Investigation of Minimal Clinically Important Difference of the Extended Barthel Index in Inpatients With Acute to Subacute Stroke in China.

209 stroke inpatients (mean age 59, 150 men, 59 women) at one rehabilitation hospital in Kunshan, China, were assessed with the Extended Barthel Index (16 items, total 0–64 points) at admission and again 4 weeks later, after receiving the hospital's standard post-stroke rehabilitation pathway. At the 4-week mark, three groups — the patient, the occupational therapist, and a family member acting as the expense payer — each rated the patient's change in daily-living ability on a 7-point scale (1 = significant improvement to 7 = significant deterioration). Four statistical methods (distribution-based, ROC curve, anchor-based, and simple linear regression) were then applied to work out what size of EBI score change counts as 'clinically meaningful' for each rater group. The study was approved by the hospital ethics committee; no randomisation or blinding was involved because this was a measurement study, not a treatment trial.

The mean EBI score increased from 37.96 ± 12.37 at admission to 42.99 ± 11.91 at 4 weeks (p < .001), a 5-point gain on a 64-point scale. The distribution-based method gave MCID values ranging from 3.53 to 6.51 points depending on the sub-method used. The ROC curve method yielded AUC values above 0.99 for all three rater groups but near-zero specificity (0.0%–4.5%), so the derived cutoffs (2.5–3.5 points) could not reliably separate 'improved' from 'not improved' patients. The anchor-based method, which the authors recommend, produced improvement MCIDs of 4.5 (OT), 4.0 (patient), and 6.0 (payer) points, and deterioration MCIDs of −2.0 (OT), −1.0 (patient), and 0 (payer) points. The simple linear regression method gave closely matching values (improvement 4.25/3.85/5.88; deterioration −1.94/−1.31/−0.32), supporting the anchor-based results.

When you score a stroke inpatient on the Extended Barthel Index, a 4.5-point gain is the smallest change you should expect to call 'improvement' after a 4-week rehabilitation cycle, and a 2-point drop is the threshold for 'slight decline.' These numbers give you a concrete benchmark for setting 4-week goals, for deciding whether a patient's progress is meaningful or just measurement noise, and for explaining to patients and families what a realistic gain looks like. They also help you interpret the EBI's broader domains (communication, social cognition, vision) that go beyond basic self-care and mobility.

Single hospital in China; the MCID values may not transfer to other countries, healthcare systems, or patient populations. All patients were in the acute-to-subacute window (15–180 days post-stroke, mean about 34 days); the values may not apply to chronic-phase patients (≥6 months). Only one 4-week interval was measured; the impact of longer rehabilitation courses on the MCID is unknown. The 'expense payer' role was filled by family members bearing the financial cost, not by insurance companies or government payers, so the payer MCID reflects family perception rather than institutional funding criteria. The ROC curve method produced near-zero specificity, so its cutoffs (2.5–3.5 points) should not be used as clinical thresholds despite the high AUC. No randomisation or control group; this is a single-group measurement study, so the 5-point mean gain over 4 weeks cannot be attributed to any specific intervention.

Funded by the Science and Technology Bureau of Suzhou (Project SKYD2022018) and the Science and Technology Plan Project of Kunshan (Projects KS2462, KSZ2566) — government research grants. No industry funding. Authors declared no conflicts of interest.

Do not use the 2.5–3.5 point MCID values from the ROC curve method as clinical thresholds. Despite AUC values above 0.99, specificity was 0.0% to 4.5%, meaning almost every patient the rater called 'not improved' still had a score gain above the cutoff, so the method cannot reliably tell you whether a change is meaningful. The anchor-based values (4.0–6.0 for improvement, −2.0 to 0 for deterioration) are the ones the authors recommend, and even those come from a single Chinese hospital, one stroke phase, and one 4-week interval.