"What's actually going on,
and how confident am I?"
This module builds a systematic, defensible framework for diagnosis and prognosis in MSK practice — using Bayesian reasoning as its spine. It's the framework that scales across every condition you'll encounter: from the patient who walked in this morning to the one you've never seen before.
You need a process that works when you don't."
Clinical reasoning under uncertainty is a skill — not a knowledge gap.
📋 Learning Outcomes
1 — Diagnosis & Clinical Categorisation
Understand the diagnostic features and clinical sub-categories of the most common MSK conditions in private practice — LBP, rotator cuff, anterior knee pain, lateral hip pain, and tendinopathy.
2 — Prognostic Frameworks
Understand prognostic frameworks for common MSK conditions — with specific attention to mean vs. median outcomes and what this means for patient communication and treatment mapping.
3 — Bayesian Reasoning in Clinical Practice
Understand how Bayesian reasoning underpins every clinical decision — prior probabilities, updating with evidence, and how tools like the STarT Back Tool formalise this process.
4 — Imaging Base Rates & Nocebo Prevention
Understand asymptomatic population imaging findings — and how this knowledge protects patients from over-medicalisation and nocebo-driven disability escalation.
5 — Building Treatment Plan Templates
Create basic treatment plan templates for 3–5 common MSK conditions as clinical reference tools — co-developed in session as applied outputs.
MSK Conditions — Diagnosis &
Clinical Categorisation
Diagnosis in MSK practice is rarely binary. Most conditions exist on a spectrum and have distinct sub-types that behave very differently — both in natural history and in treatment response. Understanding these categories is your first clinical reasoning tool.
🦴 The Five Core Conditions
Low Back Pain
Episodic / PersistentNeck Pain
Recurrent / High ChronicityShoulder Pain
Heterogeneous — Sub-type DependentAnterior Knee Pain
Not Self-LimitingLateral Elbow Pain (Lateral Epicondylalgia)
Generally Self-Limiting📚 Key Principle: Categorise, Don't Just Label
Why Sub-Classification Matters
A diagnosis label (e.g., "LBP") tells you almost nothing about how to treat someone. A sub-classification (e.g., "flexion-sensitive with radicular features, high fear-avoidance") tells you:
- Which direction of movement to load, unload, or explore
- Whether the nervous system is the primary driver vs. the tissue
- How much of the management should be passive vs. active
- What the most likely prognosis is — and how much you can shift it
- How to communicate the journey ahead in a way that reduces allostatic load, not adds to it
📎 Stimulus Resources — Conditions
Bayesian Reasoning
Applied to Clinical Practice
Bayesian reasoning is the most updated model of how the brain makes predictions and updates beliefs. It's also the framework that transforms clinical complexity from something overwhelming into something you can navigate systematically — and communicate confidently to your patients.
🐎 The Horses and Zebras Principle
You Already Do This — You Just Don't Know It
If you heard hoofbeats outside your clinic right now in suburban Brisbane, you'd think horse, not zebra. Not because zebras don't exist — but because the base rate probability of a horse is far higher in that context.
But if the circus was in town and a news report said animals had escaped — suddenly your prior probability shifts. A zebra becomes plausible. This is Bayesian reasoning: starting with the base rate and updating it as new evidence arrives.
Clinical translation: Knee OA in a 20-year-old is a zebra. Knee OA in an 80-year-old is a horse. The same positive test in both patients gives you entirely different diagnostic probabilities — because the base rates are different.
🔁 The Core Framework
Prior Probability
What do I already know about this condition, this population, this person? Base rates + your experience.
New Evidence
Assessment findings, validated tools, objective measures, clinical tests, patient-reported outcomes.
Updated Probability
Prior × likelihood of evidence = new posterior probability. Does this confirm or revise the expected journey?
New Prior
This updated probability becomes your new baseline. Repeat at every touchpoint.
🛠️ Clinical Tools as Bayesian Instruments
Tools That Update Your Prior
- STarT Back Tool — screens psychosocial risk; low/medium/high triage directly maps to expected journey
- ÖREBRO / PSFS — psychosocial flags and function baseline
- Stage of Change — readiness to engage in self-management
- Goal vs. current function gap — their narrative tells you where their head is at
- Range of motion + movement quality — avoidance vs. exposure orientation
What Each Piece of Evidence Tells You
- High STarT score → less likely to follow median response; shift toward mean (longer, messier)
- Goals close to current function → likely to follow natural history with light-touch support
- Large goal–function gap + catastrophising → likely needs more frequent early contact
- Avoidance + movement fear → disability plateau likely; belief structure needs addressing
- Strong social supports, positive framing → weight toward median, less frequent contact
⚠️ The Two Sources of Prior Probability
Epidemiological Evidence vs. Your Experiential Evidence
You have two independent prior probability sources — and you should use both:
- Epidemiological evidence: Population-level data on natural history, prognosis, mean vs. median outcomes. This is your objective anchor — what "should" happen based on the research.
- Your experiential evidence: Patterns you've observed in your specific patient population, context, and setting. A rural Queensland caseload has different base rates than a private city clinic. Both sources are valid.
Where they align — you can be more confident. Where they diverge — that divergence is information. Ask: "Why does my experience differ from the literature here? What is it about my population that changes this?"
💬 The Communication Payoff
You need to know your process."
Telling a patient "here's what I expect and here's what would make me want to check in sooner" builds more confidence than pretending certainty you don't have.
🧠 Why This Reduces Your Patient's Pain
Stress research is clear: things with low predictability kick off a stress response. Neuroendocrine inflammatory markers rise with uncertainty. When you give a patient a process — a map of what to expect and how you'll navigate deviations — you reduce their allostatic load. A lower allostatic load means:
- Pain is less amplified centrally
- Flare-ups feel less catastrophic (they were predicted)
- Avoidance behaviour is less likely to entrench
- They're less likely to re-present to emergency or GP with normal episodic pain
📎 Stimulus Resources — Bayesian Reasoning
Mean vs. Median —
Communicating Prognosis
This is the most underused clinical communication tool in MSK practice. Understanding the difference between mean and median outcomes — and knowing which number to anchor your patient to — changes how confident they feel about their journey and how much they trust you.
📊 The Core Distinction
Mean (Average)
- The mathematical average across the whole population
- Sensitive to outliers — a few people who do very badly (or very well) pull the average away from where most people land
- What you usually find in systematic reviews and meta-analyses
- The number to anchor patients to when their risk profile is elevated — prepare them for a longer journey
- Example: Cervical radiculopathy mean ~6–8 months
Median (Most Common)
- The middle value — where the bulk of patients actually land
- Not affected by outliers — more representative of the typical experience
- Found in individual patient data analyses and natural history studies
- The number to anchor patients to when their risk profile is low — reassure them with the most common outcome
- Example: Cervical radiculopathy median ~16 weeks
📉 Prognosis Reference Table — Key Conditions
| Condition | Median (Most Common) | Mean (Average) | Course Pattern | Key Communication Point |
|---|---|---|---|---|
| Acute LBP | 3–4 weeks for significant pain reduction | ~6 weeks | Sharp drop, then plateau | "Most people feel dramatically better in 3–4 weeks — and that's your likely path too" |
| Persistent LBP | Episodic course — 1–2 flares/year normal for anyone | Highly variable — driven by psychosocial load | Wiggly plateau, flare-and-settle | "We're not trying to get to zero — we're managing a known episodic condition" |
| Cervical Radiculopathy | ~16 weeks | 6–8 months | Gradual improvement; outliers pull mean high | Lead with median for low-risk; mean for high-risk presentations |
| PFPS / Anterior Knee Pain | 6–12 weeks moderate–high improvement | Variable — recurrence in 94% at 4 years | Initial improvement, recurrent episodes | "We'll get you better — but flare-ups are part of this condition, not a failure" |
| Lateral Epicondylalgia | ~1 year for natural resolution | 6–24 months; 90% resolve | Self-limiting, episodic flares | "This will resolve — our job is managing it so it doesn't limit your life while it does" |
| Total Knee Replacement | Near-normal function by ~12 weeks | ~12 weeks to function; pain can persist 6–12 months | Stepped improvement; walking aids off ~6 weeks | "Six weeks you're off your aids; 12 weeks most things are back — pain may linger to a year, and that's normal" |
🎨 How to Communicate the Curve
The Graph-in-the-Air Technique
You don't need to show a patient a research paper. Draw the curve with your hands. Literally gesture a bell curve in the air and say:
- "Most people with this sort of thing are here by around X weeks — that's the most common response."
- "The average is actually a little longer, because some people take quite a bit more time — and that's normal too."
- "Based on what you've told me, I think you're more likely to track toward [median/mean] — here's why, and here's how we'll know."
This does two things: it shows you understand the data, and it shows you've already thought about them specifically — not just the condition.
🧠 Why This Matters Neurologically
When a patient knows what to expect, flare-ups within that expected range are not alarming — they're predicted. The nervous system does not mount the same stress response to a known event as it does to a surprise. This is not a soft communication point — it is directly modulating central sensitisation and pain amplification by reducing the uncertainty signal that drives it.
Imaging Base Rates &
Preventing Nocebo
Imaging findings are one of the most common drivers of iatrogenic harm in MSK practice — not because imaging is wrong, but because findings are catastrophised. Understanding what is normal on a scan — in people with no pain — changes how you interpret, report, and communicate imaging results.
🖼️ The Asymptomatic Population Problem
What Scans Find in People With No Pain
Multiple large studies have scanned asymptomatic populations — people with no current pain — and found that structural changes are extraordinarily common. These are not diagnoses. They are age-related findings that exist on a spectrum from normal variation to tissue change.
| Body Region | Finding | Prevalence in Asymptomatic Adults | Clinical Implication |
|---|---|---|---|
| Lumbar Spine | Disc degeneration (any level) | 37% at age 20 → 96% at age 80 | Degeneration is normal ageing — not a pain diagnosis |
| Lumbar Spine | Disc bulge | 30% at age 20 → 84% at age 80 | |
| Shoulder | Full-thickness rotator cuff tear | 22% at age 50 → 65% at age 80 | Most RC tears are asymptomatic — structure ≠ symptom |
| Knee | Meniscal degeneration | ~60% of people over 50 | Meniscal findings on MRI in middle-aged patients are rarely the sole cause of pain |
| Cervical Spine | Disc degeneration | 40% at age 30 → 90%+ at age 60 | Cervical degeneration is the rule, not the exception in adults over 40 |
| Hip | Labral tear (any type) | ~68% of asymptomatic athletes | Labral findings should not automatically trigger surgical conversations |
⚠️ The Nocebo Risk
When a patient receives an imaging report with language like "moderate disc degeneration," "partial thickness tear," "bony spurring," or "significant wear" — without contextual education — they frequently experience:
- Increased pain intensity (catastrophising the finding)
- Reduced movement — fear of "damaging" something that isn't damaged
- Increased healthcare utilisation — seeking more scans, specialist opinions, procedures
- Reduced self-efficacy and increased dependence on passive treatment
- Worsening long-term outcomes — despite no change in underlying tissue state
Your communication about a scan can be more harmful than the finding itself if you don't use it correctly.
✅ How to Use Imaging Bayesianly
The Bayesian Imaging Framework
- Start with the base rate: What is the prevalence of this finding in asymptomatic people of this age? This is your prior probability that the finding is the pain driver.
- Ask: does this finding explain the clinical picture? Distribution of symptoms, aggravating factors, movement pattern — does the scan finding actually match?
- Communicate the finding in context: "Most people your age have some disc wear — it's like grey hair for your spine. The question isn't whether it's there; it's whether it's relevant to what you're feeling."
- Update based on response to treatment: If they respond well to loading → the structure is probably not the limiting factor. If they don't → keep updating.
📎 Stimulus Resources — Imaging Base Rates
Building Treatment
Plan Templates
Clinical reasoning without a structure for applying it becomes intuition — useful, but not scalable. A treatment plan template is not a rigid protocol. It is a Bayesian framework: a prior probability made visible, with built-in mechanisms for updating as evidence comes in. This module is co-created in session.
🗺️ The Three Layers of a Treatment Template
Natural History Anchor
What does this condition typically do? Mean, median, expected shape of the curve. This is your prior.
Stratification Gate
What tools/data at initial assessment will tell me which pathway this patient is likely on?
Feedback Loops
What will I check, and when, to confirm or update my pathway estimate? Leading vs. lagging indicators.
Decision Rules
What would I see that would trigger a pathway change — more frequent contact, MDT referral, imaging, or discharge?
📋 The Pain Avoidance / Copa Quadrant
When assessing movement and function, where a patient sits in this quadrant tells you how to approach graded exposure and belief structure work.
✅ Avoid + Validate
Avoidance is clinically appropriate (e.g. acute radiculopathy, spinal stenosis). Validate the choice — other movements available. Most likely to follow natural history. Lower frequency of contact needed.
⚠️ Avoid + Violate
Avoiding is entraining fear when loading is actually safe and needed. Require movement experiments — covert exposure (game, task-based). Likely to plateau on disability unless belief structure changes. More frequent early contact needed.
✅ Copa + Validate
Pushing into pain is actually appropriate (e.g. tendinopathy loading). Validate the effort and channel it. Good response to structured progressive loading. Likely to track median.
⚠️ Copa + Violate
Pushing repeatedly into pain is creating a pain-sensitisation cycle. Needs explicit behaviour modification. Stop checking the painful movement in every session. Re-educate without dismissing the pain.
📄 Example: Persistent LBP Template
Persistent Low Back Pain — Clinical Template Framework
Co-created · Session ReferenceEpisodic course. 1–2 flares/year is normal for any adult. Persistent LBP follows a wiggly plateau pattern — initial modest improvement, then oscillation. Mean recovery is highly variable and driven by psychosocial load. Median: moderate improvement in 4–8 weeks for pain; disability improvement is slower and more individual. Flare acute phase: 3–7 days sharp worsening, then returns to baseline. This is the prior you communicate session one.
STarT Back Tool · Goals vs. current function gap · Stage of change · Range of motion + movement orientation (avoider vs. copa) · Boom-bust cycle history
STarT low, positive framing, goals close to function. Anchor: Median
1–2 education sessions. 4–6 week follow-up. Flare plan given session 1. Boom-bust tool for independence.
STarT medium–high, catastrophising, goal-function gap large. Anchor: Mean
Fortnightly early. MDT co-management. Explicit flare management plan. Check-in tool between visits.
Daily activity tolerance (boom-bust log). Flare frequency and duration. Mood and coping language. Movement willingness in session.
STarT Back re-score at 6–8 weeks. Oswestry / PSFS. Overall disability vs. goals. Return to avoided activities.
Escalate if: increased flare frequency, new neurology, flares not returning to baseline within 7 days, STarT score worsens. Discharge if: self-managing, goals met, lagging indicators stable.
📄 Example: Total Knee Replacement Template
Total Knee Replacement — Post-Operative Template Framework
Inpatient → Outpatient BridgeNear-normal function by 12 weeks is the norm. Walking aids typically off by 6 weeks. Pain can persist to 12 months and this is normal — communicate this early. Stiffness and swelling may fluctuate with activity level changes. Key time points to anchor: 2 weeks (wound healing, initial ROM), 6 weeks (aids off, driving typically safe), 12 weeks (return to most daily function), 12 months (pain fully settles in most cases). Give this map in the first session.
Pre-existing psychosocial load? Bilateral TKR? High BMI / cardiovascular comorbidity? These shift the pathway toward more frequent contact and broader discharge criteria.
Protocol-guided progression. Weeks 2, 4, 8, 12 with discharge if milestones met. Patient-led progression within ROM and strength markers. Anchor: Natural History
Escalate if: ROM tracking well below protocol, acute swelling with systemic signs, new neurology, DVT symptoms. Escalate psychosocial if plateau in function with intact tissue healing.
It is a Bayesian prior made visible — a starting expectation that gets updated with every data point until you are confident in where this patient is going.
Build the template. Trust the process. Update when the evidence tells you to.
Module Complete
You now have the foundational framework for systematic, defensible clinical reasoning across the most common MSK presentations in private practice.
- ✅ Conditions: Sub-classification drives management — a label tells you little; a category tells you everything about the expected journey and treatment approach
- ✅ Bayesian Reasoning: All clinical decisions are probability estimates. Prior + evidence = updated probability. Repeat at every touchpoint.
- ✅ Mean vs. Median: Mean is pulled by outliers. Median is the most common response. Your risk stratification tells you which to anchor your patient to.
- ✅ Imaging Base Rates: Most structural findings are normal in asymptomatic populations. Your communication about a scan can be more harmful than the finding itself if delivered without context.
- ✅ Treatment Templates: Natural history anchor + stratification gate + feedback loops + decision rules. A prior made visible. Not a protocol — a process.
- ✅ The Pain Quadrant: Avoider vs. copa × validate vs. violate. Where a patient sits determines how you approach exposure, loading, and belief structure work.
- ✅ Communication: Telling a patient your process — what you expect, and what would make you check in sooner — builds more confidence than pretending certainty you don't have.
→ Repeat at every touchpoint
→ Communicate the process openly
Clinical confidence is not knowing the answer. It is trusting the process that will find it.
📚 Full Evidence Base — This Module
- Hartvigsen et al. (2018) — What low back pain is and why we need to pay attention · Lancet
- Meakins et al. (2019) — Physiotherapists' recommendations for rotator cuff-related shoulder pain · ResearchGate
- Crossley et al. (2016) — Patellofemoral pain clinical practice guidelines · JOSPT
- Lateral hip pain systematic review · BMC Musculoskeletal Disorders (2020, 2021)
- Lateral epicondylitis natural history · RACGP AJGP (2020)
- Asymptomatic structural imaging findings · Nature Scientific Reports (2024)
- Bayesian clinical reasoning in MSK · PubMed 21833252
- Clinical decision analysis · Journal of Clinical Epidemiology (2020)
- Veritasium — The Bayesian Trap (video)
- Collins (2024) — CSC Tuition Mentoring Curriculum · Bayesian Reasoning Applied Framework