Why IT people can’t build information systems

(on their own)

Every so often I remember how we were taught to build information systems and software. One of the steps is called ‘requirements capture’. The IT people are supposed to go and interrogate domain experts, in a step called ‘use case modelling’, obtaining those diamonds of information that will allow them to build the system those experts want.

There’s only one problem with that. In all real domains, the IT people and domain experts have no clue what each 0ther is saying. And yet the IT people still go and build a system. Any system. And that’s why most information systems are a) semantically broken and b) can’t keep up to date with new requirements.

We have known this doesn’t work for 20 years in health (some savvy people knew it for a lot longer). In health, just like any other real domain, you can’t afford to put any of the following in the software:

  • domain content / information models
  • workflows
  • terminology
  • ontology
  • higher-level artefacts such as guidelines (health), business rules etc.

The solution is to find ways of enabling domain experts themselves to build models and descriptions of their domain. Our contribution in openEHR was to provide a way of doing the first, the so-called ‘archetypes’.

To see just how unlikely it is that IT people could build a model of health domain content, such as the information used to record ‘problems’ and ‘diagnoses’, have a look at the contributors list of the problem/diagnosis archetype in the openEHR CKM:


Here’s the list of contributors, in a more digestible form (total = 59):


Norway (10)

  • Nasjonal IKTNorwegian Review Summary
  • DIPSTomas Alme, Bjoern Naess, Lars Karlsen
  • Bergen Hospital Trust*Silje Ljosland Bakke
  • Helse BergenJohn Tore Valand
  • Haukeland University hospitalLars Bitsch-Larsen, Sabine Leh
  • National Centre for Integrated Care and TelemedicineEinar Fosse
  • Oslo University HospitalHallvard Lærum

New Zealand (1)

  • University of AucklandKoray Atalag

Australia (33)

  • Royal Australian College of General PractitionersChris Mitchell, Camilla Preeston, Kylie Young
  • NEHTADavid McKillop, Stewart Morrison, Margaret Prichard, Cathy Richardson, Robyn Richards, Richard Townley-O’Neill, John Bennett, Matthew Cordell, Mary Kelaher, Robert L’egan, Andrew Goodchild
  • Queensland HealthStephen Chu, David Evans
  • Nursing Informatics Australia: Paul Donaldson
  • Flinders Medical CentreGail Easterbrook
  • Ocean Informatics*Heather Leslie, Sam Heard, Hugh Leslie, Shahla Foozonkhah
  • ACT HealthIan Bull
  • (individuals): Peter Garcia-Webb, Thilo Schuler, Gordon Tomes,
  • Ambulance VictoriaRohan Martin
  • Melbourne East GP NetworkChris Pearce
  • Adelaide Northern Division of General Practice LtdJodie Pycroft
  • ACCTI-UoWDonna Truran
  • Llewelyn Grain InformaticsHeather Grain
  • cpcTrina Gregory
  • EJSH ConsultingEvelyn Hovenga

Sweden (2)

  • Karolinska InstitutetNadim Anani
  • Cambio Healthcare SystemsRong Chen

United Kingdom (4)

  • AllscriptsSteve Bentley
  • Cardiff UniversityEd Conley
  • freshEHR Clinical Informatics*Ian McNicoll
  • University College LondonAnoop Shah

Brazil (1)

  • National University of BrasiliaJussara Rotzsch

Chile (1)

  • (individuals): Sergio Carmona

Spain (1)

  • UPNA (Public University of Navarre) – CHN (Complejo Hospitalario de Navarra)Aitor Eguzkitza

Germany (2)

  • CompugroupJörg Niggemann
  • Ocean InformaticsSebastian Garde

Netherlands (1)

  • Results 4 CareAnneke Goossen

Canada (1)

  • IRIS Systems, Inc.Eugene Igras

Japan (1)

  • Ehime Univ.:  Eizen Kimura
  • Kyoto UniversityShinji Kobayashi

Slovenia (1)

  • Marand d.o.o.Andrej Orel

So let’s just try to imagine by what means the IT experts trying to build a system or component that knows (properly) about recording Diagnoses and Problems could get access to the expert input of 59 people from 13 countries, many of them exceedingly busy senior clinical professionals?

Answer: there isn’t one. Developers on their own have no way to achieve this. They don’t even know who to ask. Now, multiply this problem by the other 400+ archetypes in the CKM, and remember that CKM has 1,500 or so registered clinical reviewers. And let’s assume that the ultimate number of models needed to cover medicine is more like 2,000.

It’s clear that there is no hope whatever of software developers figuring this out. Some health informatics types might imagine that the developers could go to some standards meetings (ISO, HL7, ASTM etc). But the clinical people won’t be there; most of them don’t know the first thing about health informatics standards.

The only way to achieve semantic modelling of the health domain is for domain models to be built as a completely independent activity, separated from specific software products, solutions, or technologies, run by and for domain experts.

Semantically rich system development is only possible in the health domain in this ‘inside-out’ way. Clearly, making it happen for real requires significant vision, planning and resources.

The question is: do today’s health IT programmes and funding frameworks have it?

About wolandscat

I work on semantic architectures for interoperability of information systems. Much of my time is spent studying biomedical knowledge using methods from philosophy, particularly ontology and epistemology.
This entry was posted in Computing, Health Informatics, openehr and tagged , , , . Bookmark the permalink.

4 Responses to Why IT people can’t build information systems

  1. markcsg says:

    Good points, well made.

  2. I love it – especially as I’m mentioned by name 🙂 !

  3. Pingback: Would you like to build health and care software 100 times faster? | Woodcote Consulting

  4. Pingback: Ewan Davis: the content challenge – Digital Health Intelligence

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s