,000 products) using a variety of sources for data, e.g., vendor catalogs, sales reps, etc., reports a 90 percent error rate due to inaccurate data in the POs and POA reconciliation process," she said. "An eight-hospital IDN with only one person authorized to maintain data for a system-wide item master with approximately 51,000 items reports only minimal errors.

"A high percentage of order and invoice discrepancies can indicate the extent of a data inaccuracy issue," she continued. "Inaccurate data can also skew tracking total costs for product categories and total spend, contract pricing and contract tier performance. For example, something as simple as inconsistent and therefore duplicate entries for the same product make it next to impossible for a hospital to determine exactly how much of a particular product it is buying. This becomes increasingly important as hospitals attempt to perform more analysis of their total spend. Without ensuring data accuracy from the start, the validity of spend analytics will be compromised."

In addition, older systems typically have poor search capabilities, Natenstedt added. And time plays a big role, too, because the longer you wait to correct data discrepancies the larger a problem it becomes.

Mile-high data mastery
Pettigrew’s facility, the 373-licensed bed Denver Health Medical Center, embarked on a data-cleansing project last year that ran for 16 weeks. The hospital upgraded its Lawson system in order to integrate product data between various internal computer systems, implement distributed electronic requisitioning, expand its online electronic commerce capabilities and automate each step of the supply chain process from "req to check," as Pettigrew put it. "These systems are only as good as the data that are in them," he said, describing the project at a recent World Research Group conference in Chicago.

Pettigrew wanted to be able to track spending by product category, manufacturer level and contract to identify price parity and contract variance. Denver Health worked with Neoforma to build product descriptions, inactivate unused products, recruit suppliers to cooperate and establish a project dashboard.

Because Denver Health had made certain procedural and structural investments in item file management they had an item file with fewer than 15,000 products, no missing supplier or part number information, no product duplicates and 43 percent of active products carried contract identifiers.

However, they did find that less than 50 percent of active items had been purchased in the previous 18 months so about one-third of the item file was put on inactive status. Numerous reporting inaccuracies were linked to a limited knowledge of units-of-measure conversion factors and non-uniform product descriptions and supplier naming conventions further complicated records.

Their efforts identified nearly $76.4 million in purchases and uncovered between $300,000 and $500,000 in potential pricing overpayments during a 12-month analysis. They also identified nearly $74,000 in overpayments on GPO contract pricing. From this they were able to tighten audit controls, attack non-item file purchases (maverick or off-contract spending) and improve contract monitoring and tracking.

"Once we cleaned this up we wanted to stay clean so we established ongoing maintenance procedures," Pettigrew said. "For the first time I have an MMIS with usable data. Now we can slice and dice data any way we want it. It’s amazing that we were able to take a 15,000 item file database and clean it up in 16 weeks."

Making haste with data maintenance
Not so surprisingly, investing in a new materials management information system (MMIS) or enterprise resource planning (ERP) system won’t improve cost-reduction capabilities because of the garbage in-garbage out adage. "Many ERP/MMIS installs merely replicate old data from the previous system into the new ERP/MMIS solution," Natenstedt said. "This is tantamount to buying a new Porsche and draining the oil from your ’65 Chevy and dumping it into the Porsche."

Still, Cikovsky contends that hospitals must have an accurate data source, "which we believe must be the manufacturers, and then systems that facilitate automatic uploads to item masters and/or MMIS/ERP systems."

Certainly, manufacturers have the responsibility to communicate accurate product information but it’s up to the GPOs to make sure hospitals receive accurate pricing information "in an easy and timely manner," Cikovsky said. "Some GPOs provide additional services to help automate that process, oftentimes in concert with outside service providers."

Kamani doesn’t recommend software and system investments absent a data quality improvement process in place because without it a hospital minimizes its automation and return-on-investment potential. Poor data quality forces manual workarounds, exception processing and rework, she indicated. Any reports gleaned from the MMIS/ERP systems are difficult to generate and of questionable quality. Without a clean item master, she added, hospitals fail to reap the full benefits of doing business electronically, online and offline, and of implementing distributed requisitioning. "Software can be quite expensive and is impractical for small or medium-sized organizations," she said. "Most GPOs outsource this service anyway so it’s better to go to the source."

Building systems and standardized processes to cleanse and correct data as well as maintain accuracy isn’t something a hospital can accomplish on its own, according to Wylie. "This is a very time-consuming process if handled manually and internally, and is nearly impossible without access to data that is consistently maintained by the owners of the data, the manufacturers," she said. "Oftentimes, hospitals don’t know how to fully measure the costs associated with having inaccurate data since it can impact a variety of processes."

Cikovsky, too, believes that healthcare providers, particularly hospitals, find it very difficult to handle data cleansing on their own. "Hospital purchasing and [accounts payable], as well as group purchasing organizations, are experts in sourcing and contracting, not data management nor systems design specialists," she said. "Qualified third-party content cleansing and data management providers can provide the expertise and resources as a result of the work they have done with manufacturers and their hospital customers. A third party also has a database of information available to facilitate a much faster cleansing process. Without such expertise, it’s like knowing there is a problem with your car, but not knowing exactly what the problem is and therefore not knowing exactly who can or how to fix it."

Predicting how long a data-cleansing project will last depends on a variety of factors, chief among them the size of a hospital’s item master. Wylie estimated that, generally speaking, it could take a small hospital with a single item master and dedicated internal resources several months to a year to fully cleanse its data. "And again, without systems in place to maintain data accuracy, the process could be unending," she noted. "Data cleansing experts with access to accurate data and systems to automate the process could complete the process in a matter of weeks and implement processes and systems to maintain that accuracy and avoid the need for another complete data cleansing, barring changes, such as the addition of a new hospital or item master."

Using Neoforma’s Data Management Solution projects as a barometer, Kamani estimated that a 12- to 16-week timeframe to cleanse data is typical.

Still, Kamani laments one of the biggest misunderstandings materials managers have about data cleansing software and services that can only be corrected through educational efforts. "They don’t appreciate the level of effort it takes to get the data cleansed," she said, "and so they expect to pay very little for these services."HPN