Skip to content

Outcomes and prognostication

TL;DR — Stroke outcome is multidimensional: survival, neurological impairment, global disability, activities, cognition, mood, communication, participation, quality of life, recurrence, and caregiver impact cannot be represented by one scale (Kasner 2006, PMID 16781990). The NIHSS is useful for acute deficit severity but underweights posterior-circulation, cognitive, and language-specific disability; the modified Rankin Scale (mRS) is the dominant 90-day global endpoint but is ordinal, assessor-sensitive, and clinically heterogeneous within grades (Banks 2007, PMID 17272767). Prognostic scores can estimate group-average risk from age, severity, physiology, comorbidity, and imaging, but calibration, treatment era, missingness, and self-fulfilling withdrawal of life support constrain bedside use (Hemphill 2001, PMID 11283388; Rost 2008, PMID 18556582). Outcome is modifiable: thrombectomy yields an approximate one-category mRS improvement for every 2.6 treated patients in selected large-vessel occlusion, and organized stroke-unit care reduces death or dependency (Goyal 2016, PMID 26898852; Langhorne 2020, PMID 32324916). Prognosis should therefore support decisions and communication, never predetermine treatment before response and goals are known.

The outcome hierarchy

Level Examples Appropriate question Blind spot
Survival 7-, 30-, 90-day and 1-year mortality Is the person alive at a defined time? Survival may coexist with profound dependence or acceptable adapted life
Neurological impairment NIHSS; domain-specific motor/language scores How severe is the measured deficit? Does not directly measure daily function or participation
Global disability mRS How much assistance is needed overall? Different disability profiles share one grade (Banks 2007, PMID 17272767)
Basic activities Barthel Index Can basic self-care and mobility tasks be performed? Ceiling effect; limited cognition, communication, and participation (Kasner 2006, PMID 16781990)
Participation Return to work, community mobility, relationships Has the person resumed valued roles? Strongly shaped by environment and policy
Patient-reported health Stroke-specific quality of life, fatigue, mood What does the survivor experience? Proxy answers and aphasia exclusion can bias estimates
Recurrent events Recurrent stroke, MI, bleeding, hospitalization What happens after the index event? Competing death and surveillance intensity alter estimates

The recovery field recommends fixed biological time windows and explicit separation of restitution, compensation, and adaptation. An mRS improvement can arise from any of these processes and is not a direct measure of neural repair (Bernhardt 2017, PMID 28697708).

NIHSS: acute severity, not complete phenotype

The NIHSS is reproducible after structured training and predicts early outcome, but scores are influenced by lesion location: dominant-hemisphere language items can inflate scores relative to equally disabling right-hemisphere or posterior-circulation deficits, while gait, hand dexterity, cognition, and swallowing are sparsely represented (Kasner 2006, PMID 16781990).

Use NIHSS longitudinally only with context: sedation, intubation, language discordance, pre-existing deficit, postictal state, reperfusion, edema, and hemorrhage can change the number or its interpretation.

NIHSS use Strength Caution
Baseline severity Standardized, rapid, treatment-trial compatible A low score can conceal disabling aphasia, hemianopia, ataxia, or hand weakness
Early change Detects improvement or deterioration A total score can stay stable while deficits trade domains
Prognostic covariate Strong association with death/disability Association is not individual destiny
Eligibility threshold Operationalizes trial populations Hard cutoffs exclude clinically important phenotypes

Prehospital scales derived from NIHSS components prioritize recognition rather than outcome. The three-item Cincinnati Prehospital Stroke Scale showed good reproducibility in emergency personnel, but it cannot grade disability or exclude stroke when negative (Kothari 1999, PMID 10092713).

Modified Rankin Scale

The mRS grades global disability from 0 (no symptoms) through 5 (severe disability) and 6 (death). It has construct validity and broad treatment sensitivity, but inter-rater reliability is only moderate without structured interviews and training (Banks 2007, PMID 17272767).

Analysis Question answered Advantage Limitation
Dichotomy, e.g. 0–1 vs 2–6 Did treatment increase a chosen “good outcome”? Simple absolute risk/NNT Throws away shifts elsewhere; cutoff meaning changes with baseline severity
Ordinal shift Did treatment move the distribution toward less disability? Uses all categories; more power under proportional odds Common odds ratio is less intuitive; proportionality must be assessed
Sliding dichotomy Did outcome exceed a baseline-severity-specific threshold? Accounts for prognosis Threshold choice can be opaque
Utility-weighted mRS Did treatment improve preference-weighted health outcome? Connects disability to experienced value Utilities vary by respondent and culture

Consensus nomenclature recommends saying exactly which mRS states are analyzed rather than labels such as “favorable,” “independent,” or “excellent,” because these labels have been used inconsistently (Saver 2021, PMID 34320814).

MR CLEAN demonstrated the clinical interpretability of an ordinal endpoint: intra-arterial treatment produced an adjusted common OR of 1.67 (95% CI 1.21–2.30) for a shift toward better 90-day mRS, with an absolute 13.5-point increase in functional independence (Berkhemer 2015, PMID 25517348).

HERMES pooled five thrombectomy trials and found an adjusted common OR 2.49 (95% CI 1.76–3.53) for reduced disability; the number needed to treat for at least a one-point mRS improvement was 2.6 (Goyal 2016, PMID 26898852).

DAWN used utility-weighted mRS as a coprimary endpoint in late-window selected stroke, reporting mean utility-weighted mRS 5.5 with thrombectomy versus 3.4 with control at 90 days; its highly selected mismatch population limits unselected extrapolation (Nogueira 2018, PMID 29129157).

Barthel and domain-specific function

The Barthel Index measures dependence in ten basic activities such as feeding, bathing, transfers, toileting, and mobility. It is useful for rehabilitation planning and change in basic self-care, but has ceiling effects and limited representation of executive function, language, fatigue, fine hand use, social roles, and emotional health (Kasner 2006, PMID 16781990).

Domain-specific measures should accompany the mRS when the intervention has a domain-specific mechanism. A language therapy can meaningfully improve communication without shifting mRS; an arm intervention can improve impairment without changing basic self-care; a cognitive intervention can change strategy use without changing NIHSS.

Swallowing illustrates the distinction: the seven-level Functional Oral Intake Scale showed high inter-rater reliability and sensitivity to change in 302 acute-stroke patients, but oral intake is not identical to aspiration physiology or pneumonia risk (Crary 2005, PMID 16084801).

Patient-reported and participation outcomes

Patient-reported outcome measurement is vulnerable to systematic exclusion: severe aphasia, cognitive impairment, institutional residence, and lack of proxy procedures remove people with the greatest burden from many cohorts.

Return to work is both an outcome and a systems measure. A 39-study systematic review found younger age, higher education, white-collar work, independence in activities, and lower neurological severity associated with return, but definitions and assessment times varied substantially (Orange 2024, PMID 37797913).

Work status should be reported with denominator, pre-stroke employment, hours, role modification, sustainability, and reason for non-return. “Returned” can mean one hour of altered work or full restoration of occupation.

Quality of life should be stroke-specific when possible and paired with mood, fatigue, cognition, communication, pain, and participation. Proxy ratings are not interchangeable with survivor ratings but may be necessary when inability to self-report would otherwise erase severe outcomes.

Recurrence and long-term vascular outcome

In the TIAregistry.org cohort, 3,847 patients with TIA/minor stroke had five-year follow-up. The composite of stroke, acute coronary syndrome, or cardiovascular death was 12.9%; stroke occurred in 9.5%, with approximately half of recurrent strokes occurring after the first year (Amarenco 2018, PMID 29766771).

This cohort came from specialized TIA clinics with rapid contemporary care; recurrence estimates from untreated historical cohorts, administrative data, and specialized clinics should not be pooled without acknowledging setting and treatment.

Recurrence metric Required denominator/detail
90-day cumulative incidence Index definition, competing death, ascertainment intensity
Annualized event rate Person-time and whether hazards were constant
Ischemic stroke recurrence Whether hemorrhage and TIA were excluded
Composite vascular event Exact components; hospitalization thresholds
Same-territory recurrence Imaging and mechanism adjudication

Recurrence is mechanism-sensitive. AF, large-artery atherosclerosis, small-vessel disease, and unresolved embolic sources have different time courses and prevention strategies; see secondary prevention and cardioembolic and cryptogenic stroke.

Mortality, disability, and global burden

GBD 2019 estimated 12.2 million incident strokes, 101 million prevalent strokes, 6.55 million deaths, and 143 million disability-adjusted life-years worldwide in 2019. These are modeled population estimates with uncertainty intervals, not a single prospective cohort (GBD 2019 Stroke Collaborators 2021, PMID 34487721).

Burden differs from case fatality. Population aging can increase absolute prevalence and disability while age-standardized rates fall. Country income, access to acute care, rehabilitation, prevention, and vital registration determine both outcome and estimate quality (GBD 2019 Stroke Collaborators 2021, PMID 34487721).

Organized stroke-unit care improves the combined outcomes of death, dependence, and institutional care; prognosis observed in a general ward is therefore not a fixed property of the stroke (Langhorne 2020, PMID 32324916).

Cognitive outcome and dementia

A systematic review found post-stroke dementia prevalence varies sharply with case mix and inclusion of recurrent stroke: approximately 10% after first stroke in population-based studies and substantially higher after recurrent stroke, with age, pre-stroke cognitive decline, diabetes, AF, severity, and lesion burden contributing (Pendlebury 2009, PMID 19782001).

The label “post-stroke dementia” contains multiple pathways: strategic infarction, accumulated infarcts, small-vessel injury, concomitant neurodegeneration, delirium-related unmasking, and recurrent events. A single assessment during acute hospitalization cannot reliably distinguish them.

Cognitive outcome studies should report pre-stroke cognition, delirium, language accommodations, education, recurrent events, and attrition. Otherwise severe stroke and aphasia can be misclassified as cognitive impairment or excluded entirely.

Prognostic models in ischemic stroke

ASTRAL uses age, stroke severity, onset-to-admission delay, visual-field deficit, glucose, and consciousness available soon after admission. It was internally and externally validated for three-month unfavorable outcome, but it predates widespread thrombectomy and should be recalibrated in the treatment context in which it is used (Ntaios 2012, PMID 22649218).

iScore combines age, sex, severity, subtype, risk factors, comorbidity, preadmission disability, glucose, and care setting; in two Canadian cohorts it predicted death/dependency and institutionalization, but administrative availability does not remove transportability concerns (Saposnik 2011, PMID 21960583).

The Six Simple Variables model performed at least as well as several more complex systems and informal clinician prediction in an independent cohort. Its lesson is methodological: added variables and algorithmic complexity do not guarantee improved discrimination or calibration (Counsell 2004, PMID 14966155).

Model Target Strength Critical limitation
ASTRAL 3-mo mRS >2 after ischemic stroke Six early variables; external validation Treatment-era and population calibration (PMID 22649218)
iScore Mortality and poor functional outcome Large registry derivation/validation Depends heavily on severity and comorbidity coding (PMID 21960583)
Six Simple Variables Alive/independent at 6–12 mo Parsimonious; compared with clinicians Does not encode modern reperfusion response (PMID 14966155)

Discrimination answers whether higher-risk people rank above lower-risk people; calibration answers whether a predicted 30% risk is actually near 30%. A model can have good discrimination and dangerous calibration in a new hospital or treatment era.

Hemorrhage prognosis and self-fulfilling prophecy

The ICH Score assigns points for Glasgow Coma Scale, age ≥80, infratentorial origin, ICH volume ≥30 cm³, and intraventricular hemorrhage. In its derivation cohort, 30-day mortality rose monotonically with score (Hemphill 2001, PMID 11283388).

FUNC was designed to predict 90-day functional independence rather than death. In 629 patients, age, GCS, ICH location, volume, and pre-ICH cognitive impairment predicted independence; no patient with FUNC ≤4 achieved independence, whereas >80% with score 11 did (Rost 2008, PMID 18556582).

Both tools describe cohorts; neither establishes futility. Early do-not-resuscitate orders and withdrawal of life-sustaining treatment can turn pessimistic prediction into observed mortality. The 2022 ICH guideline cautions against using baseline severity scores as the sole basis for limiting life-sustaining treatment and recommends separating severity communication from individual treatment decisions (Greenberg 2022, PMID 35579034).

Communicating prognosis

Good prognostic communication states the outcome, time horizon, uncertainty, and conditions under which the estimate applies. “Poor prognosis” is not measurable.

Weak statement Research-grade replacement
“Likely to recover” “Among comparable treated patients, X% reached mRS 0–2 by 90 days; this patient differs in A, B, and C.”
“No meaningful recovery” State the domain, assessment time, preserved functions, and uncertainty; avoid converting a score into a value judgment
“The scan is devastating” Describe lesion volume/location, mass effect, hemorrhage, and the evidence connecting each to outcome
“The score says mortality is high” Name the model, horizon, calibration population, and why it must not determine withdrawal alone

Serial prognosis is preferable to a single early declaration. Reperfusion response, edema, hemorrhagic transformation, infection, delirium, hydrocephalus, recurrent stroke, and emerging purposeful behavior update the estimate.

Open questions

  • Which outcome set best captures survival with acceptable participation without excluding people with aphasia or cognitive impairment (Kasner 2006, PMID 16781990)?
  • Can mRS adjudication be made both reliable and sensitive to outcomes patients value across cultures (Banks 2007, PMID 17272767; Saver 2021, PMID 34320814)?
  • How should prognostic models be dynamically recalibrated after reperfusion response and early complications rather than frozen at admission (Ntaios 2012, PMID 22649218)?
  • Can ICH prognosis be estimated without contamination by early treatment limitation and institutional practice (Hemphill 2001, PMID 11283388; Greenberg 2022, PMID 35579034)?
  • What explains late recurrence after apparently successful early TIA/minor-stroke management, when nearly half of five-year recurrent strokes occur after year one (Amarenco 2018, PMID 29766771)?
  • Which measures distinguish true restitution from compensation and adaptation in recovery trials (Bernhardt 2017, PMID 28697708)?

References

  1. Kasner SE. Clinical interpretation and use of stroke scales. Lancet Neurol. 2006;5:603-612. PMID 16781990
  2. Banks JL, Marotta CA. Outcomes validity and reliability of the modified Rankin scale: implications for stroke clinical trials. Stroke. 2007;38:1091-1096. PMID 17272767
  3. Saver JL, Chaisinanunkul N, Campbell BCV, et al. Standardized nomenclature for modified Rankin Scale global disability outcomes. Stroke. 2021;52:3054-3062. PMID 34320814
  4. Kothari RU, Pancioli A, Liu T, et al. Cincinnati Prehospital Stroke Scale: reproducibility and validity. Ann Emerg Med. 1999;33:373-378. PMID 10092713
  5. Berkhemer OA, Fransen PSS, Beumer D, et al. A randomized trial of intraarterial treatment for acute ischemic stroke. N Engl J Med. 2015;372:11-20. PMID 25517348
  6. Goyal M, Menon BK, van Zwam WH, et al. Endovascular thrombectomy after large-vessel ischaemic stroke: meta-analysis of individual patient data. Lancet. 2016;387:1723-1731. PMID 26898852
  7. Nogueira RG, Jadhav AP, Haussen DC, et al. Thrombectomy 6 to 24 hours after stroke with a mismatch between deficit and infarct. N Engl J Med. 2018;378:11-21. PMID 29129157
  8. Crary MA, Mann GDC, Groher ME. Initial psychometric assessment of a functional oral intake scale for dysphagia in stroke patients. Arch Phys Med Rehabil. 2005;86:1516-1520. PMID 16084801
  9. Orange C, Lanhers C, Coll G, et al. Determinants of return to work after a stroke: a systematic review and meta-analysis. Arch Phys Med Rehabil. 2024;105:359-368. PMID 37797913
  10. Amarenco P, Lavallée PC, Monteiro Tavares L, et al. Five-year risk of stroke after TIA or minor ischemic stroke. N Engl J Med. 2018;378:2182-2190. PMID 29766771
  11. Pendlebury ST, Rothwell PM. Prevalence, incidence, and factors associated with pre-stroke and post-stroke dementia: systematic review and meta-analysis. Lancet Neurol. 2009;8:1006-1018. PMID 19782001
  12. GBD 2019 Stroke Collaborators. Global, regional, and national burden of stroke and its risk factors, 1990–2019. Lancet Neurol. 2021;20:795-820. PMID 34487721
  13. Langhorne P, Ramachandra S. Organised inpatient (stroke unit) care for stroke: network meta-analysis. Cochrane Database Syst Rev. 2020;4:CD000197. PMID 32324916
  14. Ntaios G, Faouzi M, Ferrari J, et al. An integer-based score to predict functional outcome in acute ischemic stroke: the ASTRAL score. Neurology. 2012;78:1916-1922. PMID 22649218
  15. Saposnik G, Raptis S, Kapral MK, et al. The iScore predicts poor functional outcomes early after hospitalization for an acute ischemic stroke. Stroke. 2011;42:3421-3428. PMID 21960583
  16. Counsell C, Dennis M, McDowall M. Predicting functional outcome in acute stroke: comparison of a six-variable model with other systems and informal prediction. J Neurol Neurosurg Psychiatry. 2004;75:401-405. PMID 14966155
  17. Hemphill JC 3rd, Bonovich DC, Besmertis L, et al. The ICH score: a simple, reliable grading scale for intracerebral hemorrhage. Stroke. 2001;32:891-897. PMID 11283388
  18. Rost NS, Smith EE, Chang Y, et al. Prediction of functional outcome in patients with primary intracerebral hemorrhage: the FUNC score. Stroke. 2008;39:2304-2309. PMID 18556582
  19. Greenberg SM, Ziai WC, Cordonnier C, et al. 2022 Guideline for the Management of Patients With Spontaneous Intracerebral Hemorrhage. Stroke. 2022;53:e282-e361. PMID 35579034
  20. Bernhardt J, Hayward KS, Kwakkel G, et al. Agreed definitions and a shared vision for new standards in stroke recovery research. Int J Stroke. 2017;12:444-450. PMID 28697708