Pathology Harmonisation Roadmap
In our previous article, we discussed what `harmonisation` of Pathology (Laboratory Medicine) is and how it enables cross-site working, data interoperability for networks/regions/hubs. As abrief recap, here is an extract:
- Harmonisation of Pathology data covers more than just codes and includes specimens, units, reference ranges, methods, analysers, etc.
- Harmonisation aims to enable cross-site working and cross-site data interoperability in a network/region/hub/cluster.
- Harmonisation aims to ensure that legitimate variations in lab test processes and data are accurately recorded, so they can be addressed if legitimate clinical safety concerns can be addressed.
In this article, let us look at a roadmap approach for achieving clinical excellence and interoperability applying the principles of harmonisation.
Why a Pathology Harmonisation Roadmap
We sometimes see people aim/share that they’ll go `big bang` and `one fine day somehow (magically)` that the entire network/region will wake up to fully standardised pathology data. While we wish them luck, our past experiences of large programme implementations (e.g. Singapore national EHR, NHS Connecting for Health) show that getting data, systems, processes and people across multiple sites/regions/hospitals to `converge/harmonise` is a complex challenge. I am borrowing the phrase `managed convergence` from Bill Bartlett who spent a long time thinking about similar things.
Just because one has managed to successfully roll out a LIMS in your hospital/site, does not mean the same approach will work across a region/network. Using the same approach introduces the risk of having a single LIMS in name only, with each instance being just silos and replicating the past.
So if harmonisation is the objective, then the route to get there has to be in stages, so every organisation can determine for themselves how far they want to go and what the next stage is. We also made attempts to align this using a `digital maturity` approach (e.g. HIMMS EMRAM), so there is a coherent story for how this would fit alongside your existing plans for digital maturity. More importantly, this also allows leaders of digitally mature organisations to understand where they are in the journey instead of thinking they need to start from scratch.
Enough talk! Here is the roadmap at the highest level — with the high level stages:
Stages of Pathology Harmonisation
We believe that there are 4 principal stages in the road towards achieving `operational clinical excellence in near real time`.
Stage I — Single Test Catalogue with Governance
Everyone who has worked in the lab space understands the notion of a `test catalogue` (effectively the list of tests being offered/performed). However, as you remember from the last article, there is more to a `lab test` than just `names`. So while it is easy to knock up a list test for your lab/site, deriving a single test catalogue across your `network/region/chain` is a more complex challenge.
- It requires the specialists to agree that two tests are the same thing and can be comparable (flash back to the previous article).
- It requires a way for specialists to determine/agree by themselves what a `test means` and to use a consensus based `governance` to approve/sign-off the test.
The challenge with Pathology (Lab Medicine) is that there are multiple sub specialities (for the uninitiated — the Royal College of Pathologists (RCPath) UK has 17 sub specialities listed!). You will be hard pressed to find an expert in one speciality willing to sign-off/approve something that isn’t their area of specialisation.
So a `single test catalogue` that works across multiple sites and specialities requires clinical oversight/governance to create and maintain. If there is no easy way to maintain it, then there is a risk it’ll be outdated after a region/network spent 3–6 months creating it.
Stage II — Lab Test Definition
Stage II entails `shared visibility of a lab test` across the region/network, in the way lab specialists refer to as a `lab test`. Without getting philosophical, understanding what a `lab test` is a tricky question. Effectively, different layers/views get applied on what constitutes a `lab test` — SNOMED CT Procedures + Observables, LOINC model, NPU model (where units are a defining part of a test), etc. However, ask a lab specialist that they want to know/say/remember about a test and they’ll given you a long list of things:
I will refer to my first mentor Alan Rector who used to say — there are layers of views that can be applied to a `thing` in an ontology`. So please, ontology/standards folks I am not your enemy and I am not saying standards/ontologies are broken and won’t work… 😬🙈 There is a really important aspect of Stage 2, which we will cover in a separate article. They are what we call `Order Sets`. Some of you might recognise that attributes in the table seem to span the `request` and `result` dichotomy! That is because a philosophical `lab test` actually spans this life cycle and introduces complex notions like `requests that lead to multiple results`, `reflex tests`, etc. This is why we refer to `Order Sets` as this thing that needs dealing with separately. So watch the space!
Just a quick reminder if you are not from the lab world -> every lab has specified standard operating procedures and ISO standards they follow to ensure all this information is recorded. So they record them, just not in the way they are exposed beyond their own lab/site. Hence labs are effectively some of the `most standardised` hubs in healthcare except often they are `silos` from the data perspective! But here is the real story…
If we do not have a way of recording all this additional information about a `lab test`, how do we know if two test results are comparable?
Just because any single current ontology does not cater for what `lab test` means does not mean that we should not harmonise them. What is needed is to combine available standards in an effective way (with informatics principles) to ensure that lab tests can be defined by lab specialists to cover all this information. The idea is not to just `standardise` them, but to allow recording this information across sites so an `effective single place` is around for information about any `single lab test` is clearly visible across all sites. Some of this information is available in labs locally as reference material — what stage 2 entails is a `shared visibility of a lab test` across the region/network. We refer to this as the `lab portal` — a single reference for lab specialists and clinicians for discovery and interpretation of lab tests.
This translates into a few immediate things from the clinical safety and efficiency perspective:
- Identify similarities and differences across sites in what is being done for a test
- Identify legitimate variations between sites and identify the barriers or enablers for cross-site harmonisation
- Identify tests that are safe to compare and others that might need some work to be comparable
- In the future, the lab portal can then be turned into something a lot more reusable — a public facing information portal, a ready reference for non-lab clinicians (e.g. a junior doctor/nurse), etc.
Of course, this also allows visibility of tests that can never be compared across those sites. This ensures that naive charts/graphs like we discussed in the last article (plot it and hope for the best) do not introduce clinical safety risk! Please note that different lab specialities have different priorities for what additional information about a lab test is important.
Stage III — Quality Assurance
This stage is characterised by operational processes that reuse the information from the single test catalogue and the lab portal to automatically track the activity in a lab/network/region. In the lab space, this notion is referred to as `clinical dashboards` — deriving operational visibility across data. We envision this stage as beyond `bean counting` xxx blood science tests were done, yyyy immunology tests were reported on, zzzz tests were unreportable because the specimen was degraded. This stage is operational intelligence and assurance using a combination of:
- Network wide Single Test Catalogue
- Lab Portal based test definitions across sites and per-site
- Order Sets in the Lab Portal
- HL7 FHIR analytics spanning the request and result cycle
It is the combination of this information to answer questions like:
- How many Total Cholesterol Measurements were performed in the last 3 months (easy stuff)?
- How many Viral Serological Studies were performed in the last 3 months (easy stuff)?
- How many Vitamin D3 tests were referred to `site B` from `site A` in the last 3 months (easy stuff)?
- How many LFTs were done at site A vs site B in the last 3 months (easy stuff)?
- Why are outcomes for our patients who turn with dyspnoea (breathlessness) in A&E worse than those that turn up say in Pulmonology? (is it because our Order Sets were different, among other things?)
- Why does an LFT in site A cost us xxxx and site B cost us zzzzz (is it because our LFT OrderSet for site A has 5 tests vs 6 in site B)?
- Across my network/region across all sites, what is the standard deviation of `Erythrocyte riboflavin` done in the last 3 months? (Is this a legitimate variation due to different kits/processes?)
As you can probably guess, some of these questions can be easily answered currently using LIMS but there are those that require pathology harmonisation to have been done to at least Stage 2. You need your clinical governance from Stage 1 to determine if your `erythrocyte riboflavin` tests are the same and the level of detail about the kit/units/etc from Stage 2 to answer such questions.
Informatics aware audiences would have spotted the reference to HL7 FHIR in this section, which is a deliberate attempt to align this with other digitisation strategies that are currently on-going in healthcare. Some aspects of such operational intelligence are also enabled by the use of terminology servers to support analytics. This is not to say that you could not do this using HL7 V2, but rather showcasing the future state where HL7 FHIR might be where the organisation is headed. Our roadmap also aligns with broader national standards like the use of SNOMED CT, LOINC, etc. Of course the recent collaboration agreement between SNOMED International and Regenstrief Institute, Inc. works well for this.
Stage IV — Continuous Learning
There has been enough said recently about `learning systems` and `artificial intelligence` that this section does not need extensive explanation. However, it is worth mentioning that Stage IV harmonisation implicitly assumes that there is an element of automation (of data streams/flows) that is built on top of the systems/processes in Stage III. Again, the clever folks who are already here already know how such things would work. But here are some examples of the sort of AI related use cases — which we think mostly falls into `sentinel intelligence`.
- Automated triggers for when average reference ranges for say `alkaline phosphatase (ALP`) from a Site A are over the ranges seen in the past month.
- Automated notification when a new serotype of E. coli is being reported from a lab that previously never reported it.
- Automated notification when susceptibility patterns for Neisseria gonorrhoeae in a particular geographical region of my network (e.g. a location in the catchment area) are different to what was previously known.
If you are already doing this, that’s great, but if you wonder why this is not being done, then the answer is because Pathology (Lab medicine) data has not been harmonised properly!
Again, you can imagine how automation and AI built on top of everything in Stage III can now signal such changes for visibility to the Chief Information Officer (CIO) or Chief Clinical Information Officer (CCIO) to be a lot more proactive in emergence and distribution of undesirable health trends. For example, we have all been told many times that Antimicrobial resistance (AMR) could go on to be the single largest contributor to mortality and morbidity by 2050!
Achieving Pathology Harmonisation
It is often difficult to imagine this entire vision and hold it focus when going through the routine processes of reporting on tests or even if we are involved in say something like LIMS requirements gathering. However, as discussed in the last article, what is currently happening in the landscape of LIMS procurement could present a once in a generation opportunity to get things in line for delivering such a vision.
It is easy for us to say just use `Pathnexus — our lab data harmonisation platform` and you are done. However, that would be only half the story. The route is not through Pathnexus technology alone, but ensuring that people, processes and systems are all brought together through leadership. We know leadership in a resource constrained and stressful space like Pathology/lab-medicine is not easy, which is why Pathnexus is designed to reduce the pain in the journey. We will cover aspects of leadership in future articles, but for now let us cover what Pathnexus provides out of the box to achieve this vision of Pathology harmonisation.
Stage I — Single Test Catalogue
You can today use Pathnexus and achieve a single test catalogue for your network, with all the necessary clinical governance and oversight. It won’t replace the need for you to establish your governance structure, but it will go a long way in reducing the effort involved in consensus generation, approval and governance of lists across sites and specialities. If you are a new network and wondering how to get started, Pathnexus derives your single test catalogue for you while each of your sites focusses on just their `test catalogue`!
Here is what the relevant functionality in Pathnexus looks like:
Stage II — Lab Portal
Over the past few years, we have wrestled with the notion of `lab test definition` and have created a ‘lab portal` for our users in Pathnexus. It allows lab specialists to capture all information about a given lab test catering for variations in sites, specialities, etc. You can pull in any existing information you might have about tests like units, reference ranges, specimens and standardise them (yes thats right) and even go through a governance process to identify legitimate variations, share a single reference across your region/network/chain. The information you can enter can also be guidance, etc and the good news is that it reuses all the information from Stage 1 and so you’ll clearly see how your `single test catalogue` ties back to the per site test, external reference standards like SNOMED CT, LOINC, etc.
Here is what some of the functionality looks like in Pathnexus:
We have deliberately left out information about Stage III and IV functionality here since this article would then read like an advertisement, which isn’t the point.
If you are trying to understand how your organisation can adopt the Pathology Harmonisation roadmap and have questions about it, please reach out. It is okay, even if you do not want to use Pathnexus and would prefer doing it in your favourite tool (e.g. hopefully not Excel 🙈). We are passionate and excited about the possibilities that NHS England and NHS Digital NHS England network consolidation activities and the US SHIELD initative open up. We have conversations with organisations with different levels of digital maturity who want to achieve a similar vision, so we are open to conversations and feedback. We would love to learn more about your own vision, priorities, challenges and see if there are ways we can help you achieve that vision.