Clinical reasoning models: what’s wrong with them?

I’ve been interested in clinical reasoning and models used in clinical reasoning for quite some time. Occupational therapy has several models, including the “occupational therapy problem solving process” by Lela Llorens, the Model of Human Occupation by Gary Kielhofner, and the Canadian Model of Occupational Performance by Polatajko, Townsend and Craik in 2007. All of these models were designed to support occupational therapy clinical reasoning processes, and to capture the essence of what occupational therapy is about.

When it comes to pain rehabilitation, I’ve found the occupational therapy models a little lacking in specificity for my clinical reasoning. I’ve also noticed similar problems with proposed clinical reasoning models for physiotherapy when considering pain.

Here’s the thing: if pain involves so many factors (call them biopsychosocial for want of a better all-encompassing term), and we don’t know which factors are relevant for this person at this time, clinical reasoning in pain rehabilitation is complex. Why? Well the problem with pain is that it’s full of ambiguity. Not so much for the person experiencing them, but certainly for the clinician trying to help.

Bear with me a minute. To me, clinical reasoning models help shape the factors we include and those we omit.

In writing that sentence I realise I’m assuming something crucial: that models are designed to help us predict and control what’s going on. Is that the purpose of a model? I quickly did a search and found this definition: “In science, a model is a representation of an idea, an object or even a process or a system that is used to describe and explain phenomena that cannot be experienced directly. Models are central to what scientists do, both in their research as well as when communicating their explanations… Models are a mentally visual way of linking theory with experiment, and they guide research by being simplified representations of an imagined reality that enable predictions to be developed and tested by experiment.” It’s from here.

OK, so in clinical reasoning what utility does a model need? I think a model needs to generate hypotheses that explain the unique presentation of this person, their problems, at this time. A nomothetic representation of what might be going on for this unique person.

Occupational therapists and physiotherapists, and probably psychologists, are all concerned less about impairment (that’s damage or dysfunction at the body structure level) than we are about the impact this has on functional limitations and on participation. This doesn’t mean we’re not interested in impairment, but our focus is much more likely to be on “and what impact does that have on what you need and want to do”. Occupational therapists, in particular, are concerned about “and how does this affect the way you participate in our world”.

But if we look at clinical reasoning models in our various professions I think there are some gaps. I don’t think our models invite us to generate hypotheses because the various clusters of information don’t seem to link together in a terribly coherent way. Yet – with all the information around us, there are some causal (or bidirectional) relationships we can consider.

For example, we know that if someone is very fearful of their pain, they’re likely to describe elevated physiological arousal, and they’re not as inclined to engage in movements they believe will exacerbate their pain.

A line of reasoning goes from Fear -> Physiological arousal and Fear -> Avoidance.

This simple set of hypotheses generates some ideas about what might help. Firstly we’d test the presence of fear – is it just happening in this moment, or is it something that’s been present consistently? Mostly we ask the person, but we could use a questionnaire measure of fear of pain. We could also test for physiological arousal – is this present? How do we know? We could use various biofeedback devices, or we could simply ask (or use a questionnaire). And of course we can test for fear-avoidance as a combined construct via questionnaire and/or behavioural testing.

This set of steps really just determines whether our hypotheses are present, so now we need to generate some treatments. In this case, we also draw on research and think about providing information – this, we hypothesise, should reduce reported fear. So we embark on some explanations about what’s going on – and we should see a reduction of fear on a measure of pain-related fear. But perhaps not on avoidance because we know that behaviour change requires more than simply information. We might also help the person down-regulate their excitable nervous system, reducing that “fear -> arousal” relationship. And finally we might begin doing some exposure work which acts on reducing fear in the presence of doing something scary (movements) and so reduce the relationship between fear -> avoidance.

What the example above shows us is what might happen once we’ve identified some potential phenomena that may be present. What it doesn’t show, and something I struggle to find in many clinical reasoning models, is how clinicians identify those phenomena. Why would someone think to ask about fear of pain? Especially if we believe that our job is to help reduce pain and pain’s the only reason the person isn’t doing things. And even more – if we think our job is to deal with “physical” and fail to recognise the relationship between “physical” and “feelings, beliefs”.

You see, I think broad “groups of factors to consider” belongs in the assessment, but we need something more tangible when formulating an individualised explanation. We need to be generating hypotheses about how these various factors interact and lead to a presentation – and while much of this will be conjecture initially, by generating various hypotheses we can then go on to test them – and ultimately establish the priorities for treatment in collaboration with the person. That’s much easier to do when we’ve fleshed out why the person isn’t able to do what’s important to them, and we’ve synthesised all the known factors in some explanatory model.

Is this complex? Yes – but who said it had to be easy? This is why we do the work we do, because it’s complex and “common sense” doesn’t cut it. And if our various professions really want to adopt a sociopsychobiological framework for pain, maybe our clinical reasoning models need to synthesise all these factors in some coherent way rather than simply plonking the groups of factors down without integrating what’s known about the relationships between variables from different domains.

Wacker, J. G. (1998). A definition of theory: research guidelines for different theory-building research methods in operations management. Journal of Operations Management, 16(4), 361-385.

Yazdani, S., Hosseinzadeh, M., & Hosseini, F. (2017). Models of clinical reasoning with a focus on general practice: A critical review. Journal of advances in medical education & professionalism, 5(4), 177-184.

Revelation: I’m experimenting on patients!!

Actually, the heading should read ‘I’m experimenting on with patients!

Does that not help?  Sorry, perhaps I should unpack what I mean!

Chronic pain, or actually, chronic disability associated with pain, is multifactorial.  What that means is there are many different factors that influence how and why a person has chronic pain and disability.  It also means that each person is likely to have a different set of factors that is contributing to why they are having this set of problems in this specific situation.

And the implications of this are that it’s highly unlikely that any one single treatment will ‘work’ to ‘fix’ the problem! In fact, the only time we can be certain about our treatments is when the following conditions are met:

  • a reliable and valid diagnosis
  • with a well-defined explanation for the cause of the pain
  • and known patient selection criteria
  • that predict a positive response to treatment
  • with known mechanisms of response

This doesn’t happen often, especially with chronic low back pain – and as a result, we’re probably using a working hypothesis when we’re choosing a treatment.  And guess what? That’s exactly what an experiment is – following a systematic process to establish whether the results support a specific hypothesis.

Oooops.  Are you guilty too? (more…)

What should I include in my pain assessment?

With such a wide array of factors influencing a person’s pain experience, it can be difficult to decide exactly what to include in a pain assessment.

We do know that the model we use to view pain will influence the factors that are included – and although the internationally accepted model of pain is a biopsychosocial one, there are any number of versions of this model that can be adopted.

Within each domain of the biopsychosocial model the research over the past few years has exploded, meaning there are more and more factors than can be considered – and these need to be organised in a systematic way so that we can make sense of them, make good clinical decisions about interventions and then work with the person who has the pain so they can understand them and contribute.

There are a couple of fundamental things we should always have as guiding principles:

  1. No single element in the biopsychosocial model of pain is more (or less) important than any other
  2. All three domains must be assessed to fully understand the ‘four p’s’ of a pain presentation:
    1. Predisposing factors
    2. Precipitating factors
    3. Perpetuating factors
    4. Protective factors
  3. The fundamental questions to be answered through assessment are
    1. ‘What brought this person to this place with this problem today?’
    2. ‘What can be done to reduce distress and disability?’
  4. Simply asking the person with pain provides some good information, but on its own is probably inadequate.  Interviews need to be supplemented with:
    1. History – from relevant documentation (from the referrer, other health care notes, previous consultations within your facility)
    2. Observation – structured or unstructured observation from the moment the person enters your clinic, to the time they leave
    3. Clinical examination or testing – including functional performance as well as pen and paper questionnaires
    4. Other people – particularly partners or other family members
  5. Assessment only begins the process of developing a working set of hypotheses about what might be ‘true’ for this person at this time for these problems

A couple of models that can be helpful:

This one is from Robert Gatchel (Gatchel, 2004).

, American Psychologist, 59, 792–805.

Another model I like is by Tim Sharp, published in 2000, which is somewhat less complicated than Gatchel’s one, but still has a whole lot of arrows!  Dr Sharp now runs a successful consulting practice listed in my blogroll, worth a look!

Of course, no matter what model you use, under each ‘heading’ you will need to continue to update relevant research into specific factors to include (eg ‘appraisals’ would now routinely include catastrophising and pain-related anxiety, while ‘motor behaviours’ would include avoidance, safety behaviours, as well as task persistence).  And after deciding what to include, it will be just as important to determine the best way to access the information – through questionnaire, observation, history, testing or interview.

Finally, it will be important to work out a structured way to put the information collected together so it can be readily understood and used as the basis for hypothesis testing.

I’m not sure I’ve got a handle on this part yet – but I’m keen to hear what you use, or how you think this part can be structured.  I think we’ll have to draw on research from small group/teamwork literature into decision-making, and on human cognition and information processing to inform us on the best way to integrate such complex information without jumping to conclusions.

Isn’t it great the way that answering one question leads to a whole lot of new bits of research?  Can ya tell how much I love questions?!

If you’ve enjoyed reading this post, and want to keep up to date with my blog, you can click on the ‘RSS’ button at the top of the page to subscribe to updated feeds (I post most days during the working week), or you can simply bookmark this page, and come back every now and then.

I love comments – so please, keep ’em coming!  I do reply usually – oh and you can get hold of me using my email, just head over to my ‘about’ page and you’ll find it there.

(One day I’ll work out how to do a form so you can simply complete a form and it will send it to me, but that’s for another day!)

Gatchel, R. (2004). American Psychologist, 59, 792-805

Sharp, T. (2000). Chronic pain: A reformulation of the cognitive-behavioural model.  Behaviour Research and Therapy, 39, 787–800

Gatchel, R. J.(2004), American Psychologist, 59, 792–805

On the challenges of ‘doing research’

As a clinician and an educator I need to be aware of, and ‘do’ research. It’s not easy, even when I’m enrolled in a PhD programme! The problem is that so much ‘research’ is not finding out new stuff, it’s gathering existing information into a format that allows for ‘gaps’ in knowledge to be identified and then prioritised. And this isn’t sexy, isn’t visible, doesn’t result in publications, and is often undervalued. A bit like teaching/educating itself!

I don’t know what it’s like in the rest of the world, but here in NZ there are an awful lot of hoops to jump through to even begin collecting data, and this is pretty off-putting. And much of our research has to be based on short-term, fixed end-point research that can tie into the duration of a funding contract. With the thrust for research to be able to directly contribute to something pragmatic, the emphasis isn’t on philosophical or even basic science research, it’s on ‘what is happening now and what can be done about it’. Short, sharp, and applied.

So all the background, really integral parts of research are essentially invisible.

In putting together a PhD proposal, I feel like I’m breaking some new ground. And a whole lot of what I’m doing just isn’t going to be visible in my completed thesis – yet it’s incredibly important. I work primarily amongst scientists who like the hypothetico-deductive model – and I’m certainly not against that in any way. I just think there is a whole lot that happens before a hypothesis can be generated that is absolutely essential, but ignored. And this isn’t easy for scientists of the H-D persuasion. What do I mean?

Well – where did any hypothesis come from?

If we’re honest, it probably came from some sort of observation where someone saw a pattern or phenomenon that was ‘interesting’. It had to be more than a single occurrence because that could just be a ‘blip’ in the time-space continuum (ok, perhaps not there, but a ‘statistical blip’). It only becomes interesting when it happens often enough to be identified as a pattern. Pattern detection itself is influenced by past learning and models, or to bring the analogy back to research, to theory or past research. But it’s not entirely from past learning – new patterns are detected as people spend time immersed in data (otherwise known as ‘information’).


(it’s only after you’ve spent some time immersed in this picture that you’ll pick up the hidden man!)

The pattern recognition phase is not mentioned once in the H-D scientific method – yet it’s fundamental to original research (technically it generates what is called abductive reasoning, or inference to the best explanation, but more of that in later posts!) And Ethics Committees, and funding agencies seem to have trouble coping with non-H-D studies.

Exploring data to find patterns raises questions. Oh no – questions are messy and inconvenient. Questions like ‘I wonder why…’ – and all the ‘w’ questions (who, what, when, where, why – and how which is a complimentary ‘w’ word in this!) are about thoroughly describing a phenomenon or pattern. It’s only once we have a good description of it, can we then start to think about mechanisms to explain the ‘w’ questions.

But think back to your ‘research methods’ course…

How much time was spent on these topics:

– wondering and pondering

– graphing and describing

– philosophy and history

and how much was on:

– experimental design

– statistical analysis

– qualitative vs quantitative

In other words, I think many therapists may learn ‘techniques’ for research, but not an awful lot about why it needs to be done. And therapists learn a lot about confirmatory data analysis such as hypothesis development and statistical analysis, but not a lot about the messy end of making sense of raw data.

Yet – most of us deal with messy raw data every day. It’s the bread and butter of therapy. What we don’t do is systematically collect that data, then explore it so we can identify patterns.

Let’s think of some examples:

– there ‘seem to be’ a number of people who come to pain management with an erroneous understanding of what ‘pacing’ is (ie pacing equals ‘never allowing my pain to go above a certain level’)

– ‘some people’ attending pain management have trouble limiting their activity level rather than being deactivated

– ‘it looks like’ referrals come at different rates from different specialties

– ‘many’ people attend ‘many’ different pain management programmes, but are referred for more therapy

– people being referred to pain management ‘seem to have a lot of’ assessment but ‘less’ treatment

Each thing inside apostrophes is an unquantified phenomenon, and an opportunity for collecting information, then exploring it. Once the information is explored it’s possible to generate some competing explanations for the phenomenon.

For example, if we can identify that 70% of people with a pain duration of more than 18 months have 8 pain assessments, but 2 pain management programmes, and these people are funded by compensation funders; and that the remaining 30% of people with a pain duration of 18 months have 2 pain assessments, no pain management programmes, and are funded by general health funding, we can start to develop some explanatory hypotheses that can be systematically studied.

But this is messy, ‘nonproductive’ research. It doesn’t come up with ‘answers’. It doesn’t ‘explain’ things – it asks questions. It needs to lead to ‘more research’, it needs continuity of investigation so it doesn’t get dropped, and it is real.

In my PhD study, I’m looking to explore the ways people who cope in the community despite having chronic disease processes achieve important activities in their life despite pain. I’m basing this on the observation (that is supported by limited formal data) that there are more people who report pain and disability than are seeking treatment from pain management. I’m not sure what coping strategies I may find out about. I’m not sure how these people developed their skills. I don’t know whether the skills they use are the same as those developed in formal pain management programmes. I don’t have any hypotheses, I don’t have a control group (because I’m not ‘testing’ anything), I don’t want random sampling because I want to look at the widest range of coping skills there might be, I don’t want to eliminate people because they are ‘using high doses of opiates’ because this might be a coping strategy that works well for some people. And yet I’ve been asked whether I have any and all of these included or excluded in my study.

I’m using grounded theory – but grounded theory isn’t well known in the H-D world. It’s a well-recognised research strategy, but doesn’t adhere to the common precepts of the H-D model.

Part of the appeal of ‘doing research’ for me is not knowing what I may find out. The H-D model deliberately forces a researcher to decide before a study what he or she wants to find out. It’s certainly necessary for learning about ‘reality’ and ‘truth’, but in itself it’s not the only part of ‘doing research’.

This is one of quite a few posts I’ll be writing on science and therapy – keep watching!

Why is it important for therapists? As I said before, therapists work right inside ‘messy’ information all day, everyday. Learning to tolerate and manage this data allows us to notice important phenomenon. Perhaps things that haven’t yet been explored in depth, and that might generate new ideas and priorities for the confirmatory part of research.

If you’re keen on learning about abductive reasoning, I thank Associate Prof Brian Haig at Canterbury University for stimulating my thinking and exposing me to this line of science. Some of his writing includes:
Haig, B. D. (2007). Grounded theory. In N. J. Salkind (Ed.), Encyclopedia of Measurement and Statistics, Vol. 1 (pp. 418-420). Thousand Oaks: Sage.

Haig., B. D. (2006). Consistency tests establish empirical generalizations. Nature, 442, 632.

Haig, B. D. (2006). Qualitative research: A realist perspective. Proceedings of the 2006 Joint Australian and New Zealand Psychological Societies’ Conference, pp 150-154.

Haig, B.D. (2005). Exploratory factor analysis, theory generation, and scientific method. Multivariate Behavioral Research, 40, 303 -329.

Another paper to read is:
– this is a reasonably old paper but great reading.

One size does not fit all – people with pain are not clones

On a similar theme from my post ‘Pain management can’t be cloned’, I want to post about the need to tailor therapy to suit the person.  Pain management does not follow a recipe – principles yes, protocols … not quite so sure.

What do I mean by this?  Well let’s take two people with back pain that is making it difficult for them to work.  Both have trouble bending forward, sitting, walking and sleeping.  Both are male, mid-30’s, hard workers in manual labouring jobs.  Same date of onset, same reported pain intensity.   Neither man responding very well to NSAIDs with gastric problems and now taking regular paracetamol but no other medications.  ‘Objective’ measures of forward flexion, straight leg raising, lower limb strength are only slightly limited, and neurological testing is normal, and both present with paravertebral spasm, localised tenderness to palpation over L4/5, and localised pain over the central low back.  There are no changes on X-ray, and MRI/CT is not indicated.

Robbie is a surfer in his spare time, hasn’t been surfing recently but has been swimming, and until now had back pain intermittently but had never stopped working.  He comes from a family where pain has been ‘toughed out’, and he expects that this episode will eventually settle too.   He is surprised that his back pain hasn’t settled, but is carrying on with regular walking and intends to return to work as soon as he is able.  His main concern is that he finds during exacerbations of his pain, he has trouble settling his body down – his breathing changes, he feels nauseous, giddy, his heart rate rapidly increases, he becomes sweaty and quite distressed.  He has difficulty getting off to sleep and his sleep is interrupted.  He is typically active when his pain is settled, but then experiences a long period of intense discomfort (boom and bust pattern of activity).  He is quite careful with his movements, and in particular takes care not to move into certain positions that he has found have been associated with pain.

Andrew is a motorcyclist and is continuing with recreational motorcycling despite having had what he calls ‘grumbling’ back pain for most of his adult life.  His back pain hasn’t really settled completely, and this episode has got him quite worried because it is more intense than ever before.  His family are becoming quite irritated with the ongoing saga of his back – and want him to ‘get on with it’.  He has stopped walking, going to the gym and doesn’t like swimming.  He stays fit by using an exercycle, but doesn’t enjoy it.  His main concern is his lack of energy, and his increasing need for sleep. He does have trouble getting off to sleep on occasion, and goes to bed quite early because he is fatigued, but wakes regularly through the night and about an hour earlier than he would really like to.  He has even had an occasional afternoon nap.  He doesn’t really avoid movements that increase his pain – but has reduced his overall activity level because of his fatigue.  He notices that he has gained weight because he is not doing very much.

Sound familiar?  Clients like both of these men often attend pain management centres – and often they receive exactly the same treatment.  Lots of ‘core stability’, fitness training, daily scheduling, relaxation training and a graded return to work programme.   And these probably help in some way… but let’s take a closer look at what might be a more targeted and individualised strategy.

Robbie may have pain-related anxiety and avoidance. He describes increased physiological arousal and has learned that certain movements are best avoided.  He’s not deactivated or unfit – his cardiovascular fitness as measured by a 3-minute step test has remained high.  Treatment should probably focus on helping him develop skills to manage his physiological arousal (eg biofeedback, relaxation training, cognitive behavioural therapy to help identify automatic thoughts and replace with more helpful thoughts and behaviours), alongside graded exposure to those movements and activities that he finds concerning.   Sleep management would focus on strategies to improve sleep hygiene, reduction of unhelpful rumination as he goes off to sleep, and possibly sleep restriction.  A graded return to work would have a greater chance of success if he develops strategies to work to quota, and addresses his automatic thoughts and beliefs about needing to get everything done very fast (and very well).  He can probably return to work quite quickly once he has overcome his concerns about activities such as lifting and bending.

Andrew on the other hand, has lost fitness and lacks energy to increase his activity level.  He is assessed as being depressed and is treated with a combination of both antidepressant medication and cognitive therapy for this.  His family need to be brought into his treatment and it would help if he developed ways of communicating with them.  He needs to develop a daily activity plan with a schedule for both active and pleasurable activities.  Because he normally attends the gym and walks, his programme needs to be developed around this – swimming or hydrotherapy probably won’t be helpful.  He may need some help with ‘efficient’ ways for relaxing through the day, but doesn’t have the need for specific methods to reduce physiological arousal that Robbie does.  He may need to have a similar programme of sleep hygiene and sleep restriction, but because worry isn’t a problem for him, he may not need to focus on ways to address this before sleep.  His ‘pacing’ will need to focus on gradual increases in activity – and increases set by his fitness level.  Graded return to work will need to be progressed according to his cardiovascular fitness and fatigue/depression levels rather than avoided activities.

In effect, the two men have quite different treatment programmes despite experiencing very similar functional limitations.   This is why it’s so important to assess their presentation very carefully and develop hypotheses about what might be perpetuating their problems.  Any ‘protocol’ should ensure that the following principles are followed:

  • good assessment
  • generation of a number of competing hypotheses about cause and maintenance of the problems
  • interventions designed to confirm or disconfirm the hypotheses or based on a confirmed hypothesis
  • pre and post as well as follow-up outcome measures

A protocol may be less effective when it specifies the content of each session or the processes used to facilitate learning.  People learn at different rates, have different learning styles, varying automatic thoughts and underlying beliefs and attitudes, and live in different environments.

Health professionals have skills in applying concepts to specific situations.  This is why people are professionals, not computers or robots.

Sorry I don’t have any references for this post – unusual for me!  If you’ve got some – or any comments – let me know! And if you’ve been provoked by this post and don’t want to miss any others – use the RSS feed above, and subscribe!