Clinical library · Outcome measures
Functional Independence Measure
Eighteen items across motor and cognitive domains, each scored 1 to 7 by burden of care. Its motor items are reliable and its cognitive items are not, and that difference decides how the total should be read.
In one line. Thirteen motor items and five cognitive items, each rated 1 (total assistance) to 7 (complete independence), summed to a total from 18 to 126. The scoring anchor is how much another person has to do, which is why it maps onto staffing and length of stay.
It is the standard inpatient rehabilitation measure across much of the world. The thing to know before using the total is that its two domains do not behave alike.
The motor/cognitive split is not cosmetic. In the reliability meta-analysis of 11 studies, 1,568 patients and 221 reliability coefficients, median subscale reliability ran from 0.95 for Self-Care down to 0.78 for Social Cognition, and individual item reliability from 0.90 for Toilet Transfer down to 0.61 for Comprehension. Motor items were generally more reliable than cognitive or communication items. [1] A total score adds a well-measured domain to a poorly measured one.
The numbers you actually need
| Property | Value | Source and caveat |
|---|---|---|
| Inter-rater reliability, total | Median 0.95 | 11 studies, 1,568 patients, 221 coefficients [1] |
| Test-retest / equivalence | Median 0.95 and 0.92 | Same review [1] |
| Subscale reliability | 0.95 (Self-Care) down to 0.78 (Social Cognition) | Same review [1] |
| Item reliability | 0.90 (Toilet Transfer) down to 0.61 (Comprehension) | Same review [1] |
| MCID, older adults after hip fracture | 22 points total; 21 points motor | 701 patients admitted with an FIM total between 40 and 80. AUC 0.82 to 0.85 depending on anchor; the cognitive subscore was not responsive [3] |
| Responsiveness, hip fracture | Effect size 1.38 total; 1.78 motor | 9 of 10 a priori hypotheses met [3] |
| Dimensionality | Motor scale fits the Rasch model only after adjustment | Adequate fit and unidimensionality required partial credit parameterisation, rescored categories and four testlets to handle local dependency; all other analytical pathways required item deletion [2] |
| Physiotherapy effect on FIM in spinal cord injury | Robotic vs overground gait training SMD 0.38 (0.08 to 0.67) | Low-quality evidence; upper limb training with functional electrical stimulation gave SMD 1.31 (0.62 to 1.99) from two trials [5] |
What it measures
Burden of care. Every rating answers the same underlying question: how much of this task does someone else have to do, and does that helper need to be trained? That framing is why the measure has been used for casemix and reimbursement as much as for clinical change.
The motor domain covers self-care, sphincter control, transfers and locomotion. The cognitive domain covers comprehension, expression, social interaction, problem-solving and memory. The evidence above says these should usually be reported separately.
Where it misleads
1. The total score mixes reliable and unreliable measurement
Comprehension has a median reliability of 0.61 and Toilet Transfer 0.90; Social Cognition as a subscale sits at 0.78 against 0.95 for Self-Care. [1] Adding them produces a number whose precision is neither of those. Where the clinical question is motor recovery, the motor subscore is the better instrument — and in the hip fracture responsiveness study, the motor subscore had the larger effect size (1.78 against 1.38) while the cognitive subscore was not responsive to change at all. [3]
2. Summing the items assumes a unidimensionality the scale does not naturally have
The Rasch history of this instrument is a cautionary tale in itself. Adequate fit and unidimensionality for the motor scale were achieved only with partial credit parameterisation, rescored disordered thresholds, and four testlets to accommodate local dependency; every other analytical pathway required deleting items. [2] Treating the raw total as an interval-level measurement is a convenience, not a property.
3. An MCID derived in one population and admission band does not transfer
The best-derived MCID here — 22 points total, 21 motor — comes from 701 older adults after hip fracture whose admission FIM total was between 40 and 80. [3] The authors state that band explicitly. A patient admitted at 100 has nowhere near 22 points of headroom, and the figure does not apply.
4. FIM change is not a clean measure of what your treatment did
In spinal cord injury, a systematic review of 33 trials found low-quality evidence that a small number of physiotherapy interventions increase FIM or SCIM scores — robotic against overground gait training gave a pooled standardised mean difference of 0.38, and upper limb training with functional electrical stimulation 1.31 from two trials. The review's purpose was to establish what needs controlling for when FIM is used as an outcome for novel interventions. [5] If background therapy moves the score, an uncontrolled before-and-after FIM change tells you little about a specific treatment.
What the evidence supports — and what it does not
Supported
- Excellent reliability across settings, raters and patients for the total and for motor items. [1]
- Reporting the motor subscore where motor recovery is the question. [1][3]
- An MCID of 22 points total and 21 motor in older adults after hip fracture admitted between 40 and 80. [3]
- Using admission FIM patterns to anticipate length of stay and discharge destination. [4]
Not supported
- Reading the cognitive subscore as responsive to change. It was not, in the population where responsiveness was formally tested. [3]
- Treating the raw total as interval-level measurement without Rasch adjustment. [2]
- Transferring the hip fracture MCID to other populations or admission bands. [3]
- Attributing FIM change to a single intervention without controlling for background physiotherapy. [5]
- Relying on Comprehension or Social Cognition items individually. Reliability 0.61 and 0.78. [1]
How certain is this?
Evidence grade: Moderate.
The reliability evidence is unusually solid for a rehabilitation instrument: a quantitative review aggregating 221 coefficients from 1,568 patients, with the coding itself checked for inter-rater agreement. [1] Its age is not a defect — it remains the definitive synthesis, and the scale has not changed.
The MCID rests on a single large retrospective cohort in one institution, with ten a priori hypotheses of which nine were met and areas under the curve of 0.82 to 0.85. [3] That is a well-designed responsiveness study, and it is one study in one population.
The Rasch work is methodological rather than clinical, and its conclusions concern how the scale must be handled statistically rather than whether it is useful. [2] The stroke subtype analysis is a single-centre retrospective study of 373 patients. [4] The spinal cord injury review explicitly rates its own evidence as low quality, with a median PEDro score of 6.0. [5]
What would change the grade: MCID replication in stroke and in other admission bands, and responsiveness data for the cognitive domain in a population where it might plausibly change.
Common questions
Should I report the total or the subscores?
Usually the subscores, and always the motor one if motor recovery is the question. Subscale reliability runs from 0.95 for Self-Care to 0.78 for Social Cognition, [1] and in the one population where responsiveness was formally tested the motor subscore showed a larger effect size than the total (1.78 against 1.38) while the cognitive subscore was not responsive at all. [3]
What change counts as clinically important?
In older adults after hip fracture admitted with a total between 40 and 80, 22 points on the total and 21 on the motor subscore, derived by triangulating anchor-based methods with areas under the curve of 0.82 to 0.85. [3] Those conditions are part of the answer — outside that population and that admission band, published MCIDs for the FIM are not established, and this page will not invent one.
Can I use FIM change to show my treatment worked?
Not on its own. A systematic review of 33 randomised trials in spinal cord injury found low-quality evidence that several physiotherapy interventions move FIM or SCIM scores — the review existed precisely to identify what must be controlled for when these scales are used as outcomes for new interventions. [5] Background therapy is a confounder, not a constant.
Is the total score a proper interval scale?
Not without work. Achieving adequate Rasch fit and unidimensionality for the motor scale required partial credit parameterisation, rescoring of disordered thresholds and grouping items into four testlets to handle local dependency; other analytical approaches required deleting items altogether. [2] Ordinary summing is a practical convention.
Does admission FIM predict anything useful?
It appears to. Latent class analysis of 1,592 FIM records from 373 stroke patients identified six admission classes based on motor and cognitive independence, with median length of stay ranging from 126 days in the most dependent group to 29 in the most independent, and the proportion discharged home ranging from 27% to 98%. [4] That is a single-centre retrospective study, so treat the figures as illustrative of the pattern rather than as benchmarks.
References
- Ottenbacher KJ, Hsu Y, Granger CV, et al. The reliability of the functional independence measure: a quantitative review. Archives of Physical Medicine and Rehabilitation. 1996 Dec;77(12):1226–32. doi:10.1016/s0003-9993(96)90184-7 PMID 8976303 Quantitative review
- Lundgren Nilsson Å, Tennant A. Past and present issues in Rasch analysis: the functional independence measure (FIM™) revisited. Journal of Rehabilitation Medicine. 2011 Oct;43(10):884–91. doi:10.2340/16501977-0871 PMID 21947180 Rasch analysis and methodological review
- Arcolin I, Godi M, Giardini M, et al. Minimal clinically important difference of the functional independence measure in older adults with hip fracture. Disability and Rehabilitation. 2024 Feb;46(4):812–819. doi:10.1080/09638288.2023.2175386 PMID 36750763 Retrospective responsiveness study
- Furuta H, Mizuno K, Unai K, et al. Functional Independence Measure Subtypes among Inpatients with Subacute Stroke: Classification via Latent Class Analysis. Progress in Rehabilitation Medicine. 2022;7:20220021. doi:10.2490/prm.20220021 PMID 35528116 Latent class analysis
- Harvey LA, Glinsky JV, Chu J. Do any physiotherapy interventions increase spinal cord independence measure or functional independence measure scores in people with spinal cord injuries? A systematic review. Spinal Cord. 2021 Jul;59(7):705–715. doi:10.1038/s41393-021-00638-0 PMID 34099880 Systematic review
About this resource
Using this in clinic
Every figure here is traceable to its source.
Every reliability value on this page is given at the level it was measured — total, subscale or item — because for this instrument those three numbers are not the same. 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.