Clinical Reasoning: It's Not About Right or Wrong

A working framework for MSK therapists

When you first start training as a therapist, clinical decisions are often taught in black and white. If this test is positive, they have X. If it’s negative, it’s Y. Work through the algorithm, find the root cause, treat it. Easy.

Then you start treating real people and it turns out that the algorithms don’t always fit, the tests don’t always agree, and the “correct” answer isn’t always obvious. That’s when the doubts creep in.

The shift you need to make is this: clinical reasoning is not about being right or wrong. It’s about probabilities.

It’s all probabilities

Every working hypothesis sits on a sliding scale. Some are very likely. Some are very unlikely. Most live somewhere in between. Your job isn’t to find the one correct answer — it’s to gather information that pushes the probability of each hypothesis higher or lower, until you have enough confidence to act.

In theory, probability runs from 0% (absolutely sure they don’t have it) to 100% (absolutely sure they do). In practice, neither end is achievable in medicine. You’ll work somewhere between 0.01% and 99.9%. What matters is where a hypothesis sits right now, and what you need to do next.

A simple decision rule:

·       Above ~70% — confident enough to treat it.

·       Between 30% and 70% — uncertain; gather more information and retest.

·       Below ~30% — confident enough to scrap it and move on.

These thresholds aren’t fixed. If missing a diagnosis could kill or permanently disable your client, the bar for scrapping it has to be much higher. You don’t rule out cauda equina because you’re 70% sure it isn’t there — you rule it out when you’re virtually certain.

The four steps

The information-gathering loop has four stages, and you cycle through them repeatedly:

·       Acquire data. Initial observation, history, physical examination, results of any tests you run. Start before the client even sits down — how do they walk in, what’s their colour, are they limping, what’s their demeanour?

·       Interpret and organise. Make sense of what you’ve gathered. Which findings cluster? What story do they tell together?

·       Make a hypothesis. Build a differential — a list of what might be causing the problem. Rank it by importance: life-threatening first, life-changing next, systemic and serious conditions, then the more everyday MSK and psychological contributors.

·       Test the hypotheses. Work through the list and ask, for each one, how likely is this and what would change my mind?

You don’t do this once. You do it after the initial observation, again after the history, again after the physical, again after any tests. At every stage, each hypothesis gets one of three verdicts: treat it, test it further, or scrap it.

A worked example: “my back hurts”

Step 1. All you know is that they’ve got back pain. At this point you genuinely cannot rank anything — the list is wide open and every hypothesis is uncertain. You need more data.

Step 2. You observe and take a history. How did they walk in? Any limp? Skin colour? Pain down the legs — one side or both, front or back? Bowel and bladder function? Sexual function? Abdominal pain? Cardiovascular history, previous injuries, family history, recent falls, weight changes, night pain?

Your differential starts to firm up:

·       Abdominal aortic aneurysm — no cardiovascular history, non-smoker, no pulsatile pain, no family history. Very unlikely.

·       Cauda equina syndrome — no saddle anaesthesia, no bowel or bladder changes, no bilateral leg symptoms, no sexual dysfunction. Unlikely.

·       Inflammatory or systemic disease (Crohn’s, ankylosing spondylitis) — no bowel changes, no weight loss, no family history, no night sweats. Unlikely.

·       Mechanical MSK — muscle fatigue, DOMS, facet irritation — came on after a gym session he hadn’t done in a while; pain is worse after inactivity, easier with movement; paracetamol helps; aggravated by side-bending and flexion. Likely.

·       Psychological load — he’s self-employed, worried about losing work, and his father had a chronic back problem that ended his career. Likely, and almost certainly contributing.

Step 3. At this point you can scrap the dangerous stuff with reasonable confidence, treat the mechanical and psychological contributors, and keep a quiet eye on anything that doesn’t resolve as expected. If the mechanical hypothesis doesn’t respond to treatment as predicted, that’s a signal to loop back to step 1 and widen the differential again.

Where MSK therapists can get stuck

Our training is strong on MSK. It’s often thinner on visceral, endocrine, neurological, vascular, and rheumatological conditions, the things that can masquerade as mechanical pain. That’s the silent risk in our practice: not the diagnoses we consider and get wrong, but the ones we never put on the list in the first place.

A few common pressure points:

·       Acquiring data — not asking the systemic questions, because we weren’t trained to think beyond the tissue in front of us.

·       Test quality — using orthopaedic tests without knowing their sensitivity, specificity, or whether clusters outperform individual tests.

·       Interpretation — not knowing what a positive or negative result actually shifts, probabilistically.

·       Hypothesis generation — a differential that’s all MSK because that’s the only shelf in the mental library that’s well stocked.

The fix isn’t to become a physician. It’s to widen the differential early, ask systemic questions routinely, and know when your knowledge has run out.

When to refer

If you’ve cycled through the loop, exhausted the questions and tests you know, and you’re still uncertain — refer. Someone with a different training or a broader knowledge base will run a similar looking process, but the hypotheses they generate and the tests they value will come from a wider library than yours. That isn’t a failure of your reasoning. It’s the reasoning working correctly.

The goal is not to know everything. It’s to know what you know, know what you don’t, and keep your differential wide enough that the dangerous possibilities get considered before they get ruled out.

With thanks to Dr Rahul Patwari, whose YouTube series on clinical reasoning was the inspiration for this post. The probability-based framework, the four-step loop, and the “treat it / test it / trash it” decision are all his; any clumsy adaptation to sports therapy practice is mine.