A few days ago I singled out Integrating the Healthcare Enterprise (IHE), perhaps unfairly, in a posting about what seems to me to be widespread ignorance of a profound issue in e-health – that of information usefulness ( or the lack thereof ). That might be hard to believe, given that the USA government has committed $25Billion to try to improve the “meaningful use” of electronic health records. Although the definition of “meaningful use” is still rather vague, I have yet to see any evidence that the measures being adopted over there actually promote meaningful use. I contend that the major problem in supporting meaningful use is the lack of coherence in clinical information representation. $1000Billion won’t magically create meaningful use. Not until people understand the problem.
With respect to IHE, when I say “perhaps unfairly”, it is because IHE builds on standards that are already being used to some extent in the particular health domains to which IHE contributes. I’m not saying IHE is more culpable than other organisations or the Standards Development Organisations (SDOs) producing the underlying standards referenced by IHE. But because IHE are promoting adoption of their products without an overarching information architecture, they are too, in effect, building silos. The standards don’t standardise in a meaningful way!
My concern, as outlined in a previous post, is about the excessive and debilitating heterogeneity in information representation. I’d like to try to convey my concerns through the use of analogy – starting with three that hopefully will strike cords with clinicians; managers/administrators; and IT personnel and software developers, respectively. Unfortunately I’m not able to think of a good analogy of incoherence in clinical practice (hopefully it doesn’t happen!) , so this analogy is a particularly lame one. I’d welcome a better one.
Clinical: Imagine a clinician fronting up at the administration counter of her clinic/ward and asking for the case notes for the very sick patient Browne, only to be told the following:
- the case notes are spread across 7 physical locations
- you’ll have to go to the basement and read Mr Browne’s prior clinical history etched into the stone tablet in Bay A.13
- there is an audio cassette (#4568) sent from radiologist Clayton (id=84511207), reporting on the investigation (184.108.40.206088.12.45.19) you ordered on 3/4/10 for effectiveTime=20100902.
- you can find pathologist Oakbridge’s annotated reports on the biopsied specimens on microfiche, filed under code Z465-389
- current Rx for patient HI=8016745939 (aka Browne) includes warfarin, COX inhibitor ( code=A4451008:unable to resolve name, record invalid), “EES 400 mg/5 mL oral liquid: powder for, 100 mL”
- followup PET images of the nonhilar mediastinal nodes requested are stapled to the patient’s chart in ward 12B.
Administrative: Imagine the task of installing a new red printer cartridge in the HP printer in ward 12B:
- Cameron devises strategic plan and documents on A4 paper for circulation
- PM Kim uses PMBOK PHRM for organisation planning and staff acquisition
- Risk Management PM Susan uses p2msp to assess risk against business case parameters identified and imported from Project in a Box.
- James ( responsible for PL5 Scheduling) maps project data from MS-Project and produces Gantt chart – sends to printer in admin area ( which of course has red cartridge).
- QA Manager Pauline scans Gantt chart into her ISO 9001 documenting system and notices timeline has slipped. Requests project updates from all PMs. Collates and finds initial goal missing. Escalates to internal conflict resolution manager.
- Conflict Resolution Manager Sandy instigates review. Unable to locate paper copy of Cameron’s strategic plan because of organisational restructure. Sandy files report for action by complaints manager.
IT: Imagine being asked to parse a set of text documents whereby:
- three of the first five paragraphs SHALL BE encoded in EBCDIC,
- at least two paragraphs SHALL BE encoded in UTF-8,
- 0 or more paragraphs MAY BE encoded in ISO 8859-1
- paragraphs 9,11, and 13, if they exist MUST BE encoded in ASCII-7
- all monochrome illustrations of width >= 1200px or height >= 450px SHALL BE rendered as JPG
- 8 bit colour images SHALL BE rendered as GIF
- all colour images greater than 4 bits MAY BE rendered as GIF, PNG, BMP, SVG
The above analogies may seem flippant and fanciful, but they represent precisely the sort of incoherence we currently have with clinical information standards – both across standards, but even more often, and more frustratingly within one particular standard. The sad fact is that so few people can see the problem. The result is that these types of incoherent and often conflicting specifications are extremely difficult to implement, completely unmaintainable and probably clinically unsafe.
It’s not about myriads of arcane codes, nor codesets, nor hundreds of printed conformance statements, it’s about the capturing and preserving the meaning of the information!
The best examples that I have found of health data models that make some sort of sense are the ones from the Center for Disease Control in the US (http://www.cdc.gov/nedss/DataModels/phcdm.pdf) and the Canadian Conceptual Health Data Model (http://secure.cihi.ca/cihiweb/en/downloads/infostand_chdm_e_CHDMv2_31.pdf). The data model for the NHS in the UK seems underdeveloped.
Are there any others that I should be aware of?
OH wow – so true. Data has meaning in a context – take it out of that context and it loses meaning. I think this is what is behind the concept of archetypes e.g. as used in openEHR) – grouping and relating data items in a context. Also, it presents a major challenge for extracting and summarising clinical data for use in electronic health records – if you extract some data and take it out of its original context, then does it mean the same thing or can it be compared with other data?
Re data models – they seem wonderful and do make sense, but in the end, we buy systems from vendors that have their own legacy/proprietary data models, and we can’t even see their models, because they are commecial in confidence, so we just have to buy the systems, load them up and then get a feel for how the data model works (or doesn’t work) after we put it in. The HL7 RIM was supposed to help us wih integrating systems, and by logical extension, if vendors designed their systems based on he RIM, then it may all get a little easier?