Targeting Telestroke: Benchmarking Time Performance in Telestroke Consultations
Transcript of Targeting Telestroke: Benchmarking Time Performance in Telestroke Consultations
Targeting Telestroke: Benc
hmarking Time Performancein Telestroke ConsultationsJulian P. Yang, MD,* Tzu-ChingWu, MD,† Charles Tegeler, MD,‡ Ying Xian, MD, PhD,*
DaiWai M. Olson, PhD, RN, CCRN,* and Brad J. Kolls, MD, PhD, MMCi*
From the *Departmen
Center, Durham,NC; †De
Houston; and ‡Departm
Medicine, Winston-Salem
Received August 24,
accepted March 8, 2013.
Disclosures: J.P.Y. has
vant disclosures. C.T. has
disclosures. D.M.O. recei
470
Objective: To describe the length of time physicians spend completing telestroke
consultations and examine factors associatedwith that period.Methods: This is a ret-rospective review of data from telestroke software. Clinical data obtained between
July 2010 and February 2011 from 8 hub and 24 spoke hospitals were abstracted
for 235 consecutive consultations and linked to timemetadata generated by software
interaction. Consult length was defined as the time logged on to the robot and was
exclusive of any telephone interaction or documentation time. Response time was
defined as patient arrival to physician log-on. Results: Mean consult length for
203 complete, time-stamped cases was 14.5 minutes. There was no independent as-
sociation between consult length and age, diagnosis, time of arrival from symptom
onset, neurological examfindings, known recombinant tissue plasminogen activator
(r-tPA) contraindications, and absence of vascular risk factors. Mean consult length
was statistically longer in r-tPA–recommended cases (20.0 versus 15.3 minutes; P5.04). Mean response timewas 76.3 minutes.Conclusions: The relatively short consultlength suggests a workflow model in which acute stroke care is largely completed
before telestroke consultation with a specialist rendering an expert opinion on pre-
viously gathered data performed off-line. The findings for prolonged response times
indicate an area for improvement. Future workflow models for telestroke consulta-
tion will need to be reconsidered to optimize quality of care and clinical
efficiency. Key Words: Acute stroke—telemedicine—telestroke—care delivery
models.
� 2013 by National Stroke Association
Background and Significance
Despite being a leading cause of death and disability in
the United States, stroke continues to be an under-
recognized and undertreated diagnosis in the emergent
setting.1 Treatment rates of ischemic stroke with intrave-
nous recombinant tissue plasminogen activator (r-tPA)
continue to be low in the United States with only 3%-
8% of ischemic stroke patients receiving treatment.2 As re-
cently as 2009, nearly two thirds of US hospitals did not
t of Neurology, Duke University Medical
partment of Neurology, University of Texas,
ent of Neurology, Wake Forest School of
, NC.
2012; revision received March 8, 2013;
no relevant disclosures. T.-C.W. has no rele-
no relevant disclosures. Y.X. has no relevant
ves funding from the Duke Clinical Research
Journal of Stroke and Ce
offer r-tPA treatment at all with a trend for smaller hospi-
tal size and rural location being the associated factors for
nontreatment.3 It is estimated that only half the US popu-
lation resides within timely access (,60 minutes) to
a primary stroke center.4
In recent years, the emergence of telestroke—systems
of care employing high-quality, 2-way, audio–video
technology—has been seen as a potential solution to
bridge the gap between stroke centers and underserved
Institute, which serves as the statistical co-ordinating center for Get
With The Guidelines. B.J.K. has no relevant disclosures.
Grant support: None.
Address correspondence to Brad J. Kolls, MD, PhD,MMCi, Depart-
ment of Medicine–Neurology, Duke University Medical Center, Box
2900, Durham, NC 27710. E-mail: [email protected].
1052-3057/$ - see front matter
� 2013 by National Stroke Association
http://dx.doi.org/10.1016/j.jstrokecerebrovasdis.2013.03.010
rebrovascular Diseases, Vol. 22, No. 4 (May), 2013: pp 470-475
Table 1. Patient demographics
Variable
Age, mean 69.3 6 15.9 y
Gender*
Male 85 (41.9%)
Female 111 (54.7%)
Present, n (%) Absent, n (%)
Diabetes 58 (28.6) 145 (71.4)
Hypertension 74 (36.5) 129 (63.5)
Prior stroke or TIA 63 (31.0) 140 (68.9)
Atrial fibrillation or
flutter
22 (10.8) 181 (89.2)
Other vascular disease 27 (13.3) 176 (86.7)
Abbreviation: TIA, transient ischemic attack.
*Gender not recorded in 7 cases.
TELESTROKE PERFORMANCE 471
populations. Telestroke networks have been implemented
around the world, and the feasibility and the efficacy of
telestroke systems have been effectively established.5-7
With the ongoing adaptation of telestroke systems, there
is a growing consensus among stroke specialists
regarding minimum hardware requirements.8 However,
little attention has been paid to the data that is a product
of software used to manage workflow, clinical informa-
tion, and documentation during the telestroke encounter.
The electronic capture and storage of this data may
greatly aid in describing the metrics of telestroke care.
Electronic metadata, essentially data about data, can
emerge as a powerful tool to analyze the performance
and efficiency of clinical processes. For example, time
stamping of physician interaction with software can yield
new metrics regarding physician behavior.
Some data exist in the current literature to describe the
time commitment required by stroke specialists involved
with telestroke consultation. Previous reports have
outlined target evaluation times for telestroke consulta-
tion.9 Prior efficacy trials have also listed average consul-
tations times and treatment times.10 However, they have
been limited to the controlled settings of formal telestroke
trials and may not reflect the ‘‘real-world’’ application
of teletechnology. Ultimately, time metrics, such as con-
sultation length and response time, may have large
implications on best practice standards of treating ische-
mic stroke patients with r-tPA. Previous studies do not ex-
plore which clinical variables may affect these times. The
purpose of this study was to examine these relationships.
Methods
This is a retrospective analysis of demographic
and clinical data of telestroke encounters captured elec-
tronically in StrokeRESPOND documentation software
(InTouch Technologies, Inc., Santa Barbara, CA) corre-
lated with metadata generated from data logs recording
time points at which physicians interacted with hardware
end point devices (ie, telestroke robots).
From 8 hub and 24 spoke hospitals, there were 235 dis-
tinct, consecutive telestroke encounters between July 2010
and February 2011. The clinical information in these en-
counters was matched to data logs from both StrokeRES-
POND and end point devices. Time data in these logs
were automatically captured at the point of care during
telestroke encounters. Times of physician log-on and
log-off to the end point devices were recorded to the near-
est centisecond on a global network clock. Cases in which
patient arrival times were not charted were excluded (n5
11). Additionally, cases in which charted arrival times
conflicted logically with available metadata were ex-
cluded (eg, arrival times charted after telestroke consulta-
tion initiation) (n 5 21).
After exclusion, there were 203 telestroke encounters
with complete, time-stamped data logs encompassing
14 physician users. Demographic data electronically ab-
stracted from StrokeRESPOND included age, gender,
and the presence of diabetes, hypertension, prior stroke
or transient ischemic attack (TIA), atrial fibrillation or
flutter, and/or other vascular disease. Other information
collected about telestroke encounters included presence
of normal or baseline exam, final consultation diagnosis,
administration of r-tPA, and patient disposition. All per-
sonal health information was removed, and data were
stored in a Microsoft Excel file, which was shared directly
with investigators. Because the data set was completely
de-identified, this study received exempt status from
the Institutional Review Board at Duke University Medi-
cal Center (protocol ID: Pro00038896). Statistical analysis
relating the effect of clinical variables on mean times
was performed with analysis of variance and paired t
test using SAS-JMP and SASv9.4 software (SAS Institute,
Inc., Cary, NC).
‘‘Response time’’ was defined as the length of time be-
tween patient arrival at a spoke hospital and physician
user log-on to the end point. ‘‘Consult length’’ was de-
fined strictly as the time a physician user was directly
logged on. This definition was exclusive of any telephone
interaction or documentation time that occurred either
before or after the telestroke consultation.
Results
This sample had 85 males and 111 females; gender was
not recorded in 7 cases. Mean age was 69.3 years with
a standard deviation of 15.9 years (Inter Quartile Range
[IQR] 58-81). In terms of vascular risk factors, including
diabetes, hypertension, prior stroke or TIA, atrial fibrilla-
tion or flutter, and other vascular disease, 158 patients
(77.8%) had at least 1 risk factor present at admission,
and 92 patients (45.3%) were noted to have the presence
of 2 or more risk factors (see Table 1). Of note, 63 patients
(31.0%) were known to have a history of a stroke before
telestroke consultation.
Table 2. Mean response time by arrival times
Time of patient
arrival
Number of
cases
Mean response
time (minutes) P value
6:01 AM-12:00 PM 80 (39.4%) 74.6 .82*
12:01 PM-6:00 PM 84 (41.4%) 77.0
6:01 PM-12:00 AM 30 (14.8%) 74.2
12:01 AM-6:00 AM 9 (4.4%) 91.9
Total 203 76.3
*Omnibus test of means.
J.P. YANG ET AL.472
A final diagnosis at the end of the telestroke consulta-
tion was not recorded in 51 cases (25.1%). Categories of
consultation diagnoses consisted of ischemic stroke or
TIA (60/203, 29.6%), hemorrhagic stroke (6/203, 2.9%),
psychiatric illness (3/203, 1.5%), other medical illness
(6/203, 2.9%), and undiagnosed neurological symptoms
(77/203, 37.9%). In this sample, there were no docu-
mented diagnoses of seizure or other specified neurolog-
ical disease (eg, brain tumor). In the cases of ischemic
stroke or TIA, the administration of intravenous r-tPA
was recommended in 13 cases (13/60, 21.7%). Reasons
for not recommending r-tPA were inconsistently re-
corded, but in several cases, low National Institutes of
Health Stroke Scale and rapidly improving symptoms
were listed as reasons.
Mean response time was 76.3 minutes (IQR 39.4-94.0) in
this sample. Most telestroke consultations occurred dur-
ing daytime hours with 80 cases logged between 6:01
AM and 12:00 PM (39.4%) and 84 cases logged between
12:01 PM and 6:00 PM (41.4%) (see Table 2). In the 9 cases
in which patients arrived to emergency department in be-
tween 12:01 AM and 6:00 AM, mean response times were
Table 3. Mean consult leng
Variable
Age Age #55 (
15.4 m
Age $80 (
13.4 m
Arrival time #4.5 h (
15.0 m
Neurological exam Normal/baseli
15.6 m
tPA contraindications Absent (
14.7 m
Vascular risk factors Absent (
13.5 m
tPA recommended (all patients) Yes (
20.0 m
tPA recommended (ischemic stroke patients) Yes (
20.0 m
Abbreviation: tPA, tissue plasminogen activator.
noted to be longer (91.9 minutes). However, this differ-
ence was not statistically significant (P 5 .82).
Mean consult length for the entire cohort was 14.5 min-
utes (IQR 9.2-18.4). Mean consult length was significantly
longer in cases in which r-tPA was recommended (20.0
versus 15.3 minutes, P 5 .04). In the subset of patients
with a final diagnosis of TIA or ischemic stroke was
made, the association between mean consult length and
r-tPA recommendation was found to be stronger (20.0
versus 14.2 minutes, P 5 .02) (see Table 3). There was no
independent association between consult length andmul-
tiple clinical variables, including age, time of arrival from
symptom onset, absence of vascular risk factors, and con-
sultation diagnosis. Factors thought to favor shorter con-
sultation times, such as normal or baseline neurological
exam or presence of r-tPA contraindications, were also
not independently associated with consult length.
Recommendation for patient disposition (ie, whether to
transfer a patient for further care at a hub hospital or to
keep a patient at a spoke hospital for continued care)
was documented in only 67 cases. Diagnoses were noted
as follows: ischemic stroke or TIA (20/67, 29.9%), hemor-
rhagic stroke (3/67, 4.5), psychiatric illness (1/67, 1.5%),
other medical illness (1/67, 1.5%), and undiagnosed neu-
rological symptoms (25/67, 37.3%). Diagnosis was miss-
ing in 17 cases (25.4%).
In the subgroup of 67 cases with reported disposition,
transfer of care to hub hospital was recommended for
35 patients (52.2%). Transfer rates varied by diagnosis: is-
chemic stroke or TIA (15/20, 75%), hemorrhagic stroke
(3/3, 100%), psychiatric illness (0/1, 0%), other medical
illness (0/1, 0%), undiagnosed neurological symptoms
(11/25, 44%), and missing diagnosis (6/17, 35.3%). In
this group, all 4 cases in which r-tPA administration
th by clinical variables
P value
n 5 39) Age .55 (n 5 160)
in 14.3 min .43
n 5 66) Age ,80 (n 5 133)
in 15.1 min .13
n 5 79) .4.5 h (n 5 110)
in 14.1 min .44
ne (n 5 23) New symptoms (n 5 137)
in 14.8 min .65
n 5 61) Present (n 5 142)
in 14.4 min .77
n 5 45) Present (n 5 158)
in 14.8 min .33
n 5 13) No (n 5 115)
in 15.3 min .04
n 5 13) No (n 5 47)
in 14.2 min .02
TELESTROKE PERFORMANCE 473
was recorded and recommendedwere also recommended
to transfer care.
Discussion
Using temporal parameters obtained directly from tele-
stroke network data presents a unique opportunity to as-
sess telestroke quality of care. The American Stroke
Association’s ‘‘TARGET: Stroke’’ initiative with its goal
of door-to-needle times of 60 minutes or less may serve
as a frame of reference for judging the timeliness of tele-
stroke.11 The administration of intravenous r-tPAwas rec-
ommended in relatively few cases in this study, but the
times found for consult length and response time may in-
dicate how well telestroke might perform to help ensure
the quick delivery of r-tPA.
Consult Length
The mean consult length of 14.5 minutes suggests that
telestroke consultations are quick and efficient; however,
it is important to note that this figure solely represents
the time logged by a physician user directly interacting
with hardware end points and was exclusive of any un-
documented time spent communicating over the tele-
phone before or after telestroke consultation. Physician
time for documentation itself was also unaccounted for.
Consult length, found to be consistent across multiple
independent clinical variables, is relatively brief in rela-
tion to a 60-minute goal for completion of acute stroke
protocols. This finding has 2 important implications
about the use of telestroke in real-world settings. First,
the use of telestroke technology (ie, managing a patient
encounter via a robot interface) does not appear to be
overly cumbersome for physician users. Previous pub-
lished studies have documented software problems and
physician difficulty with technical troubleshooting that
limited the efficacy of telestroke consultations.12 The
group in this study, however, benefitted fromusing a stan-
dardized telestroke platform with software and hardware
end points integrated with a network managed by a third
party for connectivity and technical issues (SureCON-
NECT; InTouch Technologies, Inc.).
Second, the relatively brief consult lengths suggest
that the workflow model for managing acute stroke via
telestroke systems may be very different from stroke codes
conducted as live encounters in person. In our interpreta-
tion, physician users do not appear to be using telestroke
technology to manage stroke codes from start to finish in
entirety. Rather, after a patient’s arrival in the emergency
department of a spoke hospital, much of the initial triage
and workup, including neuroimaging, likely is conducted
off-line before telestroke consultation. On completion of
this workup, stroke specialists are then called in via tele-
stroke to review already collected clinical information,
perform a confirmatory neurological examination, and
then render an expert opinion regarding appropriate
management and disposition. The recommendation for ad-
ministration of r-tPA was the only statistically significant
variable associated with longer consult times (20.0 min-
utes), which possibly represent time needed to recheck
exam findings, personally review neuroimaging, or in-
struct spoke-site staff about drug dosing and delivery.
The relative lack of hemorrhagic stroke diagnoses (6/
203, 2.9%) recorded in this sample indirectly seems to
support this postulated workflow model for telestroke.
In live stroke codes, hemorrhagic strokes are encountered
far more frequently as both ischemic and hemorrhagic
stroke present similarly initially with focal neurological
deficits and are then distinguished by the presence of in-
tracranial blood on neuroimaging obtained during the
stroke code. Because of the relative ease in diagnosing
hemorrhagic stroke with positive neuroimaging findings,
these cases at the spoke hospitals are presumed to have
been triaged without any stroke specialist input and
that telestroke consultations were largely not initiated in
these cases. Consequently, hemorrhagic stroke appears
to be under-represented in this study’s sample.
Telestroke networks, at least in this sample, appear to
be used mainly as a way to connect to stroke specialists
for the clinical diagnosis of ischemic stroke in a confirma-
tory capacity and, if warranted, decision support regard-
ing thrombolytic therapy. This model is very dissimilar to
live-encounter stroke protocols found in many academic
centers in which stroke specialists manage the entire
stroke code from patient arrival to treatment. If our inter-
pretation of clinical workflow in this sample is correct, tel-
estroke networks may not be optimally providing
patients with early access to specialist expertise.
Indeed, a remarkable finding in this sample is that is-
chemic stroke or TIA was found to be the final diagnosis
in only 29.6% of cases. Teletechnology in its application
for telestroke networks may be underused as a means of
addressing the larger scope of neurological emergencies,
including hemorrhagic strokes, which are prone to the
same limitations in acute care because of a lack of timely
access to expertise. For example, quick evaluation by
a neurointensivist or neurosurgeon may change treat-
ment or triage plans for hemorrhagic strokes. The role
of telestroke beyond r-tPA treatment may need further
consideration as telestroke networks are developed and
implemented. Given the frequent activation of telestroke
encounters for nonstroke diagnoses, one may argue that
telestroke protocols in real-world execution are actually
functioning in a broader scope of practice as teleneurol-
ogy services. The results and outcomes of this practice,
however, do not appear to be followed and remain un-
known at this time. Teleneurology itself is a nascent prac-
tice, and the lack of publications indicates an area of
needed research and growth.13
In general, further insights garnered from research di-
rected toward elucidating workflow patterns in telestroke
consultations will become important on several levels: the
J.P. YANG ET AL.474
development of clinical protocols for best practice, the
design of software that best supports workflow efficiency,
and future research in the larger scope of acute stroke
care. Additionally, there are implications for telestroke re-
imbursement schema. Traditional, time-based billing
schedules may not ensure fair compensation for stroke
specialists managing acute and highly complex clinical
encounters.
Response Time
The mean response time of 76.3 minutes and its consid-
erable range (IQR 39.4-94.0) in this group clearly indicate
an area for improvement and a need for additional data.
Even in a model of care in which response time may corre-
spond to productive, off-line work being performed at the
spoke hospital prior to telestroke consultations, the pro-
tracted figures in this study do not support achievement
of adequate door-to-needle times less than 60 minutes. In
light of the relative brevity of consult length, the response
times in this data set suggest that quality improvement
may need to shift focus to an earlier point in the clinical en-
counter. Initial evaluation and triage of stroke patients
should be directed toward rapid acquisition of pertinent
history and quick identification of those patients requiring
specialist consultation. The efficiency of the process by
which the request for consultation is made also deserves
consideration (eg, transfer center versus direct page).
Last, the factors that may delay consultants from promptly
answering requests need to be identified.
Time of day is one important variable that may affect all
these aspects of early acute stroke care. Recent attention to
‘‘off-hour admissions’’ has shown outcome differences for
stroke patients.14,15 The small sample in this study did not
show a statistically significant trend; however, the 9 cases
arriving in early morning hours (12:01 AM to 6:00 AM) did
have longer mean response times. The reasons for longer
response times are not documented, but future studies
could be directed toward collecting information about
physician user practices in regards to ‘‘off-hour’’ call
structure.
Telestroke will benefit from the development of
standardized clinical protocols that are optimized for effi-
ciency. Beyond the installation of high-quality technolog-
ical platforms, a recurring theme of successful telestroke
systems is the necessity of education and continual re-
education of personnel at both hub and spoke sites.
Clinical protocols would require all personnel involved
to gain familiarity with the aims of telestroke consulta-
tions and the best practice standards of acute stroke
care. Despite the availability of clinical expertise via tech-
nology, quality care is still dependent on the ready and
able local execution of recommendations made remotely.
Limitations and Future Directions
The findings of this study should be interpreted with
caution given several limitations. Most notably, no datum
or information was available regarding the occurrence or
timing of any telephone interactions and how initial con-
tact from spoke hospital to hub hospital was made. Con-
sequently, potentially important factors, such as the
amounts of time elapsed between patient arrival, decision
to request consultation, first contact with consulting
physicians, and then actual telestroke initiation, were
not considered. Also, the time required for off-line docu-
mentation or follow-up was not recorded.
In general, the retrospective design and small sample
size may have introduced information bias and selection
bias. Individual cases in this sample were not indepen-
dent in that repeated observations were taken from only
14 physician users at 8 hub hospitals. A clear example
of selection bias in this sample is the likely under-
identification of hemorrhagic strokes, which can be ex-
plained by a lack of uniform clinical criteria for telestroke
initiation. Further selection bias may have occurred from
data loss by excluding 32 cases from the initial sample
with conflicting or illogical charted arrival times.
The source and method of data acquisition for different
variables figure largely into the limitations of this study.
Outcomes, such as consult length, were calculated strictly
with time points obtained from automatically generated
electronic metadata and reflect a certain measure of data
integrity. Conversely, response time was calculated using
patient arrival times that were charted into electronic doc-
umentation. Because these arrival times essentially repre-
sent retrospective chart review, they are prone to human
error in both initial documentation and retrospective
data abstraction. In this way, the presence of conflicting
or illogical times is explained. Similarly, there is an incon-
sistent record of basic demographic information with 7
cases missing any notation of patient gender.
As an initial foray into analyzing the metrics of tele-
stroke efficacy from software audit trails, the results of
this study can serve as a point of comparison for future
studies. A very important point to distinguish in the de-
sign of telestroke studies is to standardize clinical proto-
cols so that the technology is used in the same way,
from user to user and from site to site. The conjectured
model of care in this study with physician users logging
in only at the end of the clinical pathway for acute stroke
management, if true, should be formally outlined into
prospective research design. Future studies should also
require clear documentation about key points in the off-
line workflow, including any telephone interactions be-
tween hub and spoke hospitals, and also all traditional
acute stroke protocol metrics at the spoke hospital. Other-
wise, a different model of care may be more appropriate
for research.
This study should draw attention to the importance of
telestroke software design to take advantage of a unique
opportunity to study acute stroke care and physician de-
cision making. The elimination of secondhand data ab-
straction with electronic data acquisition can minimize
TELESTROKE PERFORMANCE 475
data corruption. More importantly, the ability to capture
metadata can yield powerful information: by analyzing
the sequence and patterns of clinical information entry
and utilization, the actual thought process of a physician
user can be investigated. Physician decision-making pat-
terns and other telestroke metrics, such as consult length
and response times in this study, would ideally be tied di-
rectly to patient outcome measures.
Ultimately, the information garnered in future
outcome-focused research could aid in the optimization
of not only telestroke systems but also acute stroke care
protocols as a whole.
Acknowledgments: We gratefully acknowledge the ef-
forts of Andre Grujovski and TimWright in helping to gather
and prepare the initial data set. Additionally, we would
thank Kristina Riemen for her assistance in manuscript prep-
aration.
References
1. Roger VL, Go AS, Lloyd-Jones DM, et al, on behalf of theAmerican Heart Association Statistics Committee andStroke Statistics Subcommittee. Heart disease and strokestatistics—2011 update: a report from the American HeartAssociation. Circulation 2011;123:e18-e209.
2. Reeves MJ, Arora S, Broderick JP, et al, Paul CoverdellPrototype Registries Writing Group. Acute stroke carein the US: results from 4 pilot prototypes of the Paul Cov-erdell National Acute Stroke Registry. Stroke 2005;36:1232-1240.
3. Laino C. Most hospitals don’t offer tPA to ischemic strokepatients. Neurol Today 2009;9:10.
4. Albright KC, Branas CC, Meyer BC, et al. ACCESS: acutecerebrovascular care in emergency stroke systems. ArchNeurol 2010;67:1210-1218.
5. Audebert HJ, Kukla C, Clarmann von Claranau S, et al,TEMPiS Group. Telemedicine for safe and extended useof thrombolysis in stroke: the Telemedic Pilot Project for
Integrative Stroke Care (TEMPiS) in Bavaria. Stroke2005;36:287-291.
6. Meyer BC, Raman R, Hemmen T, et al. Efficacy of site-independent telemedicine in the STRokE DOC trial:a randomised, blinded, prospective study. Lancet Neurol2008;7:787-795.
7. Sairanen T, Soinila S, Nikkanen M, et al. Two years ofFinnish telestroke: thrombolysis at spokes equal to thatat hub. Neurology 2011;76:1145-1152.
8. Schwamm LH, Holloway RG, Amarenco P, et al, Ameri-can Heart Association Stroke Council and the Interdisci-plinary Council on Peripheral Vascular Disease. Areview of the evidence for the use of telemedicine withinstroke systems of care: a scientific statement from theAmerican Heart Association/American Stroke Associa-tion. Stroke 2009;40:2616-2634.
9. Demaerschalk BM, Miley ML, Kiernan TE, et al. Stroketelemedicine. Mayo Clin Proc 2009;84:53-64.
10. Demaerschalk BM, Raman R, Ernstrom K, et al. Efficacyof telemedicine for stroke: pooled analysis of the StrokeTeam Remote Evaluation Using a Digital ObservationCamera (STRokE DOC) and STRokE DOC Arizona tele-stroke trials. Telemed J E Health 2012;18:230-237.
11. Fonarow GC, Smith EE, Saver JL, et al. Improving door-to-needle times in acute ischemic stroke: the design andrationale for the American Heart Association/AmericanStroke Association’s Target: stroke initiative. Stroke2011;42:2983-2989.
12. Demaerschalk BM, Bobrow BJ, Raman R, et al. Stroketeam remote evaluation using a digital observation cam-era in Arizona: the initial mayo clinic experience trial.Stroke 2010;41:1251-1258.
13. Rubin MN, Wellik KE, Channer DD, et al. Systematic re-view of teleneurology: methodology. Front Neur 2012;3:156. http://dx.doi.org/10.3389/fneur.2012.00156.
14. ReevesMJ, Smith E, FonarowG, et al. Off-hour admissionand in-hospital stroke case fatality in the get with theguidelines-stroke program. Stroke 2009;40:569-576.
15. Streifler JY, Benderly M, Molshatski N, et al. Off-hoursadmission for acute stroke is not associated with worseoutcomes—a nationwide Israeli stroke project. Eur J Neu-rol 2012;19:643-647.