The SF-36, short for 36-Item Short Form Health Survey, measures perceived health across eight scales scored from 0 to 100. It is free to use from its publisher, it has no legitimate global score, and three of its scales hit a ceiling. The full, programmed questionnaire is right below.
Three scales hit a ceiling: both role scales and social functioning. A patient already at 100 on one of them will not be able to show any improvement there. The detail is below.
Figures taken from the studies listed at the end of this page, every value carrying its source where it is written. The 36 items are reproduced from the RAND Corporation questionnaire and, for French, from the Haute Autorité de Santé appendix, without rewording; the two departures found in that reproduction are flagged below.
What the SF-36 measures
The SF-36 measures perceived health, not a disease. Published in 1992 by Ware and Sherbourne as part of the Medical Outcomes Study, it covers thirty-six items across eight concepts: physical functioning, role limitations due to physical health, bodily pain, general health perceptions, vitality, social functioning, role limitations due to emotional problems and mental health.
Its strength is being generic. It compares chronic low back pain with heart failure, which no region-specific scale can do, and that is why it appears in almost every large cohort.
Its reliability was tested in 3,445 patients and replicated across 24 subgroups: it runs from 0.65 to 0.94 depending on the scale, with a median of 0.85. Item completion runs from 88 to 95 %, somewhat lower among older people and those with less education.
Eight scales, no global score
The SF-36 does not yield one score out of 100. It yields eight scores out of 100, and averaging them is the most widespread misreading of all.
The 1993 validity study explains why. Physical functioning and role-physical best distinguish degrees of medical severity; mental health and role-emotional best distinguish psychiatric disorders. But social functioning, vitality and general health measure both at once: adding them up mixes dimensions the authors took care to show are not interchangeable.
Two summary scores do exist, physical and mental, but computing them rests on weightings that cannot be derived from the eight scales. A spreadsheet that recomputes them by hand gets them wrong; the profile of the eight scales, by contrast, reads with nothing but the eight values.
Three scales that hit a ceiling
This is the limit the SF-36 has carried since its validation, and the one that matters most in rehabilitation.
The 1994 study records noteworthy ceiling effects on both role scales and on social functioning. In other words, a sizeable share of respondents already scores the maximum there, and those three scales can then no longer rise, whatever the real benefit of treatment.
The 2021 meta-analysis in cardiac and pulmonary rehabilitation follows the reasoning through. The role scales explain only 24 to 27 % of the variance in change there, and the authors conclude that the SF-36 is not suited as a pre- to post-programme measure in those two fields. The minimum clinically important difference, measured in 310 patients operated on for a musculoskeletal tumour, settles around 5 points for both summary scores.
| Scale | What it best distinguishes | Ceiling effect |
|---|---|---|
| Physical functioning | medical severity | no |
| Role limitations, physical health | medical severity | yes |
| Bodily pain | mostly the physical | no |
| General health perceptions | physical and mental | no |
| Vitality | physical and mental | no |
| Social functioning | physical and mental | yes |
| Role limitations, emotional problems | psychiatric disorders | yes |
| Mental health | psychiatric disorders | no |
Free in English, another matter in French
The original questionnaire is free. Its publisher writes that its surveys are public documents, that it neither charges nor requires permission, and explicitly grants permission to use it, with no further written permission needed.
Two conditions come with that permission: any change must be identified as made by whoever made it, and printing as well as distribution must carry a credit line naming the publisher and the Medical Outcomes Study. The user also takes responsibility for any translation.
That is precisely where French changes things. The permission covers the English questionnaire; the validated French version is a distinct work, published in 1998 by Leplège et al. after a first rapid translation tested in 1995 in 1,007 Geneva residents. It is that validated version, and not a home-made translation, that the scorer above reproduces, from the Haute Autorité de Santé appendix of January 2008 where it appears with the kind permission of Alain Leplège. That permission was given to the Haute Autorité de Santé: it is not a permission for this page, which names the source and refers to its author for any other use.
Where the 36 items reproduced above come from
The English items are those of the RAND 36-Item Health Survey 1.0, which RAND Corporation permits to be used with no further written permission, provided its credit line is carried and any modification is identified: that credit line sits under the items, and therefore in the exported PDF too. The French items are those of the validated version by Leplège et al., as published in Appendix 10 of the Haute Autorité de Santé argumentaire of January 2008, where they are reproduced with the kind permission of Alain Leplège. That permission was given to the Haute Autorité de Santé, and is not a permission for this site: the source and its condition of use are named, and any other use is to be requested from its author.
The scorer reproduces both versions without rewording them, and two departures from the Haute Autorité de Santé PDF are declared here. The pain intensity scale is printed there on five lines, without the figure 4, for a question that has six of them as in English: it is restored to its six answers from a second reproduction of the same French version, Appendix 11 of a doctoral thesis deposited on theses.fr in 2016, which prints the full scale. And the printing slip « faire su sport », readable on the page of the PDF, is corrected to « faire du sport ».
Three administration pitfalls
Averaging the eight scales
The number you get looks like a result and is not one. The eight scales do not measure the same thing and three of them mix physical and mental health: the profile goes into the notes with all eight values.
Following a patient already at 100 on a role scale
A scale at the ceiling can no longer rise. If role functioning is already maximal at baseline, the absence of progress on that line says nothing about the treatment, only about the instrument.
Using a translation found online
The publisher's permission covers its English questionnaire. A French translation circulating without a reference has neither the 1998 validation nor the responsibility that goes with it, and the measurement properties quoted here do not apply to it. The scorer above reproduces the validated version as the Haute Autorité de Santé publishes it, with the source named under the items.
Frequently asked questions
What does SF-36 stand for?
SF-36 is short for 36-Item Short Form Health Survey. The questionnaire was published in 1992 by Ware and Sherbourne, as part of the Medical Outcomes Study.
How is the SF-36 scored?
In two steps, and that is the rule the scorer above applies. Each answered item is first recoded to a 0 to 100 range following its publisher's table, which turns the reversed items round on the way; then the items of one concept are averaged. That yields eight scores out of 100, a high score signalling a better state, and no global score. An item left blank does not count: the scale is calculated on the items it has answers for, and at least one is needed. The item on perceived change over one year belongs to none of the eight scales.
How much change counts?
The only value established by two convergent methods is around 5 points on the summary scores, measured in 310 patients operated on for a musculoskeletal tumour. It does not carry over as it stands to every population.
Is the SF-36 free to use?
Yes for its publisher's English questionnaire, which explicitly grants permission to use it with no further written permission, provided its credit line is carried and any change is identified.
Is there a validated French version?
Yes, published in 1998 by Leplège et al., preceded by a rapid translation tested in 1995. It is a distinct work from the English questionnaire, with its own conditions of use, and it is the one the scorer above reproduces, from the Haute Autorité de Santé appendix of January 2008.
References
7 sources, PMIDs included
- Ware JE, Sherbourne CD. The MOS 36-item short-form health survey (SF-36). I. Conceptual framework and item selection. Med Care 1992;30(6):473-83. PMID 1593914. The original publication. Thirty-six items for eight health concepts, from limitations in physical activities to mental health, self-administered from age 14 or by interview, telephone included.
- McHorney CA, Ware JE, Raczek AE. The MOS 36-Item Short-Form Health Survey (SF-36). II. Psychometric and clinical tests of validity in measuring physical and mental health constructs. Med Care 1993;31(3):247-63. PMID 8450681. The study that sorts the scales out. Physical functioning and role-physical best distinguish degrees of medical severity; mental health and role-emotional best distinguish psychiatric disorders. Social functioning, vitality and general health measure both at once and are the hardest to interpret.
- McHorney CA, Ware JE, Lu JF, Sherbourne CD. The MOS 36-item Short-Form Health Survey (SF-36). III. Tests of data quality, scaling assumptions, and reliability across diverse patient groups. Med Care 1994;32(1):40-66. PMID 8277801. 3,445 patients, replicated across 24 subgroups. Reliability from 0.65 to 0.94, median 0.85. Floor effects negligible except for the two role scales, and above all noteworthy ceiling effects on both role scales and on social functioning.
- Perneger TV, Leplège A, Etter JF, Rougemont A. Validation of a French-language version of the MOS 36-Item Short Form Health Survey (SF-36) in young healthy adults. J Clin Epidemiol 1995;48(8):1051-60. PMID 7775992. A first, rapid French translation, tested in 1,007 Geneva residents aged 18 to 44. Internal consistency from 0.76 to 0.92 and two principal components, physical and mental. The authors explicitly present it as a stopgap before the official adaptation.
- Leplège A, Ecosse E, Verdier A, Perneger TV. The French SF-36 Health Survey: translation, cultural adaptation and preliminary psychometric evaluation. J Clin Epidemiol 1998;51(11):1013-23. PMID 9817119. The official French adaptation, by forward and backward translation with conceptual equivalence given priority. The distance between response choices was checked on visual analogue scales in 30 subjects, and face validity tested in lay panels.
- Ogura K, Yakoub MA, Christ AB, Fujiwara T, Nikolic Z, Boland PJ, Healey JH. What Are the Minimum Clinically Important Differences in SF-36 Scores in Patients with Orthopaedic Oncologic Conditions? Clin Orthop Relat Res 2020;478(9):2148-58. PMID 32568896. 310 patients operated on for a musculoskeletal tumour, with two convergent calculation methods. The minimum clinically important difference settles around 5 points for both summary scores, physical and mental.
- van Rotterdam FJ, Hensley M, Hazelton M. Measuring Change in Health Status Over Time (Responsiveness): A Meta-analysis of the SF-36 in Cardiac and Pulmonary Rehabilitation. Arch Rehabil Res Clin Transl 2021;3(2):100127. PMID 34179763. A meta-analysis of the responsiveness of the SF-36 in cardiac and pulmonary rehabilitation. The role scales explain only 24 to 27 % of the variance there, and the authors conclude that the SF-36 is not suited as a pre- to post-programme measure in those two fields.
Page written by Anthony Baillon, physiotherapist, co-founder of Physio Learning. The questionnaire itself was developed at RAND Corporation as part of the Medical Outcomes Study, and published in 1992 by Ware and Sherbourne; its French version is that of Leplège et al., 1998. This page reproduces them from their sources, named under the items.
Behind this article
An author who explains, a reviewer who checks.
How we write and check our content
Physiotherapist · co-founder of Physio Learning
Marked for life by his first four-hour lecture without a single image, he took a master’s in instructional design so that it would never happen to anyone again. He hunts down publication bias and unreadable slides with the same intransigence.
Head of scientific content
Physiotherapist specialising in neuro-musculoskeletal practice and holder of a master’s in public health. He checks the methodological rigour of every article: primary sources, levels of evidence, no exceptions.