Skip to content

Stroke units and systems of care

TL;DR — Stroke outcomes depend on an end-to-end system: recognition and emergency activation, pre-notification and destination choice, immediate imaging and reperfusion, organized stroke-ward care, and measured transition to rehabilitation. The strongest organizational intervention is a geographically discrete multidisciplinary stroke ward: across 29 randomized trials (5,902 participants), organized inpatient care reduced death (OR 0.76, 95% CI 0.66–0.88) and death or dependency (OR 0.75, 95% CI 0.66–0.85), equivalent to about six additional people per 100 living independently at follow-up (Langhorne 2020, PMID 32324916). Mobile stroke units shorten onset-to-thrombolysis by about 31 minutes and improve the odds of excellent 90-day outcome, but their cost-effectiveness depends heavily on dispatch density and local geography (Turc 2022, PMID 35129584; Lund 2022, PMID 35862205). For suspected large-vessel occlusion, bypassing a nearer thrombolysis-capable hospital may accelerate thrombectomy but delay intravenous thrombolysis and expose false-positive patients to longer transport; no universal destination rule can replace region-specific modeling (Romoli 2020, PMID 33053947; Ciccone 2019, PMID 30303811). Certification and dashboards are means rather than endpoints: measured process improvement can change outcomes, while a center label alone does not eliminate case-mix, transfer, or access inequities (Fonarow 2014, PMID 24756513; Man 2018, PMID 29794035).

The care chain and its failure modes

Link Operational objective Failure that the metric should reveal Practical measures
Public recognition Convert symptom onset into an emergency call Waiting for symptoms to resolve; private transport; missed posterior-circulation signs onset/last-known-well-to-call; EMS use; stroke-code positive predictive value
Dispatch and EMS assessment Identify probable stroke without delaying transport Under-recognition; excessive on-scene examination; wrong destination call-to-dispatch; dispatch-to-scene; scene time; validated screen completion
Pre-notification and destination Alert a capable receiving team and choose the least harmful route No prenotification; inappropriate bypass; preventable secondary transfer prenotification rate; direct-to-capable-center rate; extra transport time
Emergency department Parallel registration, neurologic assessment, glucose, imaging, laboratory work and consent Serial workflow; imaging queue; treatment awaiting nonessential tests door-to-imaging; door-to-needle; door-to-puncture
Transfer Treat eligible patients locally while arranging definitive capability delayed acceptance; repeated imaging; low-priority transport; long door-in-door-out door-in-door-out; first-door-to-puncture; image-sharing success
Stroke ward Deliver protocolized multidisciplinary care and prevent complications admission to a general ward; fragmented nursing, therapy and medical plans proportion admitted to stroke unit; swallowing screen; mobilization and therapy assessment
Discharge transition Match rehabilitation and prevention to disability and mechanism no medication reconciliation; absent follow-up; inequitable rehabilitation access destination; prevention bundle; follow-up booked; patient-reported transition quality

The biological urgency behind the process measures is not metaphorical. A model of a typical untreated large-vessel supratentorial infarct estimated loss of 1.9 million neurons per minute, although the estimate is a population model rather than an individual-patient measurement (Saver 2006, PMID 16339467). Time metrics therefore need a common clock definition, explicit denominator and balancing safety measure; otherwise a faster median can conceal excluded patients, missing timestamps, or unsafe shortcuts.

Organized inpatient stroke-unit care

The defining intervention is not simply a bed labelled “stroke.” It is coordinated care by a multidisciplinary team using agreed protocols for physiological monitoring, swallowing and nutrition, complications, early rehabilitation, communication and discharge planning. The 2020 Cochrane network meta-analysis found the benefit concentrated in discrete stroke wards; compared with a general ward, stroke-ward care reduced poor outcome (OR 0.78, 95% CI 0.68–0.91), whereas a mobile consult team alone had an imprecise estimate (OR 0.80, 95% CI 0.52–1.22) (Langhorne 2020, PMID 32324916).

Comparison at median one-year follow-up Trials / participants Effect Certainty / limitation
Any organized inpatient service vs alternative 29 / 5,902 poor outcome OR 0.77 (95% CI 0.69–0.87) moderate; trials span eras and service configurations
Any organized service vs alternative 29 / 5,902 death OR 0.76 (95% CI 0.66–0.88) moderate
Any organized service vs alternative 29 / 5,902 death or dependency OR 0.75 (95% CI 0.66–0.85) moderate
Discrete stroke ward vs general ward 15 / 3,523 poor outcome OR 0.78 (95% CI 0.68–0.91) moderate
Mobile stroke team vs general ward 2 / 438 poor outcome OR 0.80 (95% CI 0.52–1.22) low; compatible with benefit or no benefit
Mixed rehabilitation ward vs general ward 6 / 630 poor outcome OR 0.65 (95% CI 0.47–0.90) moderate

The absolute translation was approximately two additional survivors, six additional people living at home, and six additional people living independently per 100 treated in organized care. Effects appeared across age, sex, severity and stroke type; length-of-stay evidence was heterogeneous (I²=85%) and did not show a systematic increase (Langhorne 2020, PMID 32324916). These trials establish the ward model, not the marginal effect of every component. Staffing ratios, seven-day therapy, nurse education and monitoring intensity remain service-design variables rather than independently randomized ingredients.

Prehospital recognition and notification

Prehospital scales serve two different tasks: recognizing any stroke and predicting large-vessel occlusion (LVO). Reviews of LVO scales found heterogeneous thresholds, incomplete prehospital validation and a sensitivity–specificity tradeoff that directly changes bypass decisions (Vidale 2018, PMID 29430622; Krebs 2018, PMID 29023166). In a prospective multicenter validation cohort of 1,316 patients with a 35.8% LVO prevalence, FAST-ED and RACE both had AUROC above 0.81, but performance in an enriched cohort does not specify the best regional threshold (Koster 2019, PMID 30209989).

Design choice Benefit Cost / hazard Evaluation requirement
High-sensitivity LVO threshold fewer LVOs left at non-thrombectomy hospitals more stroke mimics and non-LVO strokes bypass local care report sensitivity, specificity, PPV and transport penalty
High-specificity threshold fewer unnecessary bypasses more secondary transfers among missed LVOs report first-door-to-reperfusion and missed-LVO rate
Dispatcher stroke screen earlier stroke-coded response false-positive dispatch and resource use analyze all dispatches, not only confirmed stroke
EMS pre-notification receiving team can mobilize before arrival alert fatigue if criteria are broad or unreliable door-to-imaging/needle plus false-alert rate

A meta-analysis of 86 workflow studies encompassing 17,665 thrombolysis cases associated new transport protocols with greater thrombolysis use (OR 1.45, 95% CI 1.23–1.71), education/training with OR 1.38 (95% CI 1.11–1.73), and comprehensive prehospital stroke codes with OR 1.83 (95% CI 1.44–2.32) (Huang 2018, PMID 29924046). These are largely before–after organizational data: they show implementability and association, not clean isolation of a single component. A systematic review of 128 studies likewise identified EMS activation, prenotification, ambulance transport, pathway protocols and geographic access as recurrent determinants of reperfusion access (Botelho 2022, PMID 36498429).

Telestroke as a network capability

Telestroke connects a site without continuous on-site vascular-neurology expertise to remote assessment and treatment support. Its value is access and decision equivalence, not superiority over an in-person specialist already available. A 2024 meta-analysis including 12,540 patients found no significant difference between telestroke and conventional care in good functional outcome (mRS 0–2: OR 1.06, 95% CI 0.89–1.29), 90-day mortality (OR 1.16, 95% CI 0.94–1.43), or symptomatic intracranial hemorrhage (OR 0.99, 95% CI 0.73–1.34) (Mohamed 2024, PMID 37752674).

Implementation quality depends on connection reliability, a trained bedside examiner, rapid image transfer, explicit treatment authority and a transfer pathway. A systematic review of French telestroke networks found heterogeneous organization and outcome reporting, illustrating why “telemedicine available” is not an adequate quality variable (Ohannessian 2020, PMID 32147201). Tele-neurology can also substitute for an on-board neurologist in a mobile unit without adding treatment delay in a feasibility comparison, potentially changing staffing economics (Wu 2017, PMID 28082671).

Mobile stroke units

Mobile stroke units (MSUs) move CT imaging, point-of-care testing, stroke expertise and thrombolysis into the ambulance. They compress onset-to-treatment rather than only door-to-treatment.

Evidence Population / design Quantitative result Boundary
PHANTOM-S substudy 6,182 dispatch episodes; controlled availability weeks golden-hour thrombolysis 31.0% with MSU vs 4.9% usual care; discharge home aOR 1.93 (95% CI 1.09–3.41) among golden-hour-treated patients Berlin infrastructure; outcome comparison by treatment time is observational (Ebinger 2015, PMID 25402214)
B_PROUD 1,543 eligible ischemic events; prospective nonrandomized dispatch comparison common OR for worse 3-month mRS 0.71 (95% CI 0.58–0.86) availability-based exposure; Berlin (Ebinger 2021, PMID 33528537)
BEST-MSU 1,515 enrolled; alternating-week multicenter trial in tPA-eligible patients, onset-to-tPA 72 vs 108 min; mRS 0–1 55.0% vs 44.4%; adjusted OR for utility-weighted excellent outcome 2.43 (95% CI 1.75–3.36) open system intervention; US metropolitan sites (Grotta 2021, PMID 34496173)
Meta-analysis 13 time studies / 3,322 participants; 5 excellent-outcome studies / 3,228 onset-to-IVT −31 min (95% CI −39 to −23); excellent outcome aOR 1.64 (95% CI 1.27–2.13) mixes randomized and nonrandomized designs (Turc 2022, PMID 35129584)
Norwegian economic model lifetime model using local trial inputs +0.065 QALY/patient; about US$43,780/QALY at 260 ischemic strokes treated annually sharply volume- and system-dependent (Lund 2022, PMID 35862205)

The implementation question is therefore not “do MSUs save time?” but where one unit, crew and scanner produce greater population benefit than alternative investments. Reviews emphasize density, operating hours, dispatch specificity, maintenance, staffing and integration with thrombectomy routing (Bowry 2021, PMID 33511604; Alexandrov 2021, PMID 34384228). Cost per QALY should not be transported between cities without recalculating utilization, travel, wages, treatment mix and long-term care costs.

Destination strategy for suspected LVO

Two archetypes dominate. “Mothership” sends the patient directly to a thrombectomy-capable comprehensive center. “Drip-and-ship” sends the patient to the nearest thrombolysis-capable center, starts intravenous thrombolysis if eligible, then transfers confirmed LVO. The choice is a competing-delay problem.

Route Time gained Time lost Patients most exposed to error
Mothership avoids secondary transfer and repeat handoffs before thrombectomy delays IV thrombolysis when the comprehensive center is farther false-positive LVO screens; hemorrhage; mimics; non-LVO ischemic stroke
Drip-and-ship preserves rapid local imaging and thrombolysis adds door-in-door-out and transport before thrombectomy confirmed LVO, especially when local processing is slow
Mobile imaging / MSU triage establishes diagnosis before destination and can begin thrombolysis high fixed cost and limited geographic availability low-density systems with few eligible dispatches

A systematic review and meta-analysis found better functional and radiological outcomes with direct admission than transfer, but explicitly warned that nonrandomized comparisons and baseline differences limit causal interpretation (Romoli 2020, PMID 33053947). An earlier systematic review concluded that organizational-model evidence was heterogeneous and insufficient to establish one universal configuration (Ciccone 2019, PMID 30303811). In DEFUSE 3 late-window participants, transferred and directly presenting patients still benefited from thrombectomy, showing that transfer does not erase benefit when imaging selection identifies eligible tissue; it does not show that transfer delay is harmless (Sarraj 2019, PMID 30734042).

Transfer performance should be measured from the first emergency call, not reset at the receiving hospital. In one Dutch regional cohort of 198 transferred LVO patients, median call-to-comprehensive-center time was 162 minutes (IQR 137–190); highest-urgency ambulance dispatch for transfer was associated with 27.6 minutes shorter time (95% CI 3.9–51.2) (van Meenen 2021, PMID 34512538). A smartphone-supported Egyptian network reported door-in-door-out 56±34 vs 96±45 minutes and door-to-groin 50±7 vs 120±25 minutes with versus without the application, but its nonrandomized 84-vs-276-patient comparison cannot distinguish the app from concurrent pathway change (Mansour 2021, PMID 33708169).

Inside the receiving hospital

Fast pathways use parallel work: stroke-team activation before arrival, direct movement to CT, treatment where the patient is imaged, premixed thrombolytic processes, rapid weight acquisition, and early thrombectomy-team activation. In Target: Stroke, 71,169 alteplase-treated patients at 1,030 hospitals were studied before and after a national quality initiative. Median door-to-needle time fell from 77 to 67 minutes; treatment within 60 minutes rose from 26.5% to 41.3%; adjusted in-hospital mortality fell (OR 0.89, 95% CI 0.83–0.94), with the usual limitation that secular trends can accompany before–after improvement (Fonarow 2014, PMID 24756513).

Later registry analysis linked specific strategies—rapid triage, stroke-team notification, single-call activation, direct-to-CT and premixing—to faster treatment, supporting bundles rather than a single “magic” step (Xian 2017, PMID 28096207). A single-center bundle demonstrated that a <30-minute median is operationally possible, but generalizability depends on staffing and case mix (Bhatt 2019, PMID 30671160). Registry data also show that every avoidable component matters: longer door-to-needle intervals translate into longer overall treatment time and worse outcomes (Kamal 2017, PMID 28228574).

Direct-to-angiography-suite pathways bypass conventional cross-sectional imaging in carefully selected suspected LVO. Across eight studies (1,938 patients), they shortened median door-to-groin by 29.0 minutes (95% CI 14.3–43.6) and door-to-reperfusion by 32.1 minutes (95% CI 15.1–49.1), but did not establish improved functional outcome (good-outcome OR 1.38, 95% CI 0.97–1.95) (Brehm 2022, PMID 35251309). The process gain is real; the patient benefit and safe selection strategy remain less certain.

Center designation and regionalization

Stroke-center designation bundles infrastructure, protocols, expertise, data submission and quality review. In New York, instrumental-variable analysis of 30,947 ischemic strokes associated designated-center admission with lower 30-day mortality (10.1% vs 12.5%; adjusted difference −2.5%, 95% CI −3.6 to −1.4) and more thrombolysis (4.8% vs 1.7%; adjusted difference +2.2%, 95% CI 1.6–2.8) (Xian 2011, PMID 21266684). A national Medicare analysis found only a small average mortality difference between certified and noncertified hospitals (10.7% vs 11.0%) and similar readmission, with substantial overlap between categories (Lichtman 2011, PMID 21543736).

Center level is confounded by referral severity. Among 722,941 patients in a US quality registry, comprehensive centers used IV thrombolysis more often (14.3% vs 10.3%) and thrombectomy more often (4.1% vs 1.0%) than primary centers, and had shorter median door-to-IV-tPA time (52 vs 61 minutes); overall adjusted in-hospital mortality was nevertheless higher at comprehensive centers, while outcomes among reperfusion-treated patients were comparable (Man 2018, PMID 29794035). A label should therefore be interpreted with transfer-adjusted case mix and process data, not as a hospital league table.

Implementation outside high-income metropolitan networks is possible but resource-specific. A Panama stroke-center report documents staged development of protocols, training, imaging and data capture as a model for a middle-income setting; it is implementation evidence, not a comparative effectiveness trial (Novarro-Escudero 2021, PMID 34484099). The central equity metric is not merely national center count but the proportion of the population able to reach stroke-unit, thrombolysis and thrombectomy capability within clinically meaningful time.

A minimum measurement set

Domain Numerator / denominator Report with Balancing measure
EMS recognition confirmed strokes coded as stroke / confirmed strokes transported by EMS stroke type, severity, time of day false-positive stroke dispatches
Pre-notification pre-notified arrivals / EMS stroke arrivals receiving center and shift alert-to-arrival interval; false alerts
Imaging arrivals imaged within target / eligible arrivals direct vs transfer; contrast use missed hemorrhage or alternative diagnosis
Thrombolysis treated within 30 and 60 min / all treated; and / all eligible drug, onset window, contraindication reasons symptomatic ICH; protocol deviations
Thrombectomy door-to-puncture and door-to-reperfusion direct vs transferred; occlusion site futile transfer; complications
Transfer door-in-door-out; first-door-to-puncture distance, transport mode, acceptance time thrombolysis delay; cancelled transfers
Stroke-unit access stroke-unit admissions / all acute strokes ischemic, ICH, SAH; age; severity boarding time; general-ward deaths
Outcome ordinal 90-day mRS and mortality prestroke disability and follow-up completeness loss to follow-up; destination
Equity each measure stratified by geography and sociodemographic group population travel-time denominator widening absolute or relative gap

Dashboards should show distributions, not only medians; the 90th percentile detects patients stranded by nights, weekends or transfer queues. Denominators should include untreated eligible patients and suspected strokes later found to be mimics. Process improvement becomes credible when faster time is accompanied by stable or improved hemorrhage, mortality, disability and diagnostic-error outcomes (Fonarow 2014, PMID 24756513).

Evidence boundaries

  • Stroke-unit evidence is randomized and patient-centered, but much of it predates contemporary thrombectomy and current hospital organization (Langhorne 2020, PMID 32324916).
  • MSU evidence consistently shows time gain and now supports functional benefit, but most outcome evidence comes from dense urban systems with mature stroke networks (Grotta 2021, PMID 34496173; Ebinger 2021, PMID 33528537).
  • Bypass and transfer comparisons are vulnerable to geography, selection and referral bias; local travel-time distributions can reverse the preferred route (Romoli 2020, PMID 33053947; Ciccone 2019, PMID 30303811).
  • Certification studies are observational and labels identify bundles; they do not reveal which component causes benefit (Xian 2011, PMID 21266684; Lichtman 2011, PMID 21543736).
  • Most workflow studies report treatment time or treatment rate; fewer capture 90-day disability, patient experience, caregiver burden, missed diagnoses or equity (Botelho 2022, PMID 36498429).

Open questions

  • Which combination of dispatch algorithm, operating hours and base location maximizes disability-adjusted benefit per MSU-year outside high-density cities (Turc 2022, PMID 35129584; Lund 2022, PMID 35862205)?
  • Can destination policies be randomized or stepped-wedge at regional level using first-medical-contact-to-reperfusion and ordinal 90-day mRS, while retaining all false-positive bypasses in the denominator (Romoli 2020, PMID 33053947; Ciccone 2019, PMID 30303811)?
  • Which stroke-ward components account for benefit in contemporary mixed ischemic/hemorrhagic populations, and what minimum staffing model preserves it (Langhorne 2020, PMID 32324916)?
  • Does direct-to-angiography improve disability rather than only treatment time, and what is its diagnostic-harm rate among mimics and hemorrhage (Brehm 2022, PMID 35251309)?
  • Which quality metrics remain valid across high-, middle- and low-resource systems without rewarding exclusion of difficult patients (Novarro-Escudero 2021, PMID 34484099; Botelho 2022, PMID 36498429)?
  • Can telestroke networks demonstrate population-level access gains while reporting connection failure, transfer delay and 90-day outcomes rather than consultation volume alone (Mohamed 2024, PMID 37752674; Ohannessian 2020, PMID 32147201)?

References

  1. Langhorne P, Ramachandra S; Stroke Unit Trialists' Collaboration. Organised inpatient (stroke unit) care for stroke: network meta-analysis. Cochrane Database Syst Rev. 2020;4:CD000197. PMID 32324916
  2. Saver JL. Time is brain—quantified. Stroke. 2006;37:263-266. PMID 16339467
  3. Huang Q, Zhang JZ, Xu WD, Wu J. Generalization of the right acute stroke promotive strategies in reducing delays of intravenous thrombolysis for acute ischemic stroke: a meta-analysis. Medicine (Baltimore). 2018;97:e11205. PMID 29924046
  4. Mohamed A, et al. Is telestroke more effective than conventional treatment for acute ischemic stroke? A systematic review and meta-analysis of patient outcomes and thrombolysis rates. Int J Stroke. 2024;19:280-292. PMID 37752674
  5. Ohannessian R, et al. Acute telestroke in France: a systematic review. Rev Neurol (Paris). 2020;176:316-324. PMID 32147201
  6. Koster GT, et al. Clinical prediction of thrombectomy eligibility: a systematic review and 4-item decision tree. Int J Stroke. 2019;14:530-539. PMID 30209989
  7. Vidale S, et al. Prehospital stroke scales and large vessel occlusion: a systematic review. Acta Neurol Scand. 2018;138:24-31. PMID 29430622
  8. Krebs W, et al. Prehospital Stroke Assessment for Large Vessel Occlusions: a systematic review. Prehosp Emerg Care. 2018;22:180-188. PMID 29023166
  9. Sarraj A, et al. Outcomes of thrombectomy in transferred patients with ischemic stroke in the late window: a subanalysis from the DEFUSE 3 trial. JAMA Neurol. 2019;76:682-689. PMID 30734042
  10. Romoli M, et al. Mothership versus drip-and-ship model for mechanical thrombectomy in acute stroke: a systematic review and meta-analysis for clinical and radiological outcomes. J Stroke. 2020;22:317-323. PMID 33053947
  11. Ciccone A, et al. Systematic review of organizational models for intra-arterial treatment of acute ischemic stroke. Int J Stroke. 2019;14:12-22. PMID 30303811
  12. Brehm A, et al. Direct to angiography suite approaches for the triage of suspected acute stroke patients: a systematic review and meta-analysis. Ther Adv Neurol Disord. 2022;15:17562864221078177. PMID 35251309
  13. Botelho A, et al. Organizational factors determining access to reperfusion therapies in ischemic stroke—systematic literature review. Int J Environ Res Public Health. 2022;19:16357. PMID 36498429
  14. Lund UH, et al. Cost-effectiveness of mobile stroke unit care in Norway. Stroke. 2022;53:3173-3181. PMID 35862205
  15. Alexandrov AV, Nilanont Y. Improving outcomes after stroke: from stroke units to mobile stroke units. Stroke. 2021;52:3072-3074. PMID 34384228
  16. Bowry R, Grotta JC. Mobile stroke units: current and future impact on stroke care. Semin Neurol. 2021;41:9-15. PMID 33511604
  17. van Meenen LCC, et al. Pre- and interhospital workflow times for patients with large vessel occlusion stroke transferred for endovascular thrombectomy. Front Neurol. 2021;12:730250. PMID 34512538
  18. Mansour OY, et al. Using ESN-smartphone application to maximize AIS reperfusion therapy in Alexandria Stroke Network: a stroke chain of survival organizational model. Front Neurol. 2021;12:597717. PMID 33708169
  19. Bhatt N, et al. Target Stroke: best practice strategies cut door to thrombolysis time to <30 minutes in a large urban academic comprehensive stroke center. Neurohospitalist. 2019;9:22-25. PMID 30671160
  20. Kamal N, et al. Delays in door-to-needle times and their impact on treatment time and outcomes in Get With The Guidelines-Stroke. Stroke. 2017;48:946-954. PMID 28228574
  21. Xian Y, et al. Use of strategies to improve door-to-needle times with tissue-type plasminogen activator in acute ischemic stroke in clinical practice: findings from Target: Stroke. Circ Cardiovasc Qual Outcomes. 2017;10:e003227. PMID 28096207
  22. Grotta JC, et al. Prospective, multicenter, controlled trial of mobile stroke units. N Engl J Med. 2021;385:971-981. PMID 34496173
  23. Ebinger M, et al. Association between dispatch of mobile stroke units and functional outcomes among patients with acute ischemic stroke in Berlin. JAMA. 2021;325:454-466. PMID 33528537
  24. Turc G, et al. Comparison of mobile stroke unit with usual care for acute ischemic stroke management: a systematic review and meta-analysis. JAMA Neurol. 2022;79:281-290. PMID 35129584
  25. Ebinger M, et al. Effects of golden hour thrombolysis: a Prehospital Acute Neurological Treatment and Optimization of Medical Care in Stroke substudy. JAMA Neurol. 2015;72:25-30. PMID 25402214
  26. Xian Y, et al. Association between stroke center hospitalization for acute ischemic stroke and mortality. JAMA. 2011;305:373-380. PMID 21266684
  27. Lichtman JH, et al. Outcomes after ischemic stroke for hospitals with and without Joint Commission-certified primary stroke centers. Neurology. 2011;76:1976-1982. PMID 21543736
  28. Fonarow GC, et al. Door-to-needle times for tissue plasminogen activator administration and clinical outcomes in acute ischemic stroke before and after a quality improvement initiative. JAMA. 2014;311:1632-1640. PMID 24756513
  29. Man S, et al. Comparison of acute ischemic stroke care and outcomes between comprehensive stroke centers and primary stroke centers in the United States. Circ Cardiovasc Qual Outcomes. 2018;11:e004512. PMID 29794035
  30. Novarro-Escudero N, et al. Organization and implementation of a stroke center in Panamá, a model for implementation of stroke centers in low and middle income countries. Front Neurol. 2021;12:684775. PMID 34484099
  31. Wu TC, et al. Telemedicine can replace the neurologist on a mobile stroke unit. Stroke. 2017;48:493-496. PMID 28082671