Reporting requirements will undoubtedly increase with Meaningful Use. Replacing paper potentially faster than any other point in a health system will change the tenants of the traditional health establishment because structured information does a far better job at enforcing quality of care and improving population health. Health Information Technology tools like Decision Support, Clinical Repositories and databases are many times still tucked behind sophisticated Storage Area Networks, Networks, and Interface engines. A percentage of the stakeholders requiring meaningful data may really not know how to request it in an HC Enterprise. They know what they would do, left alone, with an IT Analyst to quantify it on behalf of meaningful use. As we debate our philosophies on the possibilities, we jeopardize care for the patient, and raise questions on care quality because of the applications, the lack of applied standards, and interoperability challenges.
Now we have registries, mostly undervalued and misunderstood and loosely interpreted. Today, these registries (such as disease registries) are really databases that collect clinical data on patients with a specific disease (diabetes, asthma, CHF, hypertension, etc); however is that not what the traditional decision support or a clinical repository is supposed to be for the enterprise? Lots of confusion and I would like to hear your thoughts?
Mark’s thoughts bring out another complexity for the HITECH interim final rule that all will need to work through – that concern of the component elements that must cooperate to meet all the criteria to attest to MU. Registries and reporting will become a) integral new elements for EHRs/EMRs; b) continued component elements that require continual cooperation between the EHR/EMR vendor and the registry/clearinghouse; and/or c) new market opportunities for competition amongst the noted players – the latter with the opportunity and confusion that brings to providers who need to make some expedient but lasting decisions. Our early understandings from reliability theory and cost accounting help us to appreciate that today’s way of doing this in many cases leaves one more serial component open to failure while it adds cost to the solution more – so this is an economic cleanup area. But that performance and financial opportunity must follow each healthsystem, hospital, or practice laying out its own roadmap for MU and understanding how these component elements will reshape over the next year. Mark reminds us to be planful.
Gerry Higgins says
I would to add some comments on the topics you have raised, with relevance to the EHR domain:
(1) In general, I agree with my colleague Mark Smith (at MedStar), that all data are ‘dirty’, and that needs to be represented in the EHR. Thus, XML and its derivatives tend to be the best way to represent data in the EHR.
(2) For content that needs to be structured, such as Clinical Decision Support and Disease Registries, there seems to be several choices. If we follow the path of the academic IT researcher, we need to first define an array of nomenclature and classification schema, including building ontology, before we even attempt to tie ideas to objects. There have been great efforts to do this, such as SNOMED. But given the rapid pace of change in both disease categorization, as well as uncovering more biological pathways in systems medicine, this seems like a massive effort. And much of this is irrelevant to current medical practice.
So, I believe there needs to be a ‘middle way’, where we can both be opportunistic in terms of content when it needs to be structured, but leave most health data raw for the physician to interpret.
(3) Finally, I really do believe that most EHR vendors we have today won’t be around in 5 years – the EHR domain will be dominated by large IT companies that have more experience finding solutions to both unstructured and structured data.
I believe that all this concern now about certifying various levels of EHRs, according to some organizational need for control, will turn out to be detrimental in the long run.
Donna Manley says
One of my greatest concerns that has grown out of participating in a number of organizations trying to get their arms around the MU program is the magnitude of effort that is going to be required from a data administration standpoint.
Clearly as Gerry mentions, it appears that the vendor contingent is in “mark to market” mode. It begs to question is there an organization on a state, let alone national, level that will identify basic information that will be expected to be supplied in a standard manner across all vendor products?
If the healthcare industry looks to secure this in “all or nothing” mode, the MU programs will never realize the true benefit originally expected. I’m looking for a structured, phased approach that will offer some institutionalization of information and process. Is any organization stepping up to this challenge?
Great posts Mark and Gerry,
Contextual metadata is the key. Whether it is called structure or ontology or taxonomy or data governance, the basic concept is the same i.e. a given data element must have a commonly understood and uniformly implemented defintion.
In my experience, the most pernicious data quality issues are not data types or null values but issues of homonyms and synonyms. Homonyms occur when two data elements share the same name but have different defintions; Synonyms occur when two data elements share the same definition but have different names.
The extraordinary danger of mixing and matching registries, clinical repositories, administrative data sets and/or transactional feeds arises from the high degree of homonymy and synonymy across healthcare datasets and subject domains. In my experience, 80% of the work in a new application is integration work to either remove (via transformation) or redress (via business logic) data homonyms and synonyms. This estimate only accounts for the homonyms and synonyms of which you are aware. Extracting them and their definitions, whether during the requirements phase or during the causal investigation of a newly discovered bug, is an art much more than a science.
The incredible amount of work that has gone into efforts like UMLS speak to the Augean nature of a comprehensive and systematic approach to the problems of homonymy and synonymy. Unfortunately for all the brilliance of its design and idealism of its conception, when was the last time that you saw an application making extensive use of UMLS (as compared to SNOMED, ICD9, ICD10 or MeSH)?
In my own opinion, I believe that we need a new approach to capturing and utilizing contextual metadata. I know that there are some fledgeling efforts to move in this direction, but we are still waiting for the breakthrough.