All Blood Pressures are not created equal! (Part II)

Following on from the previous post... If we want to compare and review blood pressure (BP) over time from all sources and clinical situations, then they need to be recorded in a consistent manner. We also need a common structure to facilitate accurate data exchange without manipulation.

Most systems have a proprietary database comprising a home-grown data structure – so if there are 'n' thousands of clinical vendors in existence, there are probably way more than 'n' thousands of technical approaches to specifying the clinical documentation and process that are so familiar to you. HealthVault is one that publishes its data specifications publicly but most private companies are not so open.

There are also a number of approaches that are proposing use of common data structures to support health information sharing. You can see some of my thoughts on this in previous posts – and in my opinion this is definitely the way to go if you want/need to do smart and active things with health data. Some examples are HL7 templates and openEHR archetypes.

Let's take a look at some publicly available representations of Blood Pressure...

On the right is the Microsoft HealthVault Blood Pressure specification (click image to enlarge):

It's largely in 'techspeak', although if you examine the specification itself you will find only three clinical elements:

• Systolic
• Diastolic and...
• Pulse!

Ouch... Pulse is usually independent of BP measurement for most clinicians, even if it may commonly be measured simultaneously. This model cannot currently capture event the basice clinical requirements identified in the previous post - not even recording that a large cuff was used for an obese patient, to indicate whether the measurement is accurate and able to be interpreted confidently. This model is really not rich enough to cater with our clinical recording requirements for Blood Pressure.

At right is a HL7 model for Blood Pressure (click image to enlarge):

Umm... maybe a little difficult to understand. It's hard to see the scope of the content, and almost impossible for clinicians to have input into, which is a significant issue in terms of achieving accessible and high quality data definitions.

Finally, openEHR has a number of ways of representing the data model for Blood Pressure. It is a maximal data set intending to capture all information about Blood Pressure measurements for all the clinical and research situations described previously, and more.

On the right is the simplest representation of the openEHR blood pressure archetype – displaying the content in a non-technical way so that all clinicians can focus and comment on the  clinical content itself.

At right is another way that the same BP archetype is displayed to clinicians and informaticians - specifically used when seeking detailed comments about the clinical content on a per element basis during collaborative online archetype reviews. While it may help to have an understanding of data types eg Text or Quantity or Date, no deep understanding of openEHR is required of participants.

In the interests of full disclosure, I have been intimately involved in the development of the openEHR Blood Pressure archetype which has been developed and agreed with a lot of clinical input, and places a lot of emphasis on being approachable by clinicians and incorporating their feedback.  The representations you see here are not perfect, but grassroots clinicians are participating actively in online reviews of archetypes.  The blood pressure archetype was published after consensus was reached on the content by 30 clinicians, informaticians and engineers from 13 countries.

There is no doubt that at first glance the archetype may appear too rich. One of the key differentiators in openEHR is the capability to take this maximum data set and activate/display only the parts that are relevant to the clinical scenario requirements eg for primary care or a paediatric patient. This detailed discussion is out of the scope of this post, and will be left until a later date.

The focus of the health IT domain has largely been technical. Involvement of clinicians has often been token or as a secondary process.  It is critical for successful development of EHR systems that clinicians are able to be actively involved in development and quality control of the clinical content that underpins these systems - clinicians need to request involvement. At the same time, the health IT domain needs to 'un-technify' eHealth - clinicians need support to participate, so that more than just a few elite informaticians can view these data structures and contribute to their development and quality review.

In these two posts we have explored a little about the scope of clinical models and their representation. Yet again, this is still only the tip of the iceberg re health data. The list goes on, including the use of terminology and the non-trivial technical requirements that are required to underpin these clinical models so that we have know exactly what this data means - far more complex than can be supplied by a simple XML model. Now that really starts to open a can of worms... for later!

All Blood Pressures are not created equal! (Part I)

Clinicians, have you ever had a look at the data structures underpinning your EHR - the models of clinical content? If you don't understand the principles of what your data structure is then your EHR may not do, or ever be able to do, all the things you need it to do to record your clinical notes. You need a sense of what is 'under the bonnet' and what its capabilities are. Take Blood Pressure as an example. It is probably the most easily understood clinical concept to both clinicians and patients. We clinicians tend to think of it as a very simple measurement to record on paper, and we expect our EHR applications to do it in a similar way.

So let's brainstorm about what we need to record blood pressure. The typical primary care provider such as a General Practitioner or Nurse will record the Systolic and Diastolic pressures, plus maybe a blood pressure cuff size if it is different to the 'normal' cuff, and position of the patient if recording a postural drop. This is straightforward, 'bread and butter' stuff for clinicians.

So the base data requirements are:

• Systolic – a quantity that is a pressure, measured in millimeters of mercury (mmHg)
• Diastolic – ditto
• Cuff size – most will measure using a 'normal' adult cuff, recording a size only if they use a larger or smaller cuff.
• Position – most patients in primary care will be sitting; most in hospital will be reclining

While this may cater for the large majority of our BP measurements, there are other less common situations where it is critical that BP be recorded appropriately and in context. So we need a data specification that also caters for the following situations where BP is measured, such as:

• The cardiologist or exercise physiologist may need additional data about the level of exertion for use in stress testing or research. A BP measured at rest needs to be interpreted differently to that recorded after jogging for 10 minutes.
• Recording a series of BP readings over a 24 hour period to determine the 24 hour average, plus knowing the patient's concurrent sleep status – this is now regarded as key diagnostic criteria for diagnosis of hypertension in Europe.
• Measuring data from home BP machines – many of these actually record the Mean Arterial Pressure and use an algorithm to determine the Systolic/Diastolic readings displayed.
• Automatic BP measuring every 5 minutes during surgery or an admission to Intensive Care by 'the machine that goes beep'.
• Paramedics taking a measurement of a trapped patient at the scene of an car accident.

Other questions might be clinically significant in some circumstances include:

• What Korotkoff sound was used?
• What about the implications for paediatric measurement?
• Where was the measurement taken? Which limb? Finger vs upper arm?
• How was it taken – auscultation; palpation, machine; or invasive?
• What about pregnant women who need to have their BP measured with lateral tilt to get a true reading – do we need to record that detail?

It may be very important to have this information recorded in some situations.

Maybe blood pressure is maybe not so simple, after all.

Clinicians understand clinical nuances well, recording the contextual variations appropriately in paper records – in fact it is so ingrained that it is almost done on autopilot. The problem arises in trying to get computers to do the same thing – to record simply in the usual circumstances but be able to cope with the complex and detailed where it is critically needed.

So, we need to have some understanding about the capabilities of our EHR systems.  Most systems only cater for little beyond the basic four requirements. Yet most of these additional requirements that we have identified are useful or necessary at other times to ensure unambiguous BP recording.

The mind map at left captures the basic versus the 'ideal' requirements. The 'ideal list may seem excessive to some. Yet if our EHR system could cater for this, then we could record all nuances of BP accurately and share them unambiguously if they share the same underlying structure - primary care to secondary care; PHR to EHR to research; patient to clinicians. So then, the 'art of the EHR' is to present the right amount of information and the right context to the clinician to let them record what they need - not too much for the common simple BP reading, but more when necessary.

Here we have started to explore the tip of the iceberg – the content scope of one common clinical measurement. The essential message here is that we clinicians need to have some understanding of what is 'under the bonnet' of our EHR systems. What may be adequate for the commonest situation is not necessarily enough for recording the complexities of clinical care.

Part II starts to explore clinician engagement in content models...

EHRs: the means, not the end

I remember hearing a story about a software demonstration for practice management and billing - a classic where the practice principal proudly stood up and gave testimony about the merits of his new system, including the news that he had to hire extra staff in order to run it, above those who had run the practice beforehand.  Apparently the audience's collective jaws 'hit the floor'.  This is not good eHealth.  Health IT should not add overheads, but make things smarter, quicker, more efficient, and more valuable - or it's probably not worth doing. There is something weird that appears to happen around computers and all things 'e' - sometimes we just seem to lose our analytical skills, or perspective, when sitting in front of a computer. Computers are just tools to support us in our practice of healthcare.

In KevinMD's excellent guest blog in August 2009, "How a wealth of information takes attention away from the patient", Abraham Verghese discusses the tension experienced - look at the data or the patient? With more information available at our fingertips and with only a limited amount of time per patient, how do we prioritize our focus for the best outcomes?  It's not easy as you might first think. We can be torn between the need or desire for (possibly) higher quality, detailed data, rather than the conducting a thorough patient history and exam - after all, time is limited.

When I was in medical school it was always drummed into me that you couldn't compromise a thorough medical history; that a history was possibly way and above more important than even the examination and the secret to good medical practice.  I have always considered that this is a good principle to work by.

If using an electronic health record (EHR) compromises those high standards, then it is time for a re-think.  An EHR should enhance, not hinder.  If it gets in the way of you talking or touching the patient, or causes you to spend more time or effort WITHOUT providing value back then there is something wrong. Note that the value may be direct and immediate (e.g. more efficient; data available at point of care) or indirect and delayed (e.g. better quality data; ability to do query/research, support recalls etc). The benefits should be obvious. Difficulties are usually multifactorial - the clinician; the application; the inability to touch type; or other factors - but they are an important trigger for stopping and investigating the current work processes so that the problems and barriers can be identified and tackled.

Anecdotally, it has surprised me to see some clinicians assume that data displayed on a computer screen has more authority than it warrants, because it is electronic format. They change the way they practice. However, just as in a paper record, because 'nil known' is noted on a computer screen next to 'Known allergies' does not necessarily mean that they don't have any - you still have to ask as the answer may have changed since last asked.  Whether a previous clinician has written in pen or electronically, there is no difference to how we should practise.

Are we compromising our practice? The good old doctor-patient interaction vs. the EHR?

An EHR should not change the way a clinician engages with the patient.

So, it's not rocket science, but don't EHR for EHR's sake. After all, let's face it; an EHR is only useful if it supports quality health care. It is a means to an end, the journey, but not THE end.

Yes, you might arrange the office differently and there might be more opportunity for collaboration once the patient can see their record - that's all good. But if use of an EHR compromises the doctor-patient communication, the recording of data, the use of time in a consultation, or making the history-taking and examination a priority, then the issue needs identification and quick resolution.

Don't lose those first principles that you learned in medical school - they are the foundation of the doctor-patient interaction. The EHR should just be a supportive framework to enable you to do a better job.  If it doesn't, stop and re-think as the EHR should enhance, not hinder.

But you all know that already - this is just a reminder for when we lose a little focus in the excitement...

Standardized, data-driven eHealth

What is an EHR? Most currently working within eHealth circles will have a ready answer, and the classic response will be something like "an EHR is a software application that is used by clinicians for provision of patient care". But think about it for a second...

What do we want from an electronic health record? It's not actually the application or the user interface or the workflow, but a record of health information in an electronic format.

And so what of the health information itself? What is it? What do we want? What should we want?

Let's work with these definitions: Data are the computable facts; while health information is the data after it is processed, organized, structured or presented in a given context so as to make it useful.

We definitely want health information that...

• Is available at the right time and the right place to the right person.
• Is accurate, detailed and of high quality.

I don't think there is much disagreement on these two points. How about these next suggestions?

We should also want health information that...

• Is a lifetime health record - cradle to grave. Health information that can be accumulated over time, aggregated from many sources to inform better care for ourselves, our patients and loved ones.
• Can be created in one place or by one provider and shared with others so that the meaning of the data is preserved and accurate.
• Can be dynamically and actively used - consisting of structured, atomized data that we can re-use and combine in different ways to improve our health and wellness. Documents or PDF's of our health information are useful as part of a passive record, but they are effectively a dead end - we can't do anything useful with big blobs of text except read them.
• Is not locked in to a proprietary vendor database; data should be in an open, non-proprietary format.
• Is a continuum of our health information across all aspects of healthcare and related activities, not fragmented silos of data artificially segmented for personal, clinical or secondary use purposes.
• Can be accessed via an EHR, EMR or PHR, SEHR or ICEHR - pick an acronym, any acronym!
• Can be accessed independent of any specific organizations, providers and geographical locations.
• Has had minimal transformation or manipulation. Keep in mind the Chinese Whispers effect! There is significant risk that important data can be lost or lose integrity with each data migration to a new software application or transformation between disparate systems.

It is a commercial reality that we continue to develop EHR applications in the traditional manner, building the greatest and 'best of breed' clinical software applications, with each EHR vendor doing it in their own proprietary way, as 'rugged individuals'! The resulting software usually has a rich functionality and a great user interface. It is likely that it does a fine job locally in the clinic, hospital or network on which it is installed. But what about regionally, nationally, internationally? Is it still working well if we take the big picture view? Sure, our EHR applications are full of data, but the data is isolated, fragmented, and limited in its use.

There is no doubt that all over the world, the health IT community are finding it unexpectedly hard to exchange health information between different applications - the effort and resources that are being poured into policy, research and pilots for health information exchange is enormous. Progress is glacially slow. If we want to exchange PDFs and documents, then we are doing well, but if we want our software to be able to do more than display text, to do clever things with our health information, then our data requirements are much more complex. If we want that data to be able to be shared, used in multiple applications then the current approach requiring data transformations and messaging paradigms won't be sustainable

Interestingly, the further we go down these paths many are realizing the need to change our emphasis away from the EHR application and towards the data. Back in November 2009, Clay Shirky wrote:

"This ability to separate data from transport and applications from data is the essential pre-condition for innovation — a group that has a valuable new idea for presentation of data for clinical use should not also be forced to think about the data encoding or the way the data are transported. Groups working on new data encodings should not be tied to a pre-existing suite of potential applications, nor should they have to change anything in the transport layer to send the new data out, and so on."

The bolded text, above, reflects my emphasis on an important statement. Confirming this approach, as recently as this week, Kibbe & Klepper also called for separation of the data from the applications and from the transport layer.

Change the focus to standardized, data-driven eHealth

So, innovation requires a new approach to data; a changing emphasis from application- or message-driven to data-driven eHealth. If we also insist on an open and standardized approach to health data specifications then we will be able to realize many additional advantages:

• A strong foundation of shareable and re-usable computable clinical content definitions on which to build coordinated and cohesive applications, messages, clinical decision support programs, and research activities. If we use common, standardized data definitions then the tasks of eHealth become orders of magnitude easier. Content definitions are created and agreed as the content is already specified and the processes become more generic
• An unambiguous and detailed understanding of what each piece of data means so we can do 'stuff' with it - a tight semantic 'handle'.
• A powerful enabler for managing the complex requirements involved in health data capture, integration, aggregation, inferencing and sharing.
• A continuum of our health information, independent of vendor, provider or organization - the real potential for lifelong health records, for the first time.

There is no doubt that this approach is orthogonal to the status quo and it will be challenging to many for logistical, financial and political reasons, but can we really afford to ignore this?

According to the European Commission's seminal 2009 report entitled "Semantic Interoperability for Better Health and Safer Healthcare: Research and Deployment Roadmap For Europe" (PDF, p16), standardizing the capture, representation and communication of clinical data requires three components to represent meaning: a generic reference model for representing clinical data, agreed clinical data structure definitions and clinical terminology systems. Potential standardized data definitions proposed are openEHR archetypes, ISO/EN 13606 Part 2 (which are simpler archetype structures), HL7 templates, generic templates and data sets.  Standardization of data is not 'pie in the sky' but an approach that has had significant research and implementation experience, particularly in Europe.

So, to consumers, clinicians, organizations, researchers and governments, the call should not only be "Gimme my damn data!" but give me standardized data that is application- and message-independent. Then we can actually start to use and re-use our data, not only as a detailed record of current and past health conditions, events and activities, but dynamically and pro-actively to inform and promote our future health and well-being.

Why Is The Shared EHR So Hard?

'Provision of the right data to the right provider at the right time' is the mantra we commonly hear in eHealth. It sounds deceptively simple! Some promise it; some hope for it; but pretty much no-one has it. The collective desire for a seamless and efficient shared electronic health record (EHR) which actually provides that data to the provider at a particular time is widespread and yet the EHR remains elusive. If we know what we want, why is the shared EHR so hard to achieve? Why is it taking so long? And what's more, if the financial sector can do it, why not health?

The progress that we have made in the past 10-20 years of eHealth development has been glacially slow compared to other industries and domains. The approach to health IT, and in particular the shared EHR, has been primarily linear in nature with modest incremental successes achieved. Sure, progress has definitely been made, but despite investing enormous amounts of money and resources, the solution has been more difficult than most ever anticipated. Healthcare doesn't appear to fit the same data interoperability model that has been successful in other domains such as banking or financial services. In a world where connectivity reigns and personal data can flow freely, it is not yet the norm for our health information to be connected nor to flow!

In trying to understand the problem more fully, there are 3 broad headings we need to explore:

1. Technical
2. Human interaction and activities
3. Information exchange

Let's explore some of these issues:

1. Technical

There is little doubt that traditional EHR development has been driven by technology requirements and engineering processes, so many of the typical aspects we consider when thinking of the EHR are actually quite manageable, if not solved - consider storage and retrieval of data, security, role-based access, audit trails, repository storage etc. It is not the technology holding us back here. Some even suggest that the technical aspects of the standalone EHR are the (relatively) easy part, and they may well be right!

With over 7000 EHR vendors in the United States alone, we have proof that building a standalone EHR is undoubtedly achievable. Our EHRs that are in use are traditionally rugged individuals, each created in splendid isolation with the finest data structure, processes and user interfaces! Unfortunately, building proprietary silos of health information is the almost universal approach to EHR development adopted by vendors and does not make it easy for health data to be shared.

I have seen and heard some say that we already have all the standards that we need for eHealth. This is still somewhat controversial and perhaps depends on what part of eHealth that you are trying to make work. For a coordinated eHealth system the desire is for each standards to not only achieve its purpose and goal, but to harmonize with all the others in existence to create a unified whole. Can this vision be achieved? There are certainly some examples of very successful standards.  There are also some examples of tweaking of standards for local use... which makes them non-standard again. There are also some examples of standards that have never yet been implemented... what? Let me refer you to the blog post from my colleague - The Crisis in eHealth Standards, by software engineer, Thomas Beale - for an erudite discourse on the strengths and weakness of standards and the standards processes.

2. Human interaction and activities

There is no doubt that some EHR technological achievements have been delayed or even diverted by the human, political and regulatory issues arising from practical implementation - and in most instances, rightly so.  We can develop a technological solution, the 'what', but in many jurisdictions the 'how' still has everyone tied up in knots.  For example we have technical solutions for unique patient identifiers, data security, and role-based access to data - but 'how' to apply these in practice is more elusive, often crossing into moral and ethical arguments (including confidentiality and privacy), requiring broad social and political agreement and sometimes supportive legislation. These issues have received a disproportionate amount of attention, particularly in the media, and perhaps at the expense of issues that follow which are not so well understood, yet!

Healthcare is not primarily a technical business - it is about people. Yet in our rush towards EHR development we seem to have focused on the EHR being a technical solution rather than merely a tool or medium to support the practical application of health knowledge and provision of healthcare to real people.  Some of the most underestimated and misunderstood problems in EHR development are related to:

• the scope and dynamic nature of our health knowledge domain;
• the nature of the human-human interaction that underpins quality health care;
• the approach to capturing and storing the essence of a clinical encounter; and
• the changing clinical requirements, processes and approaches to healthcare provision.

None of these are trivial concepts.  They are in the metaphorical 'EHR too hard basket', but we need to address them...

Health is possibly (more likely, probably) the most complex knowledge domain.  The extent and scope of health-related knowledge that needs to be represented in an EHR is enormous - it has huge breadth and depth, plus a fine web of complex relationships.  Most commonly underestimated is that health knowledge is dynamic - requirements distilled from a clinician and built into an EHR application by a software engineer today can be out-of-date by the time the product is launched.

An encounter between a clinician and a patient is a very complex pas de deux, an intricate communication dance between two parties.  Interestingly we clinicians have developed surprisingly effective written methods to capture the information exchange from these encounters, with all the required subtleties and nuances. Capturing the essence of this encounter into a format that can be stored, re-used, queried and shared on a computer is a daunting task and more complex than first appears. Attempting to represent both the data and our clinical processes via the user interface on a computer screen is definitely also an advanced task.

The nature of healthcare provision is also in flux. Consumers/patients/clients/citizens/individuals are increasingly mobile, more now than ever before, demanding healthcare from a range of providers in varying geographical locations.  Gone is the old-fashioned notion of the local family physician providing all our needs for all generations of the family; the local physicians are morphing into our health coordinators and facilitators in a world of collaborative and distributed models of care. Our ability to diagnose, treat and prevent conditions are moving with the assistance of new technological advances - in particular, watch out for the potential tsunami-like impact of personalized medicine/genomics on healthcare in the next few years.

3. Information exchange

There are many differing approaches to sharing health information. Why? The plain answer is that it is hard and there is no clear way forward; there are also many differing requirements and starting points:

• Logically, the technical task of sharing health information would be easier if the data was in a common format. Conversely sharing seems to become more difficult by orders of magnitude if our collective data structure is in chaos.
• Practically, some require only the simple exchange of a readable document such as a PDF or semi-structured documents. At the opposite end of the scale, there is need for EHR systems to share detailed and structured health information, so that not only can clinicians and patients read it but the meaning of each piece of data is clearly understood and it can be directly integrated into the computer and utilized in clinical decision-making - this is known as semantic interoperability.

In many places, national eHealth programs and the myriad of 'ruggedly individual' EHR vendors are pursuing technical approaches to data exchange through the set-up of information exchange hubs, message mapping and data transformations.  In this instance the focus is on exchange of complete or semi-structured documents as readable health information. Europe and some other parts of the world are taking a different approach, focusing on the exchange of standardized, atomic data  and reflected in the European decision to adopt ISO/EN 13606 as its standard for EHR extract exchange.

Exchange of unstructured health-related documents as 'blobs' of data is a great starting point, but in the long term it is a dead end. In reality it is not a sound basis for a shared EHR, nor interoperability, as we can only read them and then store them passively within the EHR. Personally, I want a dynamic, lifelong EHR for myself and my patients - one that can actively create, receive, integrate and re-use my atomic health information and put it to work for me to improve my current health and to provide the basis for future health decisions.

'Provision of the right data to the right provider at the right time' - can this actually be achieved? Clearly the relatively safe, comfortable and incremental technical innovation that has underpinned our EHR development to date hasn't provided the shared EHR solution we hoped for. So at this point let us draw from the wisdom of Einstein: “We can't solve problems by using the same kind of thinking we used when we created them." Indeed, perhaps it is time for a different approach to the shared EHR; one in which we divert our focus and are encouraged to push boundaries; to seek transformational change in our approach to eHealth.

In future posts, I hope to explore some of these issues in more depth and, in turn, some opportunities that arise...

Acknowledgments: Dr Sam Heard and Dr Hugh Leslie