Judy Faulkner, spaghetti and Elon’s I’m a celebrity – get out of here.

Its been a funny month. A few LinkedIn articles have made me stop and think.

To start with, Erik Vermeulen at EY brought my attention to a Hey Judy thoughtpiece from Judy Faulkner at Epic. My very crude summary of her piece is Epic don’t acquire comonents from third parties in their solutions because integration is hard. She is right. Composable IT solutions are hard, but so is joining up health and care across different organisations, disciplines and IT systems.

If we acquired someone else’s key product, it would have different code, different terminology, different design, different colors, different placements, and so forth.

Judy Faulkner, No Acquisitions, No Investments. 7 November 2022

The language of health and care

Judy specifically calls out terminologies (the language of healthcare) as an area where she “goes her own way”. This possibly explains why most weeks I have someone telling me Epic won’t use the International Standards for health terminologies and instead expect their potential customers to “go their way”.

A quick aside – why does this matter? If you don’t have a common way of describing health and care, you can’t join up the health and care provided by different people. Just imagine an American trying to find a cheap packet of acetaminophen in a British supermarket.

Without the common language we prevent joined up care.

Without common coded terminologies, we exclude computers from being anything more than a glorified FAX machine with a filing cabinet bolted on.

Delivering an integrated health and care system

One interpretation of Judy’s post is that you shouldn’t try to integrate the care delivered by different health care providers – those that use Epic and anyone that does not. Best use Epic throughout.

This is a credible approach for small-enough communities or where patients are locked into using a single health care provider organisation, perhaps because their insurer has a special deal. It does require people to stay within that community. This does happen – people serving the King in the military or “detained at the King’s pleasure”. But most of these people leave those communities when discharged or released. Friends in the United States tell me that simply using the same product as an electronic medical record management system doesn’t guarantee a smooth transition from service personnel to veteran.

I’m not picking on Epic here – the Thoughts and Stories piece stimulated the post. Most Electronic Medical Record Management Systems are designed to work in closed markets, where customer (insurer) and patient are encouraged to stay within one health care provider for various reasons. This market approach was encouraged in England with the “purchaser / provider” split of the 1990’s. However, joining up health and care is going to remain difficult when care records are stranded on Electronic Medical Record islands.

The alternative …

So, if using a single Electronic Medical Record management system isn’t the solution for England, what is the alternative. It’s probably best to describe the key components of a care record.

The boring bit – good master data and good reference data. All systems need these. In England they comprise the master data supplied in the Person Demographics Service and the Organisational Data Service coupled with the reference data (terminologies and classifications) in the NHS Terminology Server. If systems don’t use the up-to-date values of these, we won’t have integrated health and care.

The HL7 FHIR resources for the participants in healthcare and terminologies are my preferred way of sharing master data and reference data. They are the only credible international standards.

I’ll include the OpenEHR archetypes (domain models for clinical domains) in the reference data. They are a common resource we can refer to, though they have their own serialisation – ISO 13606-2 Archetype Definition Language.

The process bit. Health and care is like all other industries. There is a simple process. An initial contact is made, a case starts, we spend time working out how to handle the case, doing things to improve the outcome and seeking help from other people, departments and organisations. At the end of the case, we close it (discharge or death) or hand back to the person who made the initial contact (patient, GP, referrer). Unlike most industries, we’re not good at describing these processes in a way that is useful to a computer (or staff / patients). For more information on the clinical process, head over to my UML concept model conforming to ISO 13940:2015 – A System of Concepts for the Continuity of Care.

HL7 FHIR provides a great set of RESTful APIs for systems to transact with each other (though often these are implemented as messaging specifications without all the RESTful capabilities).

The specialist bit – I often hear that each clinical setting is special. They need special records only used in their area. They are often right, however I vividly remember a conversation between an orthopaedic surgeon and an obstetrician when we were trying to have a common admission form for patients admitted to a local district general hospital. I’ll paraphrase the conversation (my memory is vivid but not necessarily accurate).

• Why do I need all this obs and gynae stuff on my admission forms?
  • Do you ever use an X-Ray for diagnosis?
• Yes
  • Are any of your patients women?
• Yes
  • Could any of them be pregnant?
• Yes
  (pause)
• We’ll keep the obs and gynae bit then!

It’s important that clinical teams and the clinical networks they work in can agree and share the specialist bit, but that it’s only the clinical bits that are specialist. The usual approach is to use forms (in some systems powerful, in others, smart) with reusable components, often questions and a domain of valid answers. With common master data and reference data, these forms / templates provide a common understanding of the meaning of the record.

There are many ways of representing forms (and reports) but some examples are the World Wide Web consortium’s XForms, ISO 19763-13 and OpenEHR templates. HL7 offers FHIR Questionnaire and Structured Data Capture.

See So what does it mean? – part 3 – forms.

The missing bit – a common way to represent a health and/or care record. There are many alternatives available but the two I am most attracted to are the ISO 13606 reference information model and its close relative, the OpenEHR Reference Model. It doesn’t really matter which of these we use, but a common syntax for exchanging the care record without being focused on the clinical context (from the forms above) is critical.

I’m not suggesting we rip out all electronic medical record management systems. Instead, I think the world would be a healthier place if we expect these systems to make the records they withhold available in the common format. A format that allows a patient (or their care funder / insurer) to see the full picture and allows everyone providing care to the patient to access the pertinent information they need in a format they can use. For a working title, I’ll call this a Portable Care Record. The full record being a set of organisational or system-specific care records in a common format for an individual. See So what does it all mean? – part 1 – Lloyd George envelopes and So what does it all mean? – part 4 – Electronic records.

The spaghetti?

Dave Turner shared an excellent LinkedIn thought piece entitled Conway’s Law, Spaghetti and the NHS. The issue is …

Patients who cross these organisational and system boundaries are met with a suboptimal service. They must repeat their story because their patient record is not accessible, and the authorised health and care professionals must ask a lot of questions to try and rebuild this picture.

David Turner, Conway’s Law, Spaghetti and the NHS. Published 21 November

I’ll not restate Dave’s argument; save to say I agree!

So, what has this to do with a Portable Care Record?

Its not a single data store, not a replacement for the operational datastores and not under one person’s control. Instead it is a common structure made available by all the health care parties so the whole story can be brought together when needed.

One way to breakdown organisational communication silos is for all organisations to use a common approach to messages. The combination of SMTP, the e-mail protocol, and HTML, the text representation format, is that sending e-mails across organisations is a trivial exercise. We need a health care equivalent. The SMTP is analogous to an event broker with events that include embedded detail, though increasingly the message payload is left as a resource to be accessed when needed. The NHS Event Management System is an alternative event broker, with a pointer to the Portable Care Record as the payload.

Another approach is to open the silos for search – Schema.org metadata transforms the way we search for information with the large search engines. The NHS Record Locator Service could, with some refinement, become a common way to find various parts of a portable care record. If done properly, it could also search for common fragments (disease prevalence) across a population’s records as well.

The combination of search and notification are essential for joined up health care.

and Elon Musk’s I’m a celebrity – get out of here?

The last few weeks have shown how important the intangible asset of trust is. Advertisers have played a low key to the FIFA World Cup before the start of the competition so that the diverse approaches nation states have taken to diversity don’t cloud their “messages” enticing us to buy this, fly via here or visit there.

And seemingly, as the trust in the content being disseminated through Twitter drops with Elon’s takeover, the advertising revenue available to Twitter drops too. I, like most people, have no idea what drove Elon to buy Twitter or how he is shaping it for his vision. It is, however, a good illustration that your partners need to trust you to manage their data carefully. This includes Twitter employees, of whom there seem to be fewer than a few weeks ago.

The Twitter business model looks to be evolving, with the possibility of charging to share content. It provoked the following response:

F**k that, they should pay me. If that gets instituted, I’m gone like Enron.

Stephen King

Which is a flavour of “whose data is it anyhow” – if Twitter make advertising money out of Stephen King’s tweets the author has a point. You can imagine a similar “Whose record is it anyhow” argument for the patient being in full control of how their record is shared.

I’m reminded I was told about a couple of large health care organisations trying to join up care in the United States. They invested over $10M to get a solution working, only to find the receiving clinical teams didn’t trust the records from the sending clinical team. Trust is fickle. Trust of competitors, rare.

The globe is full of examples of health insurance companies making enticing offers to prospective customers (employers or citizens) with free Fitbits or iWatches in exchange for wellness data. We then have the unpleasant sight of other organisations offering to advise employers when a person’s wellness is about to become expensive to insure (better get rid). Examples like these won’t drive up trust.

In the UK, knowing your ambulance provider can draw on your primary care record is reassuring. Finding out there was a proposal for Border Force to request your home address may not be. Better not tell the ambulance team where you live.

Decentralised twitter

The good guys on the PC Pro Podcast have been sharing their regular thoughts on the Twitter news in their “Elon In Briefs” section. They have been exploring an alternative platform, Mastodon.

“Each Mastodon server is a completely independent entity, able to interoperate with others to form one global social network.

Mastodon gGmbH

Each organisation can have it’s own content rules – decide what can go on the platform and, more importantly, what cannot.

Decentralised care records

This is remarkably like the health and care ecosystem (separate organisational or departmental systems into independently controlled silos) however, to date, we haven’t found the way for them to effectively interoperate with each other. They talk to each other (we improve the efficiency of health care transactions) but haven’t improved the effectiveness of record sharing.

What Twitter and Mastodon illustrate is that if we did join systems, each organisation would need to trust all the other organisations it peered with to handle their care records responsibly.

And in conclusion

Judy Faulkner is right – joining up health care is hard. Keep all the data on one system and it is easier, but Conway’s law kicks us and we have independent clinical teams all wanting the best experience for their care activities which the single systems never deliver. There’s a regression to the mean. I don’t subscribe to her solution (just us Epic) but can’t fault her logic for her customers.

But joining up health care is the best way we have of living healthy and independent for longer. It’s also a way we can minimise the unnecessary questions, investigations and interventions and maximise the efficiency of the use of limited NHS resources (staff, money and equipment).

A Portable Care Record is my preferred way of achieving continuity of care across specialty, department, setting and, most of all, organisation.