In physiotherapy, measuring the outcomes of an intervention objectively and in a standardized manner is as important as the intervention itself. Without reliable measurements, we cannot know whether a patient is truly improving or whether the perceived improvement is a placebo effect, a natural fluctuation of the condition, or a bias from the therapist. Functional assessment scales, Patient-Reported Outcome Measures (PROMs), and standardized functional tests constitute the measurement arsenal of the physiotherapy researcher. However, using these instruments correctly requires understanding psychometric concepts that are not always addressed with sufficient depth in clinical training.
Choosing an inadequate measurement instrument can completely compromise a research study, no matter how rigorous its design may be in other aspects. A questionnaire that does not actually measure what it claims to measure (lack of validity), that produces inconsistent results across measurements (lack of reliability), or that is unable to detect clinically relevant changes (lack of responsiveness) generates data that, at best, contributes noise and, at worst, leads to erroneous conclusions about the effectiveness of a treatment.
Fundamental psychometric properties
The reliability of an instrument refers to the consistency of its measurements. In physiotherapy, the most relevant forms of reliability are test-retest reliability (stability of scores when the instrument is administered at two close time points to stable patients), inter-rater reliability (agreement between different physiotherapists evaluating the same patient), and internal consistency (the degree to which the items of a questionnaire measure the same construct). Test-retest and inter-rater reliability are evaluated using the intraclass correlation coefficient (ICC), which takes into account both the correlation and the absolute agreement between measurements. ICC values above 0.75 are considered good, and above 0.90 excellent.
That covers raters. When what you are comparing is two different instruments instead (a new dynamometer against the reference one, for example), it is worth complementing the ICC with a Bland-Altman analysis, which on top of overall agreement also detects whether the bias between the two methods grows with the magnitude of the measurement; the Bland-Altman calculator computes it with its limits of agreement.
Validity addresses the question of whether the instrument actually measures what it claims to measure. Content validity evaluates whether the items of the instrument adequately cover all relevant facets of the construct. Construct validity examines whether the scores relate to other variables in the theoretically expected way (for example, a lumbar disability questionnaire should correlate moderately with a pain scale and with objective mobility measures). Criterion validity compares the scores with a reference standard (gold standard), although in many areas of physiotherapy there is no clear gold standard, which complicates this evaluation.
Responsiveness is an especially critical property in physiotherapy, where the primary goal of measurement is usually to evaluate whether a patient improves with treatment. An instrument can be reliable and valid but unable to detect clinically important changes because its items are too general or because the scoring range does not have sufficient granularity. The most commonly used measures of responsiveness are the standardized effect size (SES), the standardized response mean (SRM), and the Guyatt responsiveness index. Values above 0.80 on any of these indices indicate good ability of the instrument to detect change.
Minimal detectable change and minimal clinically important difference
Two fundamental concepts for interpreting outcomes in physiotherapy are the minimal detectable change (MDC) and the minimal clinically important difference (MCID). Although they sound similar, they answer different questions and their confusion is a frequent source of interpretive errors.
The minimal detectable change is the minimum amount of change in the score that exceeds the measurement error of the instrument. It is calculated from the standard error of measurement (SEM), which in turn depends on the standard deviation of the scores and the test-retest reliability: SEM = SD x sqrt(1 - ICC). The MDC is obtained as 1.96 x sqrt(2) x SEM, which corresponds to the 95% confidence interval for the difference between two measurements. If a patient obtains a change in their score that exceeds the MDC, we can be reasonably confident that the change is real and not an artifact of measurement error. For example, if a shoulder functionality scale has an MDC of 8 points, a patient who improves by 5 points could simply be within the instrument's margin of error, while one who improves by 12 points has experienced a genuine change.
The minimal clinically important difference goes one step further: it is the minimum amount of change that the patient perceives as relevant or that has clinical significance. It is determined through anchor-based methods, which compare changes in scores with an external criterion such as the patient's global rating of change, or distribution-based methods, which use statistical criteria such as half the baseline standard deviation. Anchor-based methods are generally preferred because they incorporate the patient's perspective, which is ultimately what determines whether a change has clinical relevance.
Instrument selection for research
Choosing the appropriate measurement instrument for a physiotherapy research study requires considering several factors simultaneously. The first is the adequacy to the construct of interest: a study on the effect of an exercise program in patients with chronic low back pain could use the Oswestry Disability Index (ODI) or the Roland-Morris Disability Questionnaire (RMDQ) to measure disability, the Visual Analogue Scale (VAS) for pain, and the EQ-5D or SF-36 for health-related quality of life. Each instrument measures a different aspect of the outcome, and the choice should be guided by the research question.
The second factor is the availability of validated versions in the language of the study population. Using an unvalidated translation of a questionnaire is a serious methodological error that can invalidate the results. The process of cross-cultural adaptation of an instrument (similar to questionnaire validation in psychology) requires forward translation, back-translation, review by an expert committee, pilot testing, and evaluation of the psychometric properties in the new population. Resources such as the COSMIN (COnsensus-based Standards for the selection of health Measurement INstruments) database offer systematic reviews of the psychometric properties of health measurement instruments that greatly facilitate the selection process.
The third factor, often neglected, is clinical feasibility: administration time, patient burden, the need for special equipment, and ease of scoring. An instrument with excellent psychometric properties but that requires 45 minutes to administer may be unfeasible in a clinical context where patients are fatigued or the available time is limited. In these cases, short forms or computer adaptive testing (CAT) offer alternatives that maintain measurement quality while significantly reducing response burden.
Reporting and interpreting results
When publishing results of physiotherapy studies, it is essential to contextualize statistical changes with the MDC and MCID values of the instrument used. Reporting that a group improved significantly (p < .05) on a disability questionnaire score without indicating whether that improvement exceeds the clinically important change threshold is providing incomplete information that hinders clinical decision-making. Inferential statistics tell us whether the observed change is probably real, but only the comparison with the MCID tells us whether that change matters to the patient.
An increasingly recommended practice is to complement group-level analyses (mean differences, effect sizes) with responder analyses, that is, calculating what proportion of patients in each group achieved a change equal to or greater than the MCID. This information is directly relevant to clinical practice because it allows estimating the individual probability of benefit, something that group means cannot offer. If a treatment produces a modest mean improvement but 60% of patients exceed the clinically important change threshold, that information is much more useful for the physiotherapist making daily decisions than a simple p-value accompanied by a mean difference.