Clinical library · Outcome measures
Oswestry Disability Index
The most used disability measure in low back pain. Reliable, responsive, and carrying a minimal clinically important difference that ranged from 0.8 to 25 points across sixteen accepted ways of calculating it — in the same patients.
In one line. Ten sections — pain intensity, personal care, lifting, walking, sitting, standing, sleeping, sex life, social life and travelling — each scored 0 to 5, expressed as a percentage of the maximum. Higher is worse.
It is quick, it is condition-specific, and its psychometric credentials are sound. The thing to understand before quoting a threshold is that the threshold depends on the arithmetic you choose, and that choice can triple or halve the apparent success rate of a treatment.
The MCID is a choice, not a fact. Applying sixteen accepted statistical methods to the same cohort of lumbar fusion patients produced an average MCID of 10.5 ± 7.0 points across a range of 0.8 to 25. Distribution methods gave lower thresholds with greater variability, anchor methods higher thresholds with less. Depending on the method, between 30% and 83% of the same patients were classified as having reached the MCID. The authors conclude this calls into question the utility of a single statistically determined value for assessing surgical success. [1]
The numbers you actually need
| Property | Value | Source and caveat |
|---|---|---|
| MCID, across calculation methods | Average 10.5 ± 7.0, range 0.8 to 25 | Same cohort, 16 methods, elective 1- and 2-level lumbar fusion [1] |
| Proportion reaching MCID at 6 months | 30% to 83% | Depending only on which method was used [1] |
| Relative-change threshold | 30% improvement correctly classified 81% as "success" | Surgically treated lumbar spinal stenosis; consistent with the NORDSTEN trials [2] |
| Responsiveness | Effect sizes, standardised response means and AUC all above 0.8 | In lumbar spinal stenosis, comparable with the condition-specific Zurich Claudication Questionnaire [2] |
| Against the Roland-Morris questionnaire | Better test-retest reliability and smaller measurement error | But Roland-Morris had better construct validity as a measure of physical functioning; all included studies were poor or fair quality [3] |
| Core psychometrics | Good construct validity, acceptable internal consistency, high test-retest reliability and responsiveness | Eight articles; low burden of administration [4] |
| Versions | More than one exists | Different versions of the index are in circulation; check which you are using [5] |
What it measures
Self-reported limitation in daily activities attributed to low back pain. It is a disability measure, not a pain measure, although its first section asks about pain intensity.
Scoring is straightforward: sum the section scores, divide by the maximum possible for the sections completed, express as a percentage. The commonly omitted section is sex life, and the scoring rule handles omissions by reducing the denominator — which is why two clinicians can produce different percentages from the same form if one treats an omission as zero.
Where it misleads
1. Quoting "the MCID is 10 points" hides the choice you made
This is the finding that should change how the index is reported. Sixteen legitimate methods applied to one cohort produced thresholds from 0.8 to 25 points, and the proportion of patients judged to have improved ranged from 30% to 83%. [1] A treatment can be made to look successful or unsuccessful without changing a single patient outcome. State the method alongside the threshold, or state a relative change instead.
2. Relative change is more defensible than absolute change
In surgically treated lumbar spinal stenosis, correlations with a global perceived effect anchor were moderate (above 0.50) for absolute change but strong (above 0.67) for relative change and for follow-up scores, and a 30% relative improvement correctly classified 81% of patients as a success. [2] A patient starting at 60% and one starting at 20% do not have the same 10 points available to them.
3. There is more than one Oswestry
Several versions of the index are in circulation, differing in wording and in one section entirely. [5] Scores from different versions are not interchangeable, and the version is rarely stated in clinical records. Record which you use.
4. "Better than Roland-Morris" is not established
A systematic review of head-to-head comparisons found the Oswestry had better test-retest reliability and smaller measurement error, while Roland-Morris had better construct validity as a measure of physical functioning, with conflicting evidence on responsiveness and inconclusive evidence on internal consistency. All eleven studies were of poor or fair methodological quality, and the conclusion is that there are no strong reasons to prefer either. [3] Choose on familiarity and consistency, not on a claimed superiority.
What the evidence supports — and what it does not
Supported
- Good construct validity, acceptable internal consistency and high test-retest reliability. [4]
- High responsiveness, comparable with condition-specific measures in lumbar spinal stenosis. [2]
- A 30% relative improvement as a defensible success threshold. [2]
- Low administrative burden, suiting routine clinical use. [4]
Not supported
- Any single absolute MCID. Range 0.8 to 25 in one cohort. [1]
- Comparing success rates between studies using different MCID methods. 30% versus 83% from method alone. [1]
- Treating it as clearly superior to Roland-Morris. [3]
- Assuming two "Oswestry" scores are the same instrument. [5]
- Reading it as a pain measure. It measures disability.
How certain is this?
Evidence grade: Moderate.
The MCID-variability finding is the most useful and most secure item here, because it is an internal comparison: one cohort, sixteen methods, so the variation cannot be attributed to population differences. [1] It is a single surgical cohort, so the exact range would differ elsewhere; the principle would not.
The responsiveness and threshold work in lumbar spinal stenosis is methodologically sound and anchored to global perceived effect. [2] It is surgical, and this page is read mostly by clinicians treating non-surgical patients — a real limit on transfer.
The head-to-head comparison against Roland-Morris rests on nine articles and eleven studies, all rated poor or fair quality, and its own conclusion is that higher-quality studies are needed. [3] The general psychometric summary draws on eight articles. [4]
What would change the grade: MCID work in conservatively managed low back pain using a pre-specified method, and reporting standards that state the version used.
Common questions
What change should I treat as meaningful?
Prefer a relative threshold. A 30% improvement correctly classified 81% of surgically treated lumbar stenosis patients as a success and was consistent with the NORDSTEN trials. [2] If you must use an absolute value, state the method: across sixteen accepted methods in one cohort the MCID averaged 10.5 with a standard deviation of 7.0 and a range of 0.8 to 25. [1]
Why do published success rates differ so much?
Often because of arithmetic rather than treatment. In the same patients, the proportion reaching MCID at six months ranged from 30% to 83% depending only on the calculation method chosen. [1] When comparing two studies, check whether they used the same method before concluding one treatment worked better.
Oswestry or Roland-Morris?
Either, consistently. The Oswestry showed better test-retest reliability and smaller measurement error; Roland-Morris showed better construct validity as a measure of physical functioning; evidence on responsiveness was conflicting. All the head-to-head studies were poor or fair quality, and the review found no strong reason to prefer one. [3] Changing instrument mid-pathway costs you more than either choice gains.
Does the version matter?
Yes. More than one version of the index is in circulation. [5] They are not interchangeable and the version is almost never recorded. Fix one for your service and write it on the form.
Can I use it for spinal stenosis rather than general low back pain?
Yes. In surgically treated lumbar spinal stenosis it showed high internal and external responsiveness, with effect sizes, standardised response means and areas under the curve all above 0.8, comparable with the condition-specific Zurich Claudication Questionnaire. [2] See also low back pain and sciatica for the treatment evidence.
References
- Solomito MJ, Kia C, Makanji H. The Minimal Clinically Important Difference for the Oswestry Disability Index Substantially Varies Based on Calculation Method: Implications to Value-Based Care. Spine. 2025 May 15;50(10):707–712. doi:10.1097/BRS.0000000000005074 PMID 38887023 Methodological comparison study
- Indrekvam K, Myklebust TÅ, Austevoll IM, et al. Responsiveness of the Oswestry Disability Index and Zurich Claudication Questionnaire in patients with lumbar spinal stenosis: evaluation of surgically treated patients from the NORDSTEN study. European Spine Journal. 2024 Nov;33(11):4270–4280. doi:10.1007/s00586-024-08440-1 PMID 39134699 Responsiveness study
- Chiarotto A, Maxwell LJ, Terwee CB, et al. Roland-Morris Disability Questionnaire and Oswestry Disability Index: Which Has Better Measurement Properties for Measuring Physical Functioning in Nonspecific Low Back Pain? Systematic Review and Meta-Analysis. Physical Therapy. 2016 Oct;96(10):1620–1637. doi:10.2522/ptj.20150420 PMID 27081203 Systematic review
- Vianin M. Psychometric properties and clinical usefulness of the Oswestry Disability Index. Journal of Chiropractic Medicine. 2008 Dec;7(4):161–3. doi:10.1016/j.jcm.2008.07.001 PMID 19646379 Systematic review
- Fairbank JC. Why are there different versions of the Oswestry Disability Index?. Journal of Neurosurgery: Spine. 2014 Jan;20(1):83–6. doi:10.3171/2013.9.SPINE13344 PMID 24206036 Review
About this resource
Using this in clinic
Every figure here is traceable to its source.
Every threshold on this page states the method used to derive it, because for this index the method changes the answer more than the patients do. Where a value could not be verified against the paper it came from, it is not on this page, and the omission is stated rather than filled with a number from a secondary source.