I HAVE ATTENDED TWO CONFERENCES RECENTLY and while attending lectures with a theme of evidence-based medicine, I was pleased to hear from the lecturers that there is now an interest in hearing from those of us who are in general practice about our treatments and results.
As we generally see more cases of any condition than those routinely writing the papers and textbooks, it would be quite interesting for all concerned to try to harvest some of this information and see what is going on in the wider world.
This is especially true for conditions routinely treated at first opinion level that either rarely make it to referral or of which only a small percentage of cases do and these skew the populations studied by those in that environment.
Using data from GP practice would also help with veterinary science’s problem with small data sets. Even the best clinical trials and studies in the veterinary world manage only paltry numbers compared to the medical field. The “Quest” pimobendan study in dogs is a large study by our standards and still only looked at 260 dogs, the size of a small pilot study in human medicine.
I wrote about this a few months ago and asked how the veterinary world can make use of the masses of data generated by all of us on a daily basis and pass it all on to people who know how to interpret it. This would help improve the evidence base for what is termed “Evidence Based Veterinary Medicine” (EBVM).
A week after going to print I received (metaphorically via e-mail) a gentle tap on the shoulder, a polite cough and Dan O’Neil telling me: “Actually there is something that does all that. Ask me about the VetCompass project.” So I did, and I have reproduced some of our Q&As below. I began by asking what started it off.
Dan O’Neil: After having spent 22 years in practice and becoming so frustrated with the lack of real EBVM that was generated based on primary-care caseloads…I left practice to help develop VetCompass at the RVC. From a species perspective, these cover topics on dogs, cats and rabbits. Topics cover RTA, spinal diseases, undesirable behaviours and breed-based disorders. And these all describe what is happening to animals attending primary day-care practices and also primary emergency-care practice.
GC: The system uses standardised veterinary terms, which they call VeNoms (see http://www.venomcoding.org/VeNom/Welcome.html). I asked if these are input during clinical note writing or as a separate exercise later on.
DO’N: Both options are available.
- Practitioners can (but don’t have to) input terms using the VeNom terms (species, breed, procedures, presentations, diagnoses) during episodes of clinical care.
- We can retrospectively apply VeNom terms to vet data during research. This can be done manually (by reading records) or automatically (using Natural Language Processing).
We have just completed our automated upload process that will make it simple for PMSs across the world to participate. So we are now aiming to get every UK PMS signed up ASAP.
GC: It’s not clear to me where the info goes when it leaves the practice and who organises it.
DO’N: The data are uploaded to a secure server at the RVC. From there, the data are processed to a standard VetCompass format and loaded into the VetCompass database based also on a secure server at the RVC.
GC: Is it stored as a sort of meta- analysis PMS database for RVC researchers to dip into?
DO’N: In essence, yes. Data from all the PMSs are stored as a standard format and so are available for merged-PMS studies.
GC: Is there any computerised data analysis of the raw data based on the VeNoms?
DO’N: These can be automatically analysed. The richness of the data is really based on the free-text analyses. These are explored using combinations of NLP and then manual data extraction.
GC: How can GP vets benefit from this and how exactly do they access information from it?
DO’N: Participating practitioners are given direct access to their own practice’s data in a cleaned format and with the freedom to use the VetCompass research tools. So practitioners can complete their own clinical audits or pieces of research.
GC: That’s exciting. How can they join in supplying data?
DO’N: Provided their PMS is compliant with the VetCompass data sharing systems, it is as simple as contacting VetCompass and agreeing to sharing their data. The practices are provided with literature for their clients and can add information on participation to their website.
GC: So to look at how an example might work in the future: take, at random, diarrhoea at first presentation. Never seen in referral practice and treated in various different ways in practice. How would an investigation into this work? Would someone at the RVC be able to type “diarrhoea” into the database search and read the notes and outcomes and work out some statistics?
DO’N: Yes, it’s as simple as that to search but it can take quite a while to do the study depending on how many cases the researcher wants to read.
GC: Are all the notes copied over?
DO’N: Yes, we get all notes included; records of telephone conversations, etc. And all treatment data including diets, etc.
GC: How do you adjust for missed diagnosis, wrong diagnosis, etc.?
DO’N: We work on the final diagnosis so if there is a wrong diagnosis early on that is later revised, we take the latest one. We report on what the vets diagnose, not on whether the diagnosis is correct or not. We believe that vets are experts at what they deal with day-to-day.
GC: Can you look from the other side too, e.g. presenting sign? Could you say “let’s look at all the vomiting dogs that come in for 1,000 cases and see what they ended up being diagnosed with?”
DO’N: Yes, we can start at whatever point in the diagnosis process that we want. Starting with vomiting dogs and working forwards is easy.
GC: I remember reading in Ben Goldacre’s book about the paucity of evidence-based decisions in human medical practice. Do you have a figure for the veterinary profession?
DO’N: That is a good question re the sound clinical evidence: sadly, I don’t think we really even have the evidence on how many disorders/decisions/diagnoses have sound clinical evidence behind them. I guess this all comes down to what we accept as sound evidence: is expert opinion enough? If so, there is plenty of this about, and it often differs even between the experts so perhaps it should count as double or treble then.
If it refers to actual published evidence that relates to primary care populations and typical primary care decisions, then the reality is that the percentage is likely to be very low. In fact, it was the lack of evidence of prevalence data of common disorders in dogs and cats that triggered my PhD back in 2010 and that led to the development of VetCompass.
So if we did not even have evidence then on what was the most common disorder recorded in dog and cats, then we probably also do not have evidence on the vast majority of other clinical decisions that vets make each day.
It was this drive to plug these data gaps that was one of my major drivers to leave practice and work on primary care EBVM generation.
GC: As well as data going up to you, how can practices immediately benefit?
DO’N: Every participating practice will be given access to their own practice data (not to any others) which they can use for their own audit or research purposes. We are aiming to promote both audit and also practice-based research. This facility is provided at no charge to participating practices and is just part of the reason why so many practices are joining VetCompass.
The RCVS does promote this data-sharing process in order to support EBVM as part of the PSS:
2.5.12 The practice is contributing data towards professional benchmarking or clinical data collection, or data for future potential publication. Sharing of information to facilitate research and/or improve best practice. This could include contributing data towards undergraduate projects or clinical data to organised multi centre studies for potential publication (e.g. VetCompass or SAVSNET). 40 TOTAL POINTS AVAILABLE.
GC: EBVM seems to me to be too much of a bit of jargon to describe this project. It surely should just be the way we do things now that sharing data is so easy. We can feed up what we do as GPs or in referral practice to the epidemiologists for processing. They can look at what works and what doesn’t.
It also helps us get over the lack of big data in veterinary science.
A classic example for big data in medicine is steroid use in head trauma in humans – widely used and logical until a meta- analysis of trials and clinical case reports showed that in fact mortality is increased by using it. No single trial had convincingly shown that, but looking at a lot of trials together it was clear that lives were being lost by using what was a widely accepted and logical treatment.
Who knows what will be uncovered when we start looking at bigger data sets in veterinary practice… Maybe they will help bust some myths too, like finding out if you really can’t give ACP to boxers, or should we actually be treating feline idiopathic cystitis with antibiotics, which cruciate op really is the best…?
I asked Dan to end with a “call to arms” to get more people involved.
DO’N: Today’s veterinary practitioners dedicate themselves to improving the lives of those animals directly under their care. However, clinical decisions are often based on sketchy evidence such as poorly representative referral studies, expert opinion or anecdotes. With the development of VetCompass data collection methods that have zero impact on consulting room work flow patterns, UK practitioners now have an opportunity to be part of the long-term generation of clinical evidence that relates directly to their own primary care caseloads and management.
Such evidence may challenge many of the routine decisions that we make each day and offers the possibilities to dramatically increase our clinical effectiveness and to improve animal welfare.
By participating in VetCompass, today’s veterinary practitioners can now dedicate themselves to improving not just the lives of those animals directly under their care but to also helping all other animals in the UK both today and far in the future.
To participate, practices simply agree to display posters and leaflets that promote their involvement in this welfare-friendly project to their clients. Other than that, there are no other requirements for practices: de-identified data are simply shared directly from the practice management systems providers to a secure database at the RVC.
Participating practices are provided with online access to their own data that have been cleaned and prepared for research and which can be used for their own clinical audit or other practice-based research.
Further information can be found at http://www.rvc.ac.uk/vetcompass. To express an interest in participating, please e-mail email@example.com. Depending on which PMS provider the practice uses, many practices can immediately get involved while others may have to wait a little time until their PMS provider catches up with VetCompass compliance.