If you read one stroke trial this year, its primary outcome is almost certainly the modified Rankin Scale (mRS) at 90 days. It is the lingua franca of stroke outcomes: a single ordinal score from 0 to 6 that compresses an entire patient’s functional status into one number a trialist, a regulator, and a bedside clinician can all talk about. Its strength is exactly that compression; its weakness is the same thing. Knowing how the grades are anchored, where the clinically decisive boundaries sit, and how reliably the score can actually be assigned is the difference between using the mRS well and reciting it.

The seven grades, anchored to function

The scale is deceptively simple, but each grade is anchored to a specific functional threshold rather than to a symptom count:

  • 0 – No symptoms at all.
  • 1 – No significant disability despite symptoms; able to carry out all usual duties and activities.
  • 2 – Slight disability; unable to carry out all previous activities but able to look after own affairs without assistance (independent).
  • 3 – Moderate disability; requires some help, but able to walk unassisted.
  • 4 – Moderately severe disability; unable to walk unassisted and unable to attend to own bodily needs without assistance.
  • 5 – Severe disability; bedridden, incontinent, and requiring constant nursing care and attention.
  • 6 – Dead.

Two boundaries do most of the work. The 2/3 boundary is the line of independence: a patient who can still manage their own affairs without help is a 2, while one who needs some help is a 3. The 3/4 boundary is the line of walking and self-care: a 3 walks unassisted, a 4 does not. Note a historical wrinkle worth keeping straight — the original van Swieten description graded handicap from 0 to 5, and the grade 6 (death) was incorporated as the scale was adopted as a trial endpoint.1

Bedside pearl

When you are stuck between two grades, ask two questions in order: Can they walk without a person’s help? (separates 3 from 4.) Can they manage their own affairs without help? (separates 2 from 3.) Walking and independence, in that order, resolve most disagreements at the bedside.

Why baseline and 90 days, and not the number alone

The mRS is a measure of current global function, not of stroke severity, so it only becomes informative when you anchor it. Record the pre-stroke (baseline) mRS at presentation — the level of function the day before the event — and the outcome mRS at 90 days, the conventional endpoint for stroke recovery once early fluctuation has settled. A patient who returns to their own baseline of 2 has a very different result from one who was independent and is now a 2, even though both carry the same outcome score. Trials handle this either by enrolling patients with a favorable baseline or by analyzing the shift across the whole distribution. The 90-day window is a convention, not a biological constant; some recovery continues beyond it, and a few protocols and registries use other timepoints, so always check which is being reported.

The 0–2 cut-point and how trials count it

In the modern reperfusion era, “functional independence” is operationally defined as mRS 0–2 at 90 days — a patient who, at worst, has slight disability but remains independent in their own affairs. This dichotomy anchors the headline number-needed-to-treat figures in the thrombectomy trials. In the HERMES individual-patient-data meta-analysis pooling five positive 2015 trials, endovascular thrombectomy significantly reduced disability across the whole mRS distribution, with a number-needed-to-treat of 2.6 to reduce disability by at least one mRS level in one patient, and benefit that held across age, occlusion site, and patients ineligible for intravenous alteplase.4 Trials increasingly favor this ordinal shift analysis — testing whether the entire distribution moves toward lower scores — over a single dichotomy, because it uses information from every grade and typically requires a smaller sample size than older binary endpoints.2

Pitfall

The 0–2 dichotomy is not universal. Some trials — especially those in large-core, basilar, or elderly populations — define success as mRS 0–3, because a moderate-disability survivor can still be a meaningful gain in a population otherwise headed for severe disability or death. Before you quote an NNT or an “independence” rate, confirm which cut-point that specific trial used. The same data can read very differently at 0–2 versus 0–3.

Reliability: the case for a structured interview

The mRS’s recurring criticism is inter-rater variability. In the original van Swieten study, two physicians agreed exactly in 65 of 100 patients, differed by one grade in 32, and by two grades in 3; the unweighted kappa was 0.56 and the weighted kappa 0.91 — reassuring that large disagreements were rare, but a reminder that adjacent-grade disagreement is common.1 A systematic review of mRS reliability found overall agreement only moderate by unweighted kappa, and a mock-trial study with experienced raters still showed substantial interobserver variability even with training available.35 The practical fix is a structured, scripted interview: across the reliability literature, structured assessment consistently raises agreement above ad hoc grading, which is why centralized trials and quality programs adopt it.23

This is educational reference material, not individualized treatment advice. Definitions of functional independence, follow-up timing, and certification requirements for mRS scoring vary by trial, guideline, and institution — confirm local standards before applying any cut-point to a specific decision.