FROM MY PT SCHOOL NOTES · EVIDENCE-BASED PHYSICAL THERAPY · CHAPTER 4
Can You Trust an Outcome Measure?
Reliability, validity and responsiveness answer different questions. Students need to tell them apart before choosing an outcome measure.
What is really being measured?
Before selecting an outcome measure, name the construct you want to observe—such as balance confidence, walking performance or pain interference—and define how it will be scored. A measured outcome is a dependent variable when it serves as the response in a research question. A result is only useful when that variable matches the clinical decision.
Measurement scale affects interpretation: nominal variables name categories, ordinal values are ordered, interval scales have equal units without a true zero, and ratio scales have a meaningful zero. For example, an ordinal rating should not automatically be handled as though its numbered categories have equal distances. These distinctions come from the original variables lecture.
Is the score consistent?
Reliability describes consistency of measurements across time, raters or items. Validity asks whether a given interpretation and use are supported. The lecture distinguishes face, content, construct (convergent and discriminant) and criterion-related (concurrent and predictive) validity.
Additional terms from the scanned Chapters 5–8 glossary: test–retest reliability, interrater and intrarater reliability, parallel-forms reliability, split-half reliability, criterion-referenced and norm-referenced scores, and measurement error. These terms describe different questions about how consistently or meaningfully scores can be used.
Does the change matter?
Responsiveness concerns detection of actual change. Floor and ceiling effects can obscure decline or progress; the course links meaningful change to what matters to clinicians and patients, not only the numerical score.
Check your reasoning: If a score is consistent across repeats, does that prove it measures the intended construct?
No. Reliability concerns consistency; validity concerns whether the evidence supports the interpretation and intended use of the score.
Updated-standard note
The historical glossary defines minimal detectable change (MDC) as change just exceeding the standard error of measurement (SEM). That is not the usual individual-level MDC95 calculation; under common assumptions MDC95 = 1.96 × √2 × SEM. Keep SEM (single-score measurement error) distinct from an MDC/SDC threshold for change, and check which model and confidence level were used. COSMIN measurement-property references: https://www.cosmin.nl/
Put it into practice
Scenario: Two students use the same balance questionnaire a week apart on a patient whose function has not changed. Both get nearly identical scores, but the items mostly ask about confidence rather than actual walking performance.
What should you conclude?
Reasoning: The repeatability supports reliability in that setting; it does not prove the questionnaire measures walking performance. Define the construct first, inspect validity evidence for that interpretation, and use a performance measure if actual walking performance is the question.
Common trap: Calling a tool ‘validated’ without saying for whom, what construct, or what decision. Takeaway: Ask three separate questions: Is the score consistent? Is it measuring the intended construct? Can it detect change that matters?
Course basis: my Evidence-Based PT school notes, source lecture Chapter 7. New questions and headings are teaching aids, not official exam items. For full definitions and examples, use the class notes and assigned readings.
