EHRs struggle to keep pace with providers’ needs

Register now

Perhaps the biggest shortcoming springing from broad adoption of electronic health records is their limited ability to interoperate with one another now that they’ve been sutured into healthcare systems everywhere.

But a more fundamental shortcoming is shaping up as just as big an issue: the limited ability of EHRs to keep pace with advancing demands for more varied, easily accessed, comprehensive information for direct care and analysis.

Changing times call for maximum flexibility in how clinicians access record systems and the full picture they need to see from not only structured textual data but also diagnostic images, wave forms, free text and, coming down the pike, genomic sequencing. To manage all that, EHR technology has to leave the 20th century behind and reduce its reliance on venerable relational databases, experts say.

“You can’t take 1980, or even earlier, architectural platforms and make them do the things that people are used to being able to do on their mobile devices, or with internet connectivity on web-based devices,” says Charlotte Weaver, a long-time nurse informaticist and EHR technology consultant. “That’s what the level of expectation is, and we’ve got EHR systems that just can’t stretch to get there.”

Most EHRs are built on the data infrastructure known as MUMPS, or something MUMPS-like, notes Aaron Miri, CIO of Walnut Hill Medical Center, Dallas. A very fast relational database dating from the mid-1960s, its approach to organizing data into rows and columns was designed for text-based interaction, he says. In recent years, though, EHRs have had to deal with a proliferation of non-text information. “As these EHR engines have grown, the MUMPS data structure begins to fall apart,” Miri says. “And so the various vendors have built pseudo-repositories around the MUMPS database structure that allowed for things like wave recordings, vital signs, PACS images and whatnot, to be stored in there.”

It’s a complex shuffle in which the main database has to make separate calls to the disparate sources of these other kinds of information, seeking to grab specific patient details for display on the EHR user interface--a recipe for glitches, and also for the one performance flaw clinicians won’t stand for: time lags.

EHR vendors have a sizable installed base of such old-guard technology, thanks to the incentives of the HITECH Act, but some are innovating to address acknowledged shortcomings in their existing systems. “MUMPS is an outdated architecture that does not accommodate the moving forward of the evolutionary path,” says Ron Dobes, vice president of infrastructure and architecture for McKesson Enterprise Information Solutions. He’s chief architect of a contemporary approach to how EHRs are put together.

The limitations of legacy architecture have become more noticeable just as business and regulatory demands to produce data for myriad purposes have risen substantially. The consequences for healthcare operations include:

  • Higher requirements for speedy performance of multiple databases that drive up costs of IT investment.
  • Clunky methods of creating discrete, structured data that take away from clinical efficiency and stilt the narratives of patient stories and physician-to-physician communication.
  • Difficulty in translating legacy architecture to the tap-and-swipe world of tablets and smart phones, which impinges on clinicians’ mobility in ambulatory and community settings.

EHR vendors have figured out how to wrap secondary databases around the main platform, but it’s been challenging to maintain EHR performance without a time lag from the added interaction with different architectures, says Miri. “What you’re starting to see now are the vendors’ sort of struggling with the pace of the ‘internet of things,’ because the last thing they want is an EHR that slows down because of all these different calls.”

As a result, he says, some vendors are mandating strict requirements for computer architecture, such as a boost in the rate of operations per second it can perform to and from databases. That’s a capability the provider community is hard-pressed to supply because it means millions more dollars in infrastructure investment.

An alternative is to switch to cloud hosting instead of paying for the capacity on location, says Miri, but that doesn’t necessarily resolve the performance dilemma. After contending with internet connectivity, distances from cloud to user and issues of remote storage, “how that translates to the providers on the floor, the actual clinicians, is a slow and sluggish experience.”

Adding to the drag is the rising devotion to structured data, a great way to pack a relational database with findable nuggets but with the side effect of weighing down care givers, says Weaver. “We’ve built this documentation that is so immensely burdensome for physicians--and, in acute care, heavily nursing--and it’s not an effective communication tool that supports the team. It’s siloed, it’s templated, it’s predefined fields, so you don’t even get the patient’s story.”

As for usability, “we live in an internet cloud platform world today,” Weaver adds, and the flexibility and nimbleness of that world compared with the rigidity of the relational database is “the difference between the horse-drawn wagon and the Ford Model T. It’s profoundly different enabling technology.”

Innovative tools that comb through non-discrete data may tilt the “constant tension in medicine between structured and unstructured data” away from the current emphasis on discrete data elements, which are central to some meaningful use criteria and the grist for quality measures, says Doug Fridsma, CEO of the American Medical Informatics Association.

“Just because you can structure something doesn’t mean you should,” he says. “You need both: You need the ability to capture granular, structured data, and you need the ability to have unstructured data that doesn’t necessarily have to fit neatly into standard relational databases.”

Though structured data is valuable for delivery of care, clinical decision support and alerts, “unstructured data is really about discovery,” Fridsma says. Instead of defining in advance the data elements, a push within the Big Data community for retrieving data from free text assumes nothing and attempts to “find little triggers and word associations that you didn’t think existed, and extract those from that big pile of data and then use that to help inform patient-care delivery choices.”

Looking ahead to precision medicine, emerging techniques to unlock unstructured data can yield great benefit, “because that’s where you can mine to see if there are unknown associations that potentially could be predictive of disease, or that create associations between the patient and their genomic information,” he explains.

For reprint and licensing requests for this article, click here.