Review of conventional EMR systems

1.0) Review of conventional EMR systems

Two terms are used namely, EMR and EHR where the latter is an acronym for Electronic Health Record. These tend to be used interchangeably but, OnICS tends to think of the EMR as being a specific entity within a more comprehensive health record system that includes different types of record including public health.

Currently, four (4) methods are available for the EMR but, there are problems associated with each.

1.1) Electronic Medical Records (EMR) Methods

1.1 Conventional dictation/transcription using Microsoft WORD

1.2 Voice recognition system dictation – eliminates transcription

1.3 Templates – very large number required

1.4 Neural net - McCulloch, W.S. and Pitts, W. (1943) “A logical calculus of the ideas immanent in nervous activity” Bull. Mathematical Biophys, 5; 115-133

1.2) General Problems Associated with Conventional EMRs

1.2.1) Dictation and typing leave much to desired because everything that is needed has to be remembered

1.2.2) Systems are available for “computerising” medical notes but, most of these are hopelessly inadequate

1.2.3) Voice recognition systems rely on dictation

1.2.4) Steven Stack, and a committee of the American Medical Association (AMA Wire, 16-September-2014), identified various EMR problems in an important article entitled

“8 top challenges and solutions for making EHRs usable”

The phrase “making EHRs usable” immediately implies that EHRs are not usable, or at very best they are usable but not user-friendly, hence there must be some fundamental design problems.

The “solutions” to solve and implement the 8 top challenges were not stated. And, at the outset the author writes; “The health system desperately needs working information technology (IT) to help support quality care. ….. The current generation of EHRs and the way they are deployed is not supporting the quality of care we need it to.”

The 8 top Challenges. The EHR should;

Enhance a physician's ability to provide high-quality patient care. Under this heading Stack states that “Poor EHR design gets in the way of face-to-face interaction with patients because physicians are force to spend more time documenting required information of questionable value. Instead EHRs should be designed to enable physician-patient engagement.”
Support team-based care
Promote care co-ordination
Offer product modularity and configurability

Reduce cognitive workload but,

“. . physicians spend significant time navigation their EHRs but, the quality of the clinical paper narrative is more succinct and reflective of the pertinent clinical information”

Promote interoperability and data exchange
Facilitate digital patient engagement
Expedite user input into product design and post-implementation feedback.

The article inferred that EMRs are (1) very time consuming, (2) difficult to use, (3) not intuitively obvious and (4) require extensive training. As a result, they are not cost-effective and some US physicians are now employing “scribes” to work their EMRs, hardly an endorsement of their utility.

1.2.5) A further paper, “Improving Electronic Health Record Usability Requires Transparency” (Hodgkins, M., Ratwani, R. and Bates, D.W. JAMA Viewpoint 7-December-2018) stated “EHR usability, which is the extent to which this technology can be used efficiently, effectively and safely by clinicians to deliver care, has emerged as one of the most pressing issues in health care”

The authors note that contracts may contain “gag clauses” that require vendor authorization for clinicians to share data they identify. They state this is “EHR information blocking that can harm patient care.”

In addition, they note that this blocking can extend to 3rd party software participation

1.3) Specific Problems Generating EHRs

1.3.1) “Voice Recognition”

1.3.1.1) Specialist microphones are needed

1.3.1.2) User diction clarity and accent are both critical

1.3.1.3) Quiet room required – to exclude extraneous noise

1.3.1.4) No direct patient participation possible

1.3.1.5) No documentation by support team-based care

1.3.1.6) Promotion of care co-ordination difficult

1.3.1.7) Data acquisition cannot be digital

1.3.2) “Templates”

1.3.2.1) Large numbers of templates required

1.3.2.2) Corrections to gender, tense and other grammar are frequently needed

1.3.2.3) No direct patient participation

1.3.2.4) No support team-based care

1.3.2.5) Promotion of care co-ordination difficult

1.3.2.6) Data acquisition is not digital

1.3.3) “Neural nets”

The neural net has to be trained and generating specialist medical text in a neural net with a lexicon of 5000 words and 2 processing layers would require at least 6 × 10 pathways. With 3 processing layers this would increase to over 3 × 10 pathways.

This is an astronomical, but possible, problem. It would take an age to train the net for all possible outcomes and would be of doubtful cost effectiveness. However, IBM could do this with their super-computer (Ferrucci, D. et al (2013) “Watson: beyond Jeopardy!”. Artificial Intelligence. 199, 93-105.)

1.4) How and why did EMR problems arise?

1.4.1) Fundamental problem: The first EMRs were built by computer aficionados with limited, if any, medical insight and no demonstrable medical knowledge.

1.4.2) Designed for administration, coding, billing & reimbursement

1.4.3) Later, almost as an after-thought, clinical data were inserted into “fields” reserved for dictated notes.

1.4.4) Initially, there was no provision for clinical data-point input

1.4.5) Belated intervention by regulatory bodies (particularly US congress) to heed AMA recommendations that were implemented only after the mandatory requirement to upgrade coding from ICD9 to ICD10.