June, 2006

Ensuring Data Quality
To clean your data, use your data – and know what you’re looking for

When it comes to collecting and using data, public health agencies have become increasingly aware and concerned about its quality. Agencies need quality checks when data enters an information system and when it is used – whether for individual service, aggregate reports, or program integration. As programs grow and integrate their data with other programs, accurate data becomes exponentially more important.

So, what is data quality and how is it maintained? In essence, data quality means that the information entered and stored in a system is accurate, both in substance and in its location. This is especially critical when data is merged from separate systems because data in one field may have to match its exact counterpart in another system.

Data can be entered into a system by data entry (individual staff) or through electronic transfer such as matching and merging records. In either case, names should be spelled correctly and dates recorded accurately and in the proper field. Clean data also relies on efforts to create or maintain only one file for each individual, the “deduplication” challenge cited by so many agencies. To provide concepts and tools for deduplication, Connections has published The Unique Records Portfolio.

To address data quality proactively, most public health programs have some quality checks in place in the form of data entry rules and exception reports that flag unusual trends in the data. For example, a report might uncover that an immunization date occurred prior to a child’s birth date. In that case, either date could be inaccurate and must be corrected.

“We run a number of queries to generate our reports,” says Magaly Angeloni, KIDSNET operations manager. “For example, we look for prenatal care that has been posted after the child’s birth date – or a vaccine given before a birth date.”

Other sample queries include:

• checking for newborns still listed with the first name Baby, a holdover from prenatal care.
• checking for kids shown as “active” in WIC after the cutoff age of five.
• children in the system with a birth date prior to 1997, the earliest year in the system at this point.

To improve data quality across its programs, Rhode Island is setting up an organizational structure within the KIDSNET information system and broadly throughout the health department to make sure that business rules and protocols are in place to ensure data quality.

Business rules should be established up front for the operation of any information system, notes Amore. That is, program staff should determine and agree on what information is needed, how it will be used, and what the system requirements are to meet specific data use and data quality goals. “We need business rules that ensure that each field has valid values defined,” she says. “Does the data make sense? Program staff come to realize that accurate data doesn’t magically happen just because it’s in an information system.”

“It’s also about accountability,” says Samara Viner-Brown, Division of Family Health’s Chief of Data and Evaluation. Viner-Brown coordinates the Family Health Data Quality Team that brings together representatives from Division of Family Health programs who are responsible for or involved with databases. The team includes program managers, data managers, an epidemiologist, and data entry staff. The team is drafting policy language for the Division of Family Health to address data verification, its review, and record resolution (because finding bad data solves only half the problem).

“If you don’t look for problems, you won’t find them,” adds Kim Salisbury-Keith, KIDSNET development manager. “We have to continually look for what’s wrong, not just run edit checks. You have to troll your database, run queries looking for things that don’t make sense, unexpected changes that might be rare. For example, we might find a baby three years past the due date who is not in the system yet. You can’t have a child in utero for three years!”

When programs share data, the business rules become critical. “Once WIC changed their code set on languages, and KIDSNET found a bunch of Cambodian children listed as Italian speakers,” Salisbury-Keith recalls.

“Everyone knows that data will never be perfect,” states KIDSNET Operations Manager Angeloni. “But to improve your data, you have to clearly understand the issues.”

To improve data quality, keep these ideas from Rhode Island in mind:

• Understand how the data are going to be used.
• Agree on common definitions of terms.
• Emphasize documentation that defines the business rules.
• Train data entry staff and data managers in how forms and data fields are interpreted, business rules and data quality protocols, and how to correct or revise data and data fields.
• Learn the information system’s subtleties, such as the difference between a “null” field, a field left blank, and a field entered without data.
• Enable the system to flag errors at the time of data entry.
• Empower all staff to identify any error found in any report or system.
• Make sure that program staff determine the parameters of a report so that everyone has clear expectations for how the data will be used.
• Never completely trust the data! Remain vigilant for anything that seems unusual or out of place.

KIDSNET Development Manager Salisbury-Keith echoes a common refrain when she says: “Using your data is the best way to clean your data.”

For more information
Ellen Amore
E-mail: Ellen.Amore@health.ri.gov
Phone: (401) 222-4601

©2005 Public Health Informatics Institute
All Rights Reserved
750 Commerce Drive, Suite 400 • Decatur, Georgia 30030
TEL: 1.866.815.9704 • FAX: 1.800.765.7520

Last updated June 21, 2006 June 21, 2006