01 / 20

Evidence-Based
Decision Making

Apply research evidence, clinical guidelines, and prediction rules to inform sound patient care decisions.

National Physical Therapy Examination (NPTE)

What Is Evidence-Based Practice?

Evidence-Based Practice (EBP) is the integration of the best available research with clinical expertise and patient values to guide care decisions. It is the cornerstone of safe, effective physical therapy.

The Three Pillars of EBP

Best Research Evidence Clinical Expertise Patient Values & Preferences EBP

The Levels of Evidence Hierarchy

Systematic Reviews Randomized Controlled Trials Cohort / Case-Control Studies Expert Opinion / Case Reports Strongest Weakest

Clinical Prediction Rules (CPRs)

A clinical prediction rule combines patient findings into a count or score to estimate likelihood of a diagnosis, prognosis, or response to treatment.

Diagnostic
Predicts presence of a condition (e.g., Ottawa Ankle Rules).
Prognostic
Predicts likely outcome or recovery course.
Prescriptive
Predicts who will respond to a given intervention.

Core Statistic: Likelihood Ratios

Positive likelihood ratio: \( LR+ = \dfrac{\text{sensitivity}}{1 - \text{specificity}} \)

Negative likelihood ratio: \( LR- = \dfrac{1 - \text{sensitivity}}{\text{specificity}} \)

Rule In vs. Rule Out

SnNout High Sensitivity Negative test Rules condition OUT SpPin High Specificity Positive test Rules condition IN

Statistical vs. Clinical Significance

Statistical Significance
A result unlikely due to chance, often \( p < 0.05 \). Says nothing about how meaningful it is.
Clinical Significance
A change large enough to matter to the patient (exceeds the MCID).

How to Apply: The 5-Step Process

1
Ask a focused question using PICO (Patient, Intervention, Comparison, Outcome).
2
Acquire the best available evidence from databases and guidelines.
3
Appraise the evidence for validity, impact, and applicability.
4
Apply findings with clinical expertise and patient values.
5
Assess the outcome and adjust the plan as needed.

Worked Example: Ottawa Ankle Rules

A patient presents after an ankle inversion injury. The Ottawa Ankle Rules have very high sensitivity (\( \approx 98\% \)) for fracture.

The rule is negative: no bony tenderness, and the patient can bear weight for four steps.

Worked Example: Interpreting an LR

A special test for a lumbar condition has sensitivity \( = 0.80 \) and specificity \( = 0.90 \). Compute \( LR+ \).

$$ LR+ = \frac{0.80}{1 - 0.90} = \frac{0.80}{0.10} = 8 $$

Worked Example: MCID Decision

A trial reports a statistically significant pain reduction of \( 0.4 \) points on a \( 10 \)-point scale (\( p = 0.03 \)). The established MCID is \( 2.0 \) points.

Common Mistakes

✕ Evidence-Only Thinking
Ignoring patient values or clinical context because "the study said so."
✕ Confusing P and Importance
Treating any \( p < 0.05 \) as clinically important regardless of effect size.
✕ Misusing CPRs
Applying a prediction rule to patients unlike the original study population.
✕ Ignoring Evidence Quality
Citing low-level evidence as if it equals a systematic review.

Summary: Key Metrics

What It Tells You Best Use
Sensitivity True positive rate Ruling out (SnNout)
Specificity True negative rate Ruling in (SpPin)
\( LR+ \) Increases post-test probability Confirming a condition
\( LR- \) Decreases post-test probability Excluding a condition
MCID Meaningful change threshold Judging clinical relevance

Question 1 of 5

Which three elements form the foundation of evidence-based practice?

Question 2 of 5

A test has high sensitivity. A negative result is most useful to:

Question 3 of 5

A test has sensitivity \( = 0.90 \) and specificity \( = 0.80 \). What is \( LR+ \)?

Question 4 of 5

A study shows a statistically significant improvement (\( p = 0.02 \)) that is smaller than the MCID. The best interpretation is:

Question 5 of 5

Which level of evidence sits at the top of the evidence hierarchy?

Key Takeaways

1
Integrate all three pillars: evidence, expertise, and patient values.
2
Weigh evidence quality using the hierarchy, not just its existence.
3
Use SnNout / SpPin and likelihood ratios to rule conditions in or out.
4
Always check whether change exceeds the MCID before acting.