Physician Data Analytics – Impact of Data Quality – Part 2
By Richard Howe, PhD, Executive Director, North Texas Regional Extension Center
Last month, I briefly touched on some quality issues related to use of an electronic health record (EHR). Over the past year, I have participated in a work group, called the ONC-HIMSS Patient Matching Community of Practice, co-led by the Office of the National Coordinator for Health IT (ONC) and HIMSS, that focused on various patient identification and matching issues.
Some of these patient matching issues include:
• Multiplicity of different patient matching approaches/algorithms used by different vendors and Health Information Exchanges (HIEs). This creates a lack of uniformity in patient matching methods across the industry.
• Insufficient evaluation of the accuracy of these diverse methods, especially in real-world use. Hence, there is no gold standard for patient matching.
• High rates of unmatched and mismatched records exist, due in part to inadequate data quality at the source. This results in duplicate records within a system.
• Lack of national data standards for validation of data quality for patient matching.
The patient matching work group also noted that major improvements in patient matching rates have been achieved through the introduction of the following:
• Increase number of data attributes used in linking
• Use of nationally recognized data standards
• Decrease free text data entry
• Standardize naming conventions, such as use of legal name only
• Use of secondary data, such as use of the US Postal Service street names and addresses as a potential national standard. GPS systems in cars use the US Postal Service street names in their database.
• Analysis of the impact of the addition of specific data elements on the improvement of matching rates (for example, the addition of “mother’s maiden name” was found to dramatically reduce duplication rates)
• Use of auto-linking algorithms to improve match rates in very large retrospective data bases in which extensive manual correction is not feasible
HIMSS has recently developed three requests of Congress for the coming year. One of these requests is to remove the restriction on use of Federal Funds to develop a national patient identifier. Specifically, HIMSS asks the following:
• Support robust interoperability and health information exchange.
o One sub-theme is removing the “Congressional prohibition” on the use of government money to create a national patient identifier. In reality, if a national patient identifier had been developed years ago, many of the patient matching issues that exist today would have been significantly reduced.
o The second directs ONC to review and amend its EHR certification program to include “rigorous interoperability testing” of HIE standards and specifications.
In summary, data quality issues that impact either duplication of medical records or incorrect clinical data must be addressed to improve overall use of an EHR in a physician practice. Improving data quality will more effectively support daily practice operations and long-term data analytics.