EVIDENCE-BASED PHYSICAL THERAPY · CHAPTER 2 · CLINICAL QUESTIONS & STUDY DESIGN
A good study can still be the wrong study.
A high-quality randomized trial may answer a treatment question beautifully—but it won’t necessarily tell you what a new symptom means or how someone will function in six months. Match the design to the question before judging the headline.
1. Are you asking about treatment, diagnosis or prognosis?
Your original Evidence-Based PT lecture distinguishes different types of clinical questions. Treatment asks whether an intervention changes an outcome. Diagnosis asks how accurately a test identifies a condition. Prognosis asks what may happen over a period of time.
Take a patient with shoulder pain. ‘Does this manual technique improve short-term function compared with another approach?’ is an intervention question. ‘How accurately does this test identify the target diagnosis?’ is a diagnostic question. ‘Which features predict recovery over three months?’ is prognostic.
A precise PICO or similar framework helps identify what outcome matters. Before opening a database, decide what sort of answer would actually change the patient’s care.
Quick check: A study reports how many patients recover within six months after an injury. What kind of question is it closest to?
Reasoning: Prognosis. The relevant evidence would typically follow a defined group over time; whether a treatment caused recovery is a different question.
2. Which design gets you closer to an answer?
For treatment effects, a well-conducted randomized comparison can help separate intervention effects from baseline differences. For prognosis, a longitudinal cohort may follow outcomes after a common starting point. Diagnostic accuracy studies compare an index test with an appropriate reference standard in a relevant patient population.
Qualitative studies ask different questions: for example, why people find a home exercise program hard to maintain. Cross-sectional studies describe status or associations at one time and generally cannot establish which event came first. Case-control designs may examine past exposures among people with and without an outcome.
The key isn’t memorizing a universal ‘best’ study type. It’s asking what the design can and cannot establish.
| Clinical question | Often useful design | What to check |
|---|---|---|
| Intervention effect | Randomized comparison | Allocation, co-interventions, dropout |
| Prognosis | Followed cohort | Baseline and follow-up completeness |
| Diagnostic accuracy | Index test versus reference standard | Patient spectrum and blinding |
| Experience and barriers | Qualitative inquiry | Sampling and credibility |
Quick check: You find a cross-sectional study linking pain and weakness. Does it prove weakness caused the pain?
Reasoning: No. It describes an association measured around the same time; direction of effect and confounding remain uncertain.
3. Study labels are only the beginning.
Your Research Designs chapter compares experimental, quasi-experimental and observational approaches, prospective versus retrospective timing, and between- versus within-subject designs. These terms describe how the study was organized, not whether its conclusions are trustworthy.
Ask whether the sample resembles your patient, whether groups were selected or allocated appropriately, whether important outcomes were measured consistently, and whether missing data or other treatment could explain the result.
Finally, integrate the findings with clinical expertise, individual preferences and feasibility. The research may be strong but less applicable if the intervention or population differs greatly from the patient in front of you.
Quick check: A systematic review has a prestigious title. Can you skip checking the included studies and relevance?
Reasoning: No. The review may synthesize studies with different populations or important biases. Check its question, methods, included evidence and applicability.
Sources & learning scope
Based on the original Evidence-Based PT class notes, particularly Chapter 2 Evidence, Chapter 5 Research Designs and Chapter 21 Developing a PICO Question. The question-design table is original teaching synthesis rather than a copied hierarchy or a validated selection rule.
Next: sampling, random allocation and why groups can differ before treatment begins.
Put it into practice
Scenario: Two articles concern knee osteoarthritis. One compares treatments for function, and another follows patients for twelve months to estimate who improves.
What should you conclude?
Reasoning: The comparison study most directly addresses intervention effectiveness; the longitudinal cohort answers a prognosis question. They are not competing grades for the same purpose.
Common trap: Ranking studies without asking what question their design can answer. Remember: Study design follows the clinical question.
From My PT School Notes · Free student lesson. Examples are for education, not individualized care or a substitute for supervised clinical instruction.
