Electronic Medical Record (EMR) adoption since 2008 has been dramatic primarily driven by the American Recovery and Reinvestment Act of 2009 (HITECH) and the Affordable Care Act of 2010. While this list is not all inclusive, perhaps some of the most important issues surrounding this phenomenon include improved patient outcomes, workflow efficiencies, best practices in capturing both structured and unstructured data, cost efficiencies, improved reimbursement rates, and interoperability.
How best to capture that data?
American healthcare systems continue to be in a state of flux and development. Amidst what has become a highway of healthcare initiatives riding the technology train as it chugs down the tracks, are an ever evolving set of criteria and parameters. Beyond MU, I believe there is lack of clarity around what data will be needed to achieve the goals of lower re-admissions, more cost effective delivery of care, and overall better patient outcomes. Are we collecting the right data now? How do we come to understand what the right data is? Are EMRs nimble enough to adapt their systems to extract different data and have the ability to go back and extract data that was previously collected and make it relevant with newly collected information.
Structured versus unstructured.
EMRs are essentially data repositories or more simply, databases. Massive amounts of highly defined bits of data that can be queried to show historical medical trends and lend credence to predictive models for improved patient outcomes in the future. From one perspective this is the easy part – the structured data.
What about unstructured data, in particular, the narrative? The dictated narrative portion of patient healthcare encounters resides largely in non-XML files, usually an RTF (Rich Text Format) file. While these files do not have “fields” per se, they are searchable and are able to be parsed using natural language processing engines.
During the last five years there has been contention between EMR vendors attempting to eliminate the dictated narrative and medical transcription companies attempting to make the case that the narrative continues to be essential to an optimized accurate description of a patient encounter.
While dictation has been much maligned over the last five plus years, it is starting to become clear and acknowledged that point-and-click simply can’t capture all the necessary data. There are a couple of problems with point-and-click data capture. One issue is that it is impossible to address every possible parameter in drop-down menus. An interesting scenario was described to me recently by Simon Beaulah, Director of Healthcare Strategy at Linguamatics, a leading NLP text mining platform. He states: “Significant insight into a patient’s potential clinical risk can be gained from analyzing references to their home life. For example, consider these different comments from discharge planners: “lives alone” and “going home with husband”. Analysis of this unstructured text indicates that the social support structure is very different between the two cases and that the former is higher risk and requiring more care coordination. This type of insight, which is so vital to personalizing the patient experience, cannot be captured effectively with point and click interfaces or drop down menus.”
Recently the American Medical Informatics Association produced a report entitled: Report of the AMIA EHR 2020 Task Force on the Status and Future Direction of EHRs. The opening paragraph reads: ” . . . . with this broad adoption many clinicians are voicing concerns that EHR use has had unintended clinical consequences, including reduced time for patient-clinician interaction, transferred new and burdensome data entry tasks to front-line clinicians, and lengthened workdays. . . . These frustrations are contributing to a decreased satisfaction with professional work life. In professional journals, press reports, on wards and in clinics, we have heard of the difficulties that the transition to EHRs has created. Clinicians ask for help getting through their days, which often extend into evenings devoted to writing notes. Examples of comments include “Computers always make things faster and cheaper. Not this time.” and “My doctor pays more attention to the computer than to me.”
As we step into the next phase of improved healthcare analysis and delivery it’s important for us to recognize that the need for better data analytics is just getting started with the move to affordable care requiring better population health, clinical risk modeling and care coordination. It’s a landscape ripe with challenges and opportunities.
This is a abbreviated version of an article written by Linda Sullivan published in the October issue of “For The Record” magazine.
No related posts.