Too Much, Not Enough, All at Once
The title of this post is from a Jack Kerouac quote, having to do with falling in love. I have a more prosaic interpretation: it’s the key problem in health IT, everywhere you look. Health IT solutions provide a lot of data, but it’s usually too much, and what the user is actually looking for is often missing. Too much and not enough at the same time. Because data itself only becomes useful when context and meaning transform it into information. This can be especially true when data and information are presented mixed together, and in high volumes.
Example 1: CCDs (Continuity of Care Documents), which are Discharge document packages in C-CDA format, sent by all EMRs to physician offices as required by Meaningful Use / Promoting Interoperability rules. The hospital (sending party) puts a lot of information in the CCD: medications, allergies, problem list, discharge instructions, vital signs and results, anything the physician office (receiving party) might possibly need—as well as disclaimers and lengthy legalese that simply clutter up the report. Consequently, physician offices who receive these packages complain that these CCDs or Discharge Packages are too much, too long, spit out 80 pages on their fax machines, but don’t include the CT-scan report that they would actually like to see. Too much and not enough.
Example 2: Carequality, Commonwell, Care Everywhere, and other Interoperability Frameworks and Data Exchanges. The intentions are great and the need for data exchange is real. Providers and patients need the data to flow between all care participants. But it’s often too much and not enough: too much data to wade through (select a CCD from a list of patient encounters), and physicians find it hard to find the one thing they are looking for—last week’s CT scan report or discharge summary.
Example 3: Most EHRs (Electronic Health Records). No need to elaborate. EHRs (and the less comprehensive subset EMR or Electronic Medical Records) contain a lot of critical pieces of information, buried in masses of irrelevant data. Look around and you’ll see many more examples.
How to fix this? I offer three suggestions:
- First: Search. Thanks to Google, we all live in a Search world, and we have all become accustomed and trained to find what we’re looking for by getting really good at search. Yet EHRs, Data Exchanges, CCDs, and other health data sets don’t employ searchability to the best effect. Rather than presenting menus, tabs, indexes, links, a simple search line would allow the user to cut through all the navigation and pinpoint the data they need. eClinicalWorks is proposing a search function called Prisma, to search within a patient’s record, not just within your EHR but also in any connected data (through the interoperability frameworks). I think that’s the right approach. All EHRs and Health Information Exchanges should support searchability within a patient record.
- Second: Switch from push to pull. CCDs (see above) are usually pushed to physician offices—via secure messaging, or in many cases via fax, the most outdated version of push. EHRs, per Meaningful Use mandate, push CCD’s and discharge packages—with too much information but not the one thing the user would want to know. Going to a Search method is like switching from push to pull—let the user decide what they are looking for, and pull just that. This is the appeal of FHIR vs. the old HL7 standards. FHIR allows users and apps to define precisely what data/information they need and retrieve just that.
- Third: AI. Can we build an AI that understands relevance? Which finds and presents relevant information, based on user and setting and information need? See this vision (link below), from three writers I admire – I would like to suggest that information optimization should be a scenario for AI in EHRs. If we can design self-driving cars, self-driving EHRs may not be far behind…