October 2007

Does your IT system need scrubbing to offset the fat-finger effect?

by Rick Dana Barlow

Likening data cleansing to personal hygiene can put the practice into crystal clear perspective. Missing or skipping a day, for example, may lead to disturbing consequences, albeit for different reasons.

One of the most obvious miscues about data cleansing is the notion that it’s a singular event that doesn’t require additional fixes and ongoing maintenance – either internally or by a third-party organization, such as a software- or online-based data cleansing firm, consulting firm, electronic commerce exchange or group purchasing organization.

Yet experts tell Healthcare Purchasing News that to achieve data standardization and synchronization industry wide, materials managers must be starting the process with clean and correct data, the result of regular maintenance and internal behavioral modification.

HPN tapped several professionals to spotlight strategies and tactics necessary to maintain data integrity. Robert Stowell, solutions architect, ByteManagers Inc., Chicago; Phil Meyer, R.N., MHA, president, ChoiceMED LLC, Watkinsville, GA; Emily Cikovsky, senior manager, content business solutions, and Karen Conway, director, industry communications, GHX LLC, Louisville, CO; and Sam Muppalla, COO, Portico Systems, Conshohocken, PA; shared their insights on data cleansing.

HPN: How do you convince healthcare facilities of the need to invest in data cleansing software and services on a continual basis and to modify employee behavior to prevent dirty data from infecting their system again?

STOWELL: We help healthcare facilities understand the impact of the warning signs to their business. It’s a simple matter to walk most facilities through benchmarking exercises in the following areas: Amount of product returns/incorrect orders; amount of redundant identical product (same manufacturer, model number, etc.) clogging up store-rooms and warehouses; amount of functionally redundant product (different model numbers, exact same item sold by different manufacturers) clogging up store-rooms and warehouses; and the amount of spend wasted through lack of consolidated purchasing strategy.

Most companies and facilities are aware that these problems exist, but few realize that over 75 percent of these issues are caused by bad data or bad data processes. Companies need to understand and benchmark these issues to define ways to improve. When we analyze the aggregate impact of these issues to business and customer relationships, investing in solutions is a no-brainer.

MEYER: The decision to invest in data cleansing services is usually tied to an organization’s decision to upgrade software such as a new ERP, MMIS or clinical system, and/or they are changing from a decentralized system to a centralized ‘enterprise’ system. However, many times a thorough data cleansing, standardization and synchronization effort based on a hospital’s existing MMIS will actually achieve the desired results of improved efficiencies and visibility into spend and utilization analysis.

Direct ROI tied to just data cleansing is hard to calculate, however it reaches, and is directly related to every facet of an organization’s operation and cost saving initiatives: Purchasing, Contract Portfolio Management, Inventory Management, Product Consolidation and Standardization, Electronic Supply-Chain, Invoice Reconciliation, Enterprise Spend Analysis and Business Intelligence.

The ChoiceMED solution is based on a Master Data Management (MDM) philosophy where after the initial data cleansing and standardization effort is completed, we maintain the data, enterprise-wide, on a perpetual basis. As new vendors and/or items are added they are automatically imported into the ‘production queue’ and then synchronized back into the client’s master data set. Therefore, wholesale process changes are not required to maintain the investment of the data cleansing project.

CIKOVSKY: Show them the money! It’s possible to show the direct relationship between dirty data and higher costs. In addition to increased discrepancies that waste staff time, dirty data can cost hospitals real money when, as an example, the wrong product is ordered and they have to pay rush freight charges to get the right product in time. Without accurate data, it is difficult for hospitals to ensure they are paying the right contract price. Invoice discrepancies also increase invoice cycle time, reducing early pay discount savings. Bottom line: Having the wrong product information – and worse, the wrong product – makes it harder to care for patients.

CONWAY: With more hospitals striving to lower supply chain management costs by automating more of their business processes, most material managers recognize the need to maintain data integrity on an ongoing basis. After all, if you want a touch-less or so-called perfect order, you can no longer count on humans to intervene and correct any data errors. What’s more difficult is changing human behavior. We’ve found that hospitals that are the most successful are those that put stringent policies in place regarding who can update the item master and in what format. For example, we know of one hospital that had multiple people updating its item master (containing 15,000 products) using a variety of sources for data. That hospital had a 90 percent error rate due to purchase order errors and invoice exceptions. On the other hand, hospitals that put procedures in place, such as using information from discrepancy reports to update their item masters on a regular basis, see a significant and sustained reduction in order errors.

MUPPALLA: Succinctly articulating the enterprise-wide impact of data quality is the key to convincing healthcare facilities of the need to invest in data cleansing. There are measurable costs associated with the inefficiencies and rework caused by poor data as well as significant impacts to relationships with key constituents, such as customers, members, suppliers, etc. Furthermore, there is a direct corollary between a company’s ability to scale and data quality. Automation of data integrity rules will minimize the need for extensive employee behavior modification.

What are some of the specific danger/warning signs to determine that you need to invest in data cleansing software and/or services?

STOWELL: Is your organization staffing around data problems? Does it require a high-level of domain expertise to negotiate your product or customer information? Does your data process live in somebody’s head instead of a well-documented data governance strategy?

If so, your organization is at risk. Not only will you fail to reap the benefits inherent to a successful data governance strategy, your purchasing processes will remain specific to people, hard to improve, and impossible to consolidate.

MEYER: Specific danger signs are somewhat ambiguous, but include: Unsatisfactory and erroneous reporting, increase in the purchase of incorrect products, poor utilization of purchasing contracts, increase in purchasing and invoicing errors with supply-chain partners, and increases in unnecessary shipping charges.

CIKOVSKY: Order error rates over 10 percent with any given trading partner, more than 20 percent non-stock or special orders, or contract price discrepancies greater than 10 percent, should all be cause for concern. It’s also a pretty clear sign when the supplier calls GHX or the provider directly to request the provider clean up their data.

CONWAY: Typically, we have seen the following differences between hospitals that do not cleanse their data regularly and those that do regular content cleansing, a best practice: 21 percent have duplicates in their item master vs. less than 1 percent of the item master; 2 percent with missing catalog numbers from vendors vs. less than ½ percent; 41 percent with missing catalog numbers from manufacturers vs. less than 5 percent; 14 percent missing unit of measure vs. less than 2 percent; 31 percent missing quantity of each vs. less than 5 percent; 47 percent missing price vs. less than 10 percent; 40 percent with non-file item orders vs. less than 20 percent, and 0 percent with UNSPSC codes on products vs. 97 percent to 100 percent.

MUPPALLA: Key warning signs include increased manual rework driven by errors, increased customer service costs, and increased cycle time for core processes. All these symptoms could point to poor data quality as a root cause.

What are some common sources of dirty data?
Is software the only solution?

STOWELL: The problem of dirty data comes mostly through lack of forethought, process, or understanding that data has context. Data that is ‘good’ in one environment may suffer in others. Software alone is not the answer. It requires a combination of people, process and technology to implement an effective data governance strategy.

People should encompass three levels of expertise:

1. Domain-specific expertise, understanding of what the products are and are for.

2. Data-specific expertise, understanding how to produce and maintain quality data.

3. Information technology expertise and understanding how data is effectively aggregated and deployed.

The process should be well documented and designed with these fundamentals in mind:

1. Build local processes that can expand to apply more globally. Too many organizations get paralyzed when faced with the enormity of everything at once.

2. Build simple, easy-to-follow processes free from complex technical jargon.

3. Solicit advice and opinions from the people experiencing the problems; the best process in the world means little if it can’t be implemented.

Technology should be assessed to augment both existing and future processes. Technology for data cleansing is not the same as technology for data maintenance. Understanding the differences can be critical.

MEYER: To answer this question fully, one must first define "dirty data." Dirty data can be defined as non-standardized data (essentially correct but not in a standard format), completely inaccurate data, or data that is missing critical buying-decision or reporting information. Dirty data will also almost certainly contain duplicate or obsolete records. ‘Clean data’ is data that is accurate, consistent, intuitive, standardized and contains all essential attributes for making informed buying decisions.

‘Dirty data’ can come from a GPO, a vendor and a manufacturer, and, of course, from clients’ own system.

Software is not truly a solution, let alone the only solution. Although much of the data cleansing process is automated, the underlying ‘clean data’ is the true key to the solution, and for data to be fully ‘clean’ from a ChoiceMED perspective, it is verified by a subject matter expert.

CIKOVSKY: Using information from outdated sources, such as old catalogs, and errors in data entry, due to rekeying problems or inconsistency in how data is entered. For example, a clinician might use slang when entering a product in an item master, because he or she knows what it means, but others searching for the same product may not be able to find it.

Software and data cleansing services can help, but what is most important is that hospitals develop a tight process up front that controls how data is entered and then create a sustainable, manageable way to maintain their item masters. Software plays a role, but a short-and-sweet approach to data entry is the real key to success.

CONWAY: The biggest challenge hospitals face trying to maintain data integrity is that data is constantly changing for a variety of reasons. Sometimes it’s the result of mergers and acquisitions on the supplier side or provider side. And there are always new products being introduced and others becoming obsolete. Inconsistencies in how data is entered in materials systems cause a lot of the problems; but other times, it’s simply a matter of mistakes being made when data is manually entered into systems, the so-called ‘fat finger effect.’ Software can really help in this case, but some providers are still challenged depending on what kind of information and in what format their materials systems can accept.

During times of wholesale changes, such as when one supplier acquires another company’s product line, GHX uses software to make the necessary changes to orders during the transaction process and then notifies providers of changes they need to make to their systems. They can do that manually, or they can use a service to provide them with the information to upload electronically into their systems (again, depending on system capabilities). In either case, there is a delay between the time data changes and when the data is changed in their system. That’s why we are developing new ‘black box’ technology that will enable hospitals to not only get the information almost immediately and in the format they need for various systems and purposes.

MUPPALLA: Common sources of poor data are inaccurate data entry, multiple data repositories for the same data and external data coming into your organization that are not subject to the same quality filters as your internal data. While software is part of the solution, data cleansing initiatives need to also include continuous process improvements, adherence to data quality standards, and potentially leveraging ‘gold standard’ external master files.

What if the dirty data came from a vendor or GPO and not from employee data input errors? How do you convince them to work with you and provide you with the right data? What is the vendor’s and GPO’s responsibility in making sure your MMIS/ERP system is updated regularly so that you’re always paying the correct price?

STOWELL: These collaborative processes are critical to the sustained maintenance of data between organizations and vendors or GPOs. Companies are achieving success at building contractual incentives into contracts with vendors and GPOs, such as discounts associated to reduction in returns associated to incorrect content or larger penalties associated to data-related returns.

MEYER: ChoiceMED works directly with vendors, GPOs and manufacturers in collaboration to cleanse and standardize our mutual customer’s data. We believe that it is the responsibility of an organization’s vendor and GPO partners to maintain accurate item pricing within their purchasing system, and our advanced data synchronization software facilitates this process.

CIKOVSKY: Dirty data comes from GPOs and vendors, but each has vested interests in making sure the information is correct. Again, show them the money – show them the error rates, indicate their costs, talk about the number of calls the providers are making into the supplier customer service centers, or to their GPO reps – that’s compelling.

CONWAY: No one in the healthcare supply chain is immune from the problems of dirty data. GPOs and vendors struggle, too, with matching specific buying organizations to specific contract pricing. That’s why unique standard identifiers for products and organizations, down to the ship-to level, are so important. The industry also needs a collaborative tool that enables all the authorized parties to specific contracts share information related to contract eligibility and commitments on line and in real time, to avoid the errors caused by delays or errors in communicating information. GHX and its members are working hard on both standards and a contract commitment tool that the entire industry can use.

MUPPALLA: A starting point for ensuring data integrity from vendor sources is to publish a standard for data quality. In conjunction with data quality standards, an organization should develop an intelligent data import capability that includes your data validation and rules. Reporting capabilities are also required to facilitate reporting vendor compliance. Some companies also choose to build vendor compliance into service contracts. Like data cleansing, vendor compliance is a journey.

What’s the biggest misunderstanding supply chain managers have about data cleansing software and services? How should it be corrected?

STOWELL: There’s a huge difference between effective data cleansing and effective data maintenance. The tools and processes may share some elements, but the differences need to be understood before an organization can effectively build solutions.

For example, there is often a large cost associated to the cleansing of large amounts of legacy information. Years of bad practices often take work to analyze, strategize and solve. Data cleansing software can help with this, but it will not solve the problem. Many healthcare providers are able to successfully outsource the ‘heavy lifting’ associated to large sets of bad data and implement tools and processes designed to prevent the issues from recurring instead. This is a much more efficient way for organizations to spend their resources.

Also, many of the most effective industry solutions are more global than just supply chain. [Master Data Management] and [Product Information Management] tools are providing effective ways to get ‘all that and more’ from data initiatives.

MEYER: Most supply-chain managers think the data cleansing process is very time consuming and intrusive to their daily operations. The ChoiceMED process is very easy to implement, and because of our unique hands-on experience, we know how to work with these busy professionals to make the project a success.

CIKOVSKY: The biggest misunderstanding is that software and/or services will take care of the data problems in their item masters and they won’t have to do any work to ensure the information is correct. Unless a provider is willing to outsource all of their item file maintenance, and most are not, they will have to devote some resources to validating the information they receive, either from a software package or service. And that takes time.

The other big misunderstanding is that one size fits all; in other words, that data that works for materials systems and functions will work for finance or for clinicians. Hospitals need to effectively maintain the data in all of these areas and make sure it is accurate – and that depends on how it is used, by what system and for what purpose. In other words, accuracy is in the eye of the beholder.

To correct these problems, we recommend that hospitals form cross functional teams to understand their overall data management needs and how they can best work together to meet their combined needs. We would also support creation of an industry data management user group, perhaps under the umbrella of one of the industry associations, to help provide hospitals with objective education on the issues they face and the options available to them.

MUPPALLA: Underestimation of the impact of poor data and the frequency of data cleansing are the biggest misunderstandings. To correct the problem, it is critical to infuse quality at the source of the data, to monitor the quality at the point of data intake. Typically this is done through leveraging work flow driven applications that implement a standard set of data integrity business rules. You also need intelligent intake as well as a flexible distribution mechanism to make quality data accessible to the people who need it.

For more information on these companies visit their respective Web sites at www.bytemanagers.com, www.choicemeddata.com, www.ghx.com and www.porticosys.com.