I had the privilege to be invited to testify to the Health I.T. Policy Committee workgroup on the topic of data quality back in November. I’ve been an advocate for the work of the committee for years and am constantly impressed with the considerable insight and genuine passion they bring to their work. The opportunity to testify, however, was my first opportunity to actually participate in the policy-making process and it certainly was both a learning opportunity for me, as well as a chance to share my thoughts on the important topic of data quality.
Over the last 18 years as a healthcare IT leader and product strategist, I have experienced first-hand the data quality tradeoffs that come with balancing individual clinician freedom versus standardization when implementing new systems; the risks inherent in migrating data from one clinical system to another created by data quality problems; and worked against the huge barrier data quality issues can create in establishing interoperability between healthcare providers in a region.
My perspective on the types of solutions and best practices available to address healthcare data quality issues was significantly expanded when I joined Informatica. Although healthcare may be a bit late getting into data analytics and business intelligence, the advantage is we have the opportunity to learn from the experience of other industries.
The testimony I presented to the working group follows:
My name is Richard Cramer, and I am the Chief Healthcare Strategist for Informatica, a data integration software vendor. Thank you to the members of this working group, the Health I.T. Policy Committee and the Department of Health and Human Services for the opportunity to testify today on the important topic of EHR-Generated Data Quality Assurance and Management.
The first question posed to the panel asks what metrics and standards for data quality are currently in use for capture, transfer, storage and analysis for providers and data intermediaries that manage clinical information?
I will address this question from the viewpoint of the healthcare enterprise, rather than the design features of an individual EHR. From this perspective there are few common metrics and standards for data quality beyond attempts to standardize vocabularies and code sets.
I encounter almost universal acknowledgement among clinicians and information technologists that they have very real and pervasive clinical data quality issues. However, it is the rare exception that any of these organizations can quantify the magnitude and extent of the data quality issues in objective terms.
Tools and best practices exist that can aide in identifying and resolving data quality issues with clinical data. However, these have not yet seen widespread adoption beyond the traditional list of healthcare innovators and early adopters. Equally important, data quality must be viewed as a program, not a technology problem or a project. Done right, a data quality program doesn’t end but gathers momentum. New data is always being created; new data sources are always being added. Having the governance, stewardship, and tools that enable a continuous process of profiling data, monitoring quality and resolving discrepancies is an essential component of addressing the current and future challenge of ensuring the quality of clinical data as a key asset.
The second question posed to the panel is what are EHR database managers doing to ensure information in clinical DBs are trustworthy, complete and accurate?
Although database managers are working hard and doing the best they can, they are all-too-frequently constrained by limited access to their own data within vendor applications. Even for healthcare organizations that have invested in data quality tools and solutions that can rapidly identify and resolve data quality issues, they are limited in their ability to use these solutions with the data in their EHRs due to proprietary data structures or contractual terms that limit access.
Consequently, the first real visibility that organizations have into their data is often after it has been exported from the production application into some form of external store where it can be accessed. These external databases are seldom representative of the production application, and significant data quality artifacts can be induced by the proprietary nature of the data extract routines. In other situations, the best organizations can do is capture clinical data from HL7 messages being sent out from the EHR and persist the data themselves in a database of their own design.
If we consider that a key aspect of data being trustworthy is being able to have complete visibility into the lineage of the data – from source to target and all transformations, mappings and other alterations to the data as it is processed from system to system – we face a huge barrier to having trustworthy data on the very first step in getting the data out of the EHR.
The third and final question posed to the panel asks if data quality is important for an attestation program like the EHR Incentive Program?
I have seen meaningful change in how healthcare organizations behave with regards to data as a result of meaningful use, and it has been almost universally positive. Organizations for the first time have an inventory of data, they know where to get it, when it’s missing, what the allowed values are. It is a refreshing focus on getting the data right the first time that will become the raw material for future analysis. However, there is a fundamental truism that you only get high quality data when you use it. Therefore, the most productive focus is likely to be on meaningful clinical measures that drive demand for quality data, rather than direct metrics on data quality itself.
In addition to addressing the issue of data quality, I wanted to focus my comments on the challenges of simply getting access to data that is locked away in proprietary applications. And while my perspective is certainly influenced by being the healthcare strategist for a data integration software vendor (where job #1 is helping customers get data out of the applications and make it useful for analytics), the basis of my opinion was my own frustrations when I worked for health systems in the extreme difficulty in trying get at data in vendor applications. I hear this same complaint from virtually every one of our customers, so commenting on the need to make access to data easier under the umbrella of data quality seemed like a good idea.
Little did I know that far from being controversial, my comments were going to mirror most of the other testimony. The topic of how difficult it is to access data in electronic health records for the purpose of analytics and other reporting uses was a central theme of almost all the panelists. It was clear from the testimony of the panelists, the questions from the working group, and the public discussion that finding a way to incent improved access and use of EHR data was definitely an area of focus.
So I left the hearing as encouraged as ever that the Health I.T. Policy Committee and the working group are addressing the topics that genuinely need to be addressed; are seeking input from a broad audience of vendors, providers, patients and others; and are setting at least some of their sights on the right target – namely helping to gain access to and unlock the value of all that great clinical data that is being captured by providers using EHRs in a meaningful way.