A total of 4,870 (2,886 after deduplication) studies were identified via the search strategy (Fig. 1). One study was identified via citation searches. 94 potentially eligible articles were retained following initial screening. After obtaining and reviewing full texts (or abstracts where full texts were not available), 75 studies were excluded with reasons detailed in Fig. 1. In total, 19 studies were included, 8 of which were full text articles, 7 conference abstracts and 4 published protocols.
Blood biomarkers
We identified protocols for two currently active, large pre-hospital studies of blood biomarker technologies which are both prospective diagnostic accuracy designs [21, 22], summarised in Table 2.
Table 2
Descriptions of Biomarker Technologies and Outcomes
Study Details | Portability | Expertise and training requirements | Person interpreting output | Location (on scene; stationary ambulance; in transit) | Purpose, diagnostic accuracy, comparator and use of clinical scale | Physical invasiveness and time to acquire results | Acceptability: clinicians and/or patients | Impact on EMS clinician decisions or treatment provision | Impact on process (time metrics) or patient outcomes | Costs |
Helsinki Ultra-acute Stroke Biomarker Study Lindsberg 2018[21] Sub-type: Proteins | Portable with use of a vehicle | Expertise: EMS clinician Training: not reported | EMS clinician | In transit | Purpose: Diagnostic -Mimic -TIA -Ischaemia -Haemorrhage Diagnostic accuracy: Results not yet available Comparator: Definitive diagnosis Clinical Scale: none | Invasive Time to acquire results: not yet available | Not assessed | Destination: Results not yet available Treatment: Results not yet available | Time metrics: Not yet available Patient outcomes: will be evaluated using 3-month modified Rankin Scale (mRS) | Not reported |
Purines for Rapid Identification of Stroke Mimics (PRISM) Shaw et al. 2019[22] Sub-type: Purines (Metabolite) | Highly portable (hand-held) | Expertise: EMS clinician Training: Required | EMS clinician | On scene Stationary In transit | Purpose: Diagnostic -Mimic -TIA -Ischaemia -Haemorrhage Diagnostic accuracy: Results not yet available Comparator: Expert clinical opinion informed by brain imaging and/or other investigations Clinical scale: FAST | Invasive Time to acquire results: ~3–5 minutes | Results not yet available | Destination: Results not yet available Treatment: Results not yet available | Not yet available | Not reported |
The Helsinki Ultra-acute Stroke Biomarker Study [21] is an early stage, single-centre study aiming to establish diagnostic and predictive biomarkers for potential IVT candidates. The specific aims are to (i) identify ischemic stroke, transient ischemic attack, intracerebral haemorrhage and stroke mimics; (ii) identify patients not responding to IVT; (iii) identify patients with increased chance of IVT-related complications; and (iv) predict 90-day patient health outcomes using the modified Rankin Scale (mRS). The study will evaluate known stroke biomarkers, e.g. Glial Fibrillary Acidic Protein (GFAP) and NR2 peptide, and explore novel markers (during a discovery phase) via blood samples taken by EMS clinicians during transit. This study has yet to report on primary outcomes, although a precursor study [40] has reported the feasibility of implementing pre-hospital EMS biomarker sampling using a cannula adapter technique. It is not clear how EMS clinicians would obtain a result on scene in future (e.g. a point of care assay system) even if the panel is of value.
Purines for Rapid Identification of Stroke Mimics (PRISM) [22] is an early stage, multi-centre study aiming to assess the diagnostic accuracy of whole blood purine concentration using capillary blood sampling performed by trained EMS clinicians within 4 hours of stroke symptom onset. The goal is to differentiate between stroke and stroke mimics, with a hospital sub-study investigating LVO. Purines are measured using the point of care SMARTChip system, a hand-held reader and disposable biosensor, developed by Sarissa Biomedical Ltd in the UK, which measures purines in a finger-prick blood sample. Purines are a by-product of cellular metabolism, which accumulate rapidly during hypoxia (as occurs in stroke) and can be reliably detected in systemic arterial blood [41]. Results can be obtained within 3–5 minutes and paramedics will require training to use this technology [42]. There is currently no published diagnostic accuracy or patient outcome data, however the technology is at a late stage of development which would facilitate deployment if of value. Neither biomarker protocol commented on the potential cost of the technologies.
Pre-hospital Imaging
We identified two studies of non-invasive pre-hospital imaging technology [23, 24]. These are summarised in Table 3. EEG is included here as ‘imaging’ since the intention is to produce information which correlates anatomically with cerebral tissue injury. From a study design perspective, these are prospective diagnostic accuracy studies.
Table 3
Descriptions of Pre-hospital Imaging Technologies and Outcomes
Study Details | Portability and Resolution | Expertise and training requirements | Person interpreting output | Location (on scene; stationary ambulance; in transit) | Purpose, Diagnostic accuracy, comparator and clinical scale | Physical invasiveness and time to acquire results | Acceptability: clinicians and/or patients | Impact on EMS clinician decisions or treatment provision | Impact on process (time metrics) or patient outcomes | Costs |
“Handheld Infra-red screening device” Murphy et al. 2015[23] Electromagnetic Spectroscopy – Light
FDA approved; CE Marked; ISO certification |
Highly portable (hand-held) Low resolution | Expertise: Qualified Paramedic Training: required | EMS clinician | In transit | Purpose: Diagnostic -Stroke -Mimic -Haemorrhage Diagnostic accuracy: Haemorrhage vs Mimic False Positive Rate: 15 False Negative Rate: 2 True Positive Rate: 5 True Negative Rate: 10 Haemorrhage vs Ischaemia False Positive Rate: 8 False Negative Rate: 2 True Positive Rate: 5 True Negative Rate: 6 Comparator: CT (hospital) scan Clinical scale: stroke assessment form | Non-invasive Time to acquire results: 2–3 minutes | Clinicians: 2–3 minutes too long (+ increased FPR if time reduced) Measurements performed during fast ambulance transport appeared to cause false positive readings Unreliable measurement for combative and epileptic patients | Destination: None Treatment: Not reported | Not reported | Not reported |
Electra-Stroke Coutinho 2019[24] EEG Waveguard™ dry electrode cap and the eego™ amplifier are both CE marked | Portable with use of a vehicle Low Resolution | Expertise: Qualified paramedic Training: Not reported | Unclear | In transit Other locations unclear | Purpose: Diagnostic -LVO-a Diagnostic accuracy: Results not yet available Comparator: Final (hospital) diagnosis Clinical scale: mRS, NIHSS | Non-invasive Time to acquire results: | Results not yet available | Destination: Results not yet available Treatment: Results not yet available | Results not yet available | Not reported |
Our review identified one small, single-centre study reporting on use of a handheld Infrared screening device in the pre-hospital setting in the USA by device-trained EMS clinicians during transit [23]. The stage of development was not reported. The purpose of the device was to discern between stroke types by detecting changes in blood flow. The authors found evidence that contralesional increases in blood flow indicate LVO. The device was compared with hospital-based CT on its ability to detect haemorrhagic stroke in 46 suspected stroke patients; 7 of these 46 patients had CT-confirmed haemorrhagic stroke. True and false positive/negative results are shown in Table 3; diagnostic accuracy was extrapolated from these figures. For haemorrhage versus mimic, sensitivity and specificity within the study population were 71% and 40% respectively. Sensitivity and specificity for haemorrhage versus ischaemia were 71% and 43% respectively. Although all haemorrhagic strokes were identified by the device, the poor specificity may limit clinical use. Paramedics considered the scan too long for stroke patients and consequent speeding up of the patient scan time increased false positives. Faster ambulance speed also increased false positive rates, and the device was difficult to use reliably with patients unable to lie still. There was no impact on destination decisions or reported change in patient outcomes.
‘Dry’ electrode cap Electroencephalography (EEG) will be used in the ELECTRA-STROKE study [24], which is a large, active multi-centre study in the Netherlands, aiming to develop and validate an algorithm for automated signal analysis to detect anterior circulation LVO in suspected ischaemic stroke patients in the pre-hospital setting. The technology is fully developed; however, development of the algorithm is in early stages. The rationale for this study is underpinned by evidence that delta activity is associated with lesion location on cerebral imaging [43]. Omitting the preparation time for ‘wet’ EEG may enable even inexperienced EMS clinicians to undertake a measurement within 5 minutes. Training requirements were not reported. Across 4 phases, algorithms will be iteratively tested and developed to maximise diagnostic accuracy using the CE-marked Waveguard™ dry electrode cap and the eego™ amplifier. The algorithm will be validated in the ambulance in a large multi-centre study. The EEG results will only be analysed in hospital so destination choice will not be assessed. It is not reported whether clinical outcomes will be assessed. Expected primary outcome completion was December 2019.
There are no data available on costs, EMS clinician decisions regarding hospital destination (stroke-specialist versus non-stroke specialist centre), impact on treatment or patient outcomes for either pre-hospital imaging device.
Mobile Telemedicine
We identified 12 mobile telemedicine technologies reported across 15 studies [25–39]. Descriptions are summarised in Table 4. The outcomes are summarised in Table 5. Eleven studies were single-centre [25, 26, 29–32, 34–37, 39], one of which was large [29] and four were moderate-to-large multi-centre studies [27, 28, 33, 38]. From a design perspective, this included: development and pilot testing (n = 2), feasibility and pilot testing (n = 3), mixed methods pilot test (n = 1), retrospective before and after (n = 1), prospective cohort (n = 4), a prospective non-randomised controlled trial (n = 1), protocol: single-centre RCT (n = 1), single-centre RCT (n = 1).
Table 4
Descriptions of Telestroke Technologies
Study Details | Communication Method | Portability, Resolution and Data Transfer Speed | Expertise and training Requirements | Person interpreting output | Location (on scene; stationary ambulance; in transit) | Costs |
TeleBAT LaMonte et al. 2000[25] Xiao et al. 2000[26] | Unidirectional video and data transfer | Portable with use of a vehicle Resolution: Low (VGA) Transmission speed: Slow (2G) | Expertise: Qualified Paramedic Training: required | Remote physician | Stationary In transit | $20,000-$25,000 (~£14,000 - £17,000) in year 2000 + ‘operating cost of 4 cell phones’ |
‘peeq-box’
Bergrath et al. 2012[27] | Bidirectional video and data transfer | Portable with use of a vehicle Resolution: Low (VGA) Transmission speed: Slow to moderate (2G-3G) | Expertise: Other ambulance staff (EMS physicians) Training: required | Remote physician | Stationary In transit | Not reported |
Stroke Angel Ziegler et al. 2008[28] Rashid et al. 2015[29] | Clinical data transfer No audio or video | Highly portable (hand-held) Resolution: N/A Transmission speed: Slow (2G)[28] to moderate (3G)[29] | Expertise: Qualified Paramedic Training: required | EMS clinician | Stationary In transit | Not reported |
PreSSUB I Espinoza et al. 2016[30] | Bidirectional video and data transfer | Portable with use of a vehicle Resolution: High (HD-FHD) Transmission speed: Fast (4G) | Expertise: Other ambulance staff (EMS nurses) Training: required | Remote physician | In transit | Not reported |
PreSSUB II Espinoza et al. 2015[31] Brouns et al. 2016[32] | Bidirectional video and data transfer | Portable with use of a vehicle Resolution: (HD-FHD) Transmission speed: Moderate to fast (3G-4G) | Expertise: Other ambulance staff (EMS nurses) Training: required | Remote physician | In transit | Not reported |
InTouch Xpress Belt et al. 2016[33] | Bidirectional video | Highly portable (hand-held) Resolution: High (HD-FHD) Transmission speed: Fast (4G) | Expertise: Qualified Paramedic Training: required | Remote physician | In transit | ~$33,000 (~£27,000) in 2016: $23,000 (~£19,000) equipment + ~$10,000 (~£8,000) maintenance |
Smartphone with encrypted software Brotons et al. 2016[34] | Bidirectional video (unclear if data transfer) | Highly portable (hand-held) Resolution: Not reported Transmission speed: Not reported | Expertise: Qualified Paramedic Training: not reported | Remote physician | On scene Stationary | $2,250 (~£1,800) per unit in 2016 |
HipaaBridge Barrett et al. 2017[35] | Bidirectional video | Highly portable (hand-held) Resolution: High (HD-FHD) Transmission speed: Fast (4G) | Expertise: Qualified Paramedic Training: required | Remote physician | In transit | ~$600 (~£500) in 2017 |
iPad with video capability Shah et al. 2017[36] | Bidirectional video | Highly portable (hand-held) Resolution: Not reported Transmission speed: Not reported | Expertise: Qualified Paramedic Training: required | Remote physician | Not reported | Not reported |
Field-Telestroke Andrefsky et al. 2018[37] | Bidirectional video | Highly portable (hand-held) Resolution: Not reported Transmission speed: Not reported | Expertise: Qualified Paramedic Training: required | Remote physician | On scene In transit | Described as ‘Low cost’ |
REACHOUT project / HIPPA-compliant hand-held iPads Hackett et al. 2018[38] | Bidirectional video | Highly portable (hand-held) Resolution: Not reported Transmission speed: Not reported | Expertise: Qualified Paramedic Training: requirements were not reported | Remote physician | Not reported | Not reported |
Custom-built system Johansson et al. 2019[39] | Bidirectional video | Portable with use of a vehicle Resolution: High (HD-FHD) Transmission speed: Moderate to fast (3G-4G) | Expertise: Other ambulance staff (EMS nurses) Training: required | Remote physician | Not reported | Not reported |
Table 5
Telestroke Technology Study Outcomes
Study details | Purpose, Diagnostic accuracy, comparator and clinical scale | Time to conduct telestroke assessment | Acceptability: clinicians and/or patients | Impact on EMS clinician decisions or treatment | Impact on process (time metrics) outcomes | Impact on patient outcomes |
TeleBAT LaMonte et al. 2000[25] Xiao et al. 1997[26] | Purpose: Stratify stroke/facilitate care Diagnostic accuracy: Not reported Comparator: None. Assessed acceptability and usability of TeleBAT Clinical scale: NIHSS | Not reported | Paramedics x 2 and stroke specialists x 2[25]: Clinicians in favour of TeleBAT (privacy of video transmission, non-interference with regular tasks on ambulances; providing valuable information; & usability) Paramedics x 2 and stroke specialists x 2[26]: System did not intrude into paramedic/patient privacy and was safe. Adequate for clinical examinations: stroke specialists could score most NIHSS items, but difficulty with patients’ leg movement). Easy to learn/operate | Destination: Not reported Treatment: Not reported | Not reported | Not reported |
‘peeq-box’ Bergrath et al. 2012[27] |
Purpose: Stratify stroke/facilitate care Diagnostic accuracy: prehospital stroke diagnosis confirmed in hospital in 11 (61%, telestroke) vs 30 (67%, standard EMS) – difference non-significant Extrapolated data: Telestroke (false positives = 7; true positives = 11) vs standard EMS transport (false positives = 15; true positives = 30 (difference was non-significant) Non-significant differences between telestroke and standard EMS for other neurological/non-neurological diagnoses Comparator: Standard EMS transport (time metrics) and hospital-confirmed diagnosis Clinical scale: bespoke 14-item stroke history checklist + Glasgow Coma Scale | Not reported | In 15 of 18 missions the telemedicine system functioned faultlessly Significantly more (median 14) stroke-specific data points were transferred, in written form, from the EMS to the hospital via telestroke (versus median of 5 non-telestroke group). | Destination: Not reported Treatment: No significant impact on thrombolysis rates: 3/10 (30%) telestroke 5/27 (19%) standard EMS | Sample of patients with a suspected pre-hospital diagnosis of stroke Time on Scene: 4 min median increase with Telestroke (median 25 min) vs standard EMS (median 21 min). Difference was non-significant Scene to door time: 2.5 min median increase with Telestroke (median 37.5 min) vs standard EMS (median 35 min). Difference was non-significant Door-to-scan time: 2 min increase with Telestroke (median 59.5 min) vs standard EMS (median 57.5 min. Difference was non- significant | Not reported |
Stroke Angel Ziegler et al. 2008[28] | Purpose: Stratify stroke/facilitate care Diagnostic accuracy: Extrapolated data: Stroke vs non-stroke False positives: 27; False negatives: 53 True positives: 102; True negatives: 44 Sensitivity = 65.81%; Specificity = 61.97% Comparator: Hospital-based assessment using the same clinical scales and changes in time metrics before (prior to 2005) and after (2005–2007) introduction of Stroke Angel Clinical scales: Los Angeles Prehospital Stroke Screen, 3-item stroke scale | Not reported | Benefits stated by hospital clinicians were that EMS clinicians are "trained" by direct feedback from the PDA in dealing with the stroke patient. The use of Stroke Angel was evaluated to be consistently positive by EMS clinicians Hospital clinicians took the early warning seriously and were better prepared for the arrival of patients. Better communication between doctors and EMS clinicians, and improved perception of each other's tasks and work. | Destination: Not reported Treatment: Local lysis rate (number of lyses / all stroke patients enrolled on the stroke unit) increased from 6.1% (2005) to 11.2% (2007) | Call to scene time: unchanged (10 min before and after) Time on scene: before (17 min); after (2007) 23 min Travel time: before (26 min) after (2007) 22 min Call to-door time: before (53 min); after (2007) 55 min Door-to-CT time: before (53 min); after (2007) 30 min Patients treated with thrombolysis: Door to CT time: before (32 min); after (2007) 16 min Door-to-needle time: before 61 min, after (2007) 38 min | 1.5% of cases with symptomatic intracerebral bleeding (SITS-MOST criteria) |
Stroke Angel Rashid et al. 2015[29] | Purpose: Stratify stroke/facilitate care Diagnostic accuracy: Not reported Comparator: Standard EMS transport (time) Clinical scale: ‘Structured checklist’ | Not reported | Not reported | Destination: Not reported Treatment: Telestroke (39%), standard EMS (32%). 7% difference statistically significant | Data covered the period 2005–2013: Time on scene: 19 min (Telestroke), 20 min (Standard EMS). Not statistically significant Door-to-scan time: 12 min (Telestroke), 24 min (Standard EMS). Difference of 12 min was statistically significant | Not reported |
PreSSUB I Espinoza et al 2016[30] Certified by Autographe (Wavre, Belgium) | Purpose: Stratify stroke/facilitate care Diagnostic accuracy: Teleconsultants identified 12 patients (80%) with potential stroke or TIA, which concorded with in -hospital diagnosis in 10 patients (83%). Telestroke – no missed stroke diagnoses: Extrapolated data: False positives: 2; False negatives: 0; True positives: 10 Comparator: Hospital-based diagnosis Clinical scale: Unassisted Telestroke Scale | Median 9 minutes (IQR 8–13 min) | NIHSS was considered unsuitable for mobile telemedicine – this led to the development of a novel scale to rapidly assess stroke severity via telemedicine without assistance by a third party – the Unassisted Telestroke Scale 94% of teleconsultations were established successfully; one major technical issue occurred due to battery malfunction of the in-ambulance device | Destination: Not reported Treatment: Not reported | Not reported | Not reported |
PreSSUB II Espinoza et al. 2015[31] Brouns et al. 2016[32] Certified by Autographe | Purpose: Stratify stroke/facilitate care Diagnostic accuracy: Not reported Comparator: Standard EMS transport (time) and hospital diagnosis Clinical scale: Glasgow Coma Scale, Unassisted Telestroke Scale (UTSS) | Not reported | The proportion of successful in-ambulance telemedicine assessments was 96.2%[32] Technical and organisational feasibility was established[32] | Destination: Not reported Treatment: Thrombolysis rate (not yet available) | Call-to-CT time[32]: Standard EMS (87.1 min; 95% CI = 68.7-105.6) versus telestroke (50.8 min; 95% CI = 46.3–55.3): Statistically significant mean reduction of 36.4 minutes (95% CI = 17.5 to 55.3) | No telestroke-related adverse events. Mortality was similar in both groups[32] mRS, Barthel Index, EQ-5D and WHO-Five Well-being Index (not yet available) |
InTouch Xpress Belt et al. 2016[33] | Purpose: Stratify stroke/facilitate diagnosis: -Stroke -Ischemia Diagnostic accuracy: Extrapolated data: Stroke vs non-stroke (telestroke) False Positives: 3 True Positives: 12 Stroke vs non-stroke (Standard EMS transport) False Positives: 17 True Positives: 54 Comparator: Standard EMS transport (time) and hospital diagnosis. Clinical scale: Cincinnati Stroke Scale | With alteplase (n = 15): mean 7.3 min (95% CI = 4.9–9.8). Without alteplase (n = 74): mean 4.7 min (95% CI = 3.9–5.4) | Clinicians: 39% of teleconsults required reconnection. Connectivity was rapidly re-established in all but two cases; in all but these two cases, the tele-neurologist felt the clinical evaluation was satisfactory Acceptance among patients and EMS has been uniformly positive (but no data are presented to support this statement) | Destination: Not assessed Treatment: Not reported | Door to needle time: Telestroke - mean 28 min Standard EMS – mean 41 min (decrease of 13 minutes was statistically significant) Onset to scene time: Telestroke - mean 31.1 min Standard EMS – mean 50 min (18.9 min decrease was non-significant) Scene-to-door time: Telestroke - mean 29 min Standard EMS – mean 34 min (5 min decrease was non-significant) Onset to needle time: Telestroke - mean 92 min Standard EMS – mean 122 min (32 min decrease was significant) | Deaths: 0 (in both groups) Complications: 1 in telestroke group (vs 5 in standard EMS group) |
Smartphone with encrypted software Brotons et al. 2016[34] | Purpose: Stratify stroke/facilitate care Diagnostic accuracy: High correlations between telestroke NIHSS and NIHSS on hospital arrival Comparator: Telestroke NIHSS versus arrival at hospital NIHSS (conducted by the same physician) Clinical scales: CPSS, MEND exam | Not reported | Paramedics and physicians: easy to use and extremely valuable in making triage decision | Destination: Direct transfer to CSC Treatment: Not reported | Not reported | Not reported |
HipaaBridge on iPads Barrett et al. 2017[35] | Purpose: Stratify stroke/facilitate care Diagnostic accuracy: Not reported Comparator: None. Assessed acceptability and usability of HipaaBridge Clinical scales: NIHSS | Mean NIHSS assessment time 7.6 min (range 3 to 9.8 min) | Neurologists rated 83% of encounters as ‘satisfied/very satisfied’ EMS clinicians − 90% of encounters ‘satisfied/very satisfied’. | Destination: None Treatment: Not reported | Not reported | Not reported |
iPad with video capability Shah et al. 2017[36] | Purpose: Stratify stroke/facilitate care Diagnostic accuracy: Not reported Comparator: Standard EMS transport (time) Clinical scales: Cincinnati Stroke Scale and NIH-8 | Not reported | Not reported | Destination: Not reported Treatment: Not reported | Door to CT order: Mean decrease 6 min (95% CI = 3.6–8.5) Door to CT study start: Mean decrease 12 min (95% CI = 9.4–14.6) Door-to-CT result: Mean decrease 12.6 min (95% CI = 9.7–15.5) CT order to CT result: Mean decrease 6.9 min (95% CI = 4.5–9.3) | Not reported |
Field-Telestroke Andrefsky et al. 2018[37] | Purpose: Stratify stroke/facilitate care Diagnostic accuracy: Not reported Comparator: Standard EMS transport (time) Clinical scale: None reported | Not reported | Not reported | Destination: None Treatment: Non-significant increase in thrombolysis (10.6%-12.7%) | Door-to-scan time: Telestroke (10.7 min) Standard EMS (34.5 min) (improvement 23.8 min) Door-to-needle time: Telestroke (41 min) Standard EMS (50 min) (improvement 9 min) | Not reported |
REACHOUT Hackett et al. 2018[38] | Purpose: Stratify stroke/facilitate care Diagnostic accuracy: Not reported Comparator: Hospital telestroke (time) Clinical scale: None | Not reported | Not reported | Destination: Not reported Treatment: Not reported | Door-to-needle time: Significant median reduction of 26 minutes with EMS telestroke (median 39.5 min) compared with hospital based telestroke (median 65.5 min) | Not reported |
Custom-built system Johansson et al. 2019[39] CE Marked | Purpose: Stratify stroke/facilitate care Diagnostic accuracy: Not assessed Comparator: Acceptability / usability of the new telestroke system vs current practice Clinical scale: PreHAST and NIHSS | Not reported | 4 EMS nurses & 1 remote physician: 2 EMS nurses stated the system was reliable; 3 considered it to be safe Minor operating interference, physicians’ competence crucial and unclear efficacy emerged from analysis of free text Remote physician - image quality ‘more than satisfactory’ | Destination: Not assessed Treatment: Not assessed | 3 out 4 of EMS nurses did not believe that the system yielded a more uniform assessment or would reduce time-to-treatment | Not reported |
All telemedicine systems included video and audio components, with exception of Stroke Angel in which stroke screening information was collected and transferred from the ambulance to hospital. Earlier studies [25–29] utilised technology with lower resolution and slower transmission speeds than later studies [30–39] employing contemporary technology such as high definition, bi-directional video communication and 4G networks. Systems were either purpose-built or adapted from commercially available technology (e.g. tablet PCs). Most telemedicine systems were in the Beta stage of development, with exception of three Gamma stage systems [29, 32, 33].
The need for EMS clinician training on use of telestroke systems was reported for all but two studies [34, 38]. EMS clinicians were predominately paramedics, with three studies employing EMS nurses (equivalent to paramedics in these countries) [30, 32, 39] and one [27] EMS physicians and paramedics (with the aim of obviating the need for EMS physicians).
Costs were rarely reported, limiting comparison between studies. Where reported, [25, 26, 33–35] costs are based on year of publication prices (converted costs were calculated using historical exchange rates but not adjusted for inflation). None of the studies reported on the full range of costs required to implement telestroke (training, unit, operating and maintenance).
A variety of existing and commonly used pre-hospital and hospital-based stroke screening scales were used in conjunction with the telestroke systems (Table 5). Three studies evaluated and used a bespoke telemedicine scale. A 14-item stroke history checklist was developed by experts based on published checklists and recommendations and evaluated for use in conjunction with the ‘peeq-box’ system [27]. The PreSSUB I and II studies [30, 31] developed and evaluated the Unassisted Telestroke Scale; included items were based on existing stroke scales and evaluation of their appropriateness by experts [44, 45].
Data on diagnostic accuracy of telestroke systems were reported in five studies [27, 28, 30, 33, 34]. Pre-hospital stroke diagnosis (versus other neurological/non-neurological diagnoses) using the ‘peeq-box’ telestroke system was comparable to standard EMS transport and hospital confirmed diagnoses of stroke [27]. The Stroke Angel telestroke system, utilising the Los Angeles Pre-hospital Stroke Screen, had only moderate sensitivity (66%) and specificity (62%) for a diagnosis of stroke in the pre-hospital setting [28]. The PreSSUB I study [30] reported an equivalent rate of stroke diagnosis between telestroke and hospital-based clinical assessments (80% and 83% respectively). The InTouch Express telestroke system, using the Cincinnati Pre-hospital Stroke Scale, had equivalent rates of true/false positives for stroke diagnosis compared with standard EMS transport [33]. Finally, a smartphone telestroke system with encrypted software using the National Institute of Health Stroke Scale (NIHSS) reported ‘high’ intra-rater reliability with hospital-based NIHSS assessment [34].
Eleven of 15 studies [25–28, 30–35, 39] evaluated acceptability/usability of telestroke systems from the perspective of EMS clinicians and remote physicians using mixed methods. Results were positive, with studies reporting only minor issues related to connectivity [27, 33] and high levels of satisfaction with systems [25, 26, 28, 30, 32–35], image quality, reliability, usability or perceived safety [25–28, 32–34, 39]. One study reported only 25% of EMS nurses believed telestroke could improve assessments and reduce time-to-treatment due to concerns about clinician ability to use systems and integration into standard care processes [39]. Robust data on patient acceptance was not reported.
Time metrics were reported for 11 of 15 telestroke studies [27–33, 35–38]. Duration of telestroke consultation was reported in three [30, 33, 35]. PreSSUB I [30] consultations were 9 minutes (IQR 8–13 min). InTouch Xpress [33] consultations were 7.3 and 4.7 minutes (mean) for thrombolytic and non-thrombolytic patients respectively. Mean duration of NIHSS via the HipaaBridge system was 7.6 minutes [35].
With the Stroke Angel system, which allows transfer of relevant data to remote clinicians, travel time reduced by 4 minutes versus standard EMS transport [28]. Call-to-door time increased (2 minutes) and call-to-scene time matched standard care. The In-Touch Xpress study assessed onset-to-scene time [33] with a non-significant decrease of 18.9 minutes. Where evaluated, there were no significant differences in time-on-scene [27–29] and scene-to-door time [27, 33] between telestroke and standard EMS transport. PreSSUB II was the only study to assess Call-to-CT time [32], reporting a significant mean reduction of 36.4 minutes (95% CI = 17.5 to 55.3) with telestroke. Door-to-CT time was improved in four studies [27–29, 36] ranging from 12 minutes [29] to 24 minutes [28]. One study utilising IPads [36] reported significantly reduced door-to-CT start (12 minutes) and result (13 minutes). Four studies reported improved door-to-needle times [28, 33, 37, 38], two of which statistically significantly [33, 38], ranging from 13 minutes (InTouch Xpress versus standard EMS transport) [33] to 26 minutes (REACHOUT versus hospital-based telemedicine) [38]. InTouch Xpress telestroke significantly decreased onset-to-needle time (32 minutes) [33].
Excluding one study, where suspected stroke patients were taken directly to the nearest specialist centre [34], telestroke studies did not assess impact on EMS clinician decisions as a function of hospital destination (stroke-specialist centre versus non-specialist centre). Impact on IVT rates were assessed in four studies [27–29, 37]; two reported non-significant differences versus standard EMS transport (‘peeq-box’ [27] and Field-Telestroke [37]). Compared with standard EMS transport, the Stroke Angel system elicited significant increases in IVT, with 7% [29] and 5% [28] increases over 9 and 3-year periods respectively. However, neither study adjusted for concurrent increases in the thrombolysis rate. PreSSUB II [32] has yet to report on this.
Few data are available on patient safety outcomes. Stroke Angel [28] reported a 1.5% rate of symptomatic intracerebral bleeding and 11% mortality rate with use of the system. However, a-priori rates were not reported. PreSSUB II [32] reported no telestroke-related adverse events and equivalent mortality outcomes as with standard EMS transport. The InTouch Xpress [33] telestroke system also had equivalent mortality (zero), but a lower complication rate (1 vs 5 respectively), compared with the standard EMS transport group. None of the telestroke studies reported on patients’ functional health outcomes, although PreSSUB II [32] plans to.