Introduction
A stroke unit is a form of organized inpatient care, where acute stroke patients are managed by a specialized interdisciplinary team of physicians, nurses, and allied healthcare professionals with expertise in stroke in a geographically defined location. Reference Boulanger, Lindsay and Gubitz1 Stroke unit care is associated with lower mortality, better functional status, and increased probability of living at home 1 year following stroke. Reference Langhorne and Ramachandra2 Location of regional stroke activity in a central unit has the benefit of concentrating expertise and services but may introduce delay between first contact with the emergency department (ED) and stroke unit admission particularly if there are multiple pathways for admission. Reference Darehed, Norrving, Stegmayr, Zingmark and Blom3 Although early stroke unit admission may be associated with lower mortality, Reference Turner, Barber and Dodds4 it is yet unclear whether longer delay affects patients’ functional outcomes and hospital resources. Reference Boulanger, Lindsay and Gubitz1 The objective of this study was to determine whether shorter time to transfer of acute ischemic and hemorrhagic stroke patients from the ED to the stroke unit was associated with better patient and system-level outcomes.
Methods
This was a single-center retrospective cohort study. The stroke unit in this study is located at a comprehensive stroke center (CSC) affiliated to a comprehensive stroke network, located in Hamilton, Ontario, Canada. It is comprised of a ward-level care unit and a stepdown-level care unit. Only the CSC offers acute reperfusion therapies including thrombolysis and endovascular therapy (EVT). Hyperacute stroke patients are either sent directly to the CSC or urgently transferred to the latter from the ED of one of two primary stroke centers (PSC) involved in this study. Following hyperacute stroke care (neuroimaging with or without reperfusion therapy) in the ED, they are admitted to the stroke unit with the highest priority. Acute stroke patients who present themselves to an ED at the CSC outside of hyperacute treatment window (more than 6 h from stroke onset) are also directly admitted to the stroke unit. If no bed is available on the stroke unit, stroke patients are temporarily admitted to an alternative inpatient unit under internal medicine. Out-of-window stroke patients who present themselves to an ED at a PSC are either transferred directly to the stroke unit or temporarily admitted under internal medicine at the PSC. Patients eligible to be admitted to the stroke unit include all patients with acute ischemic and hemorrhagic stroke, as well as selected cases of transient ischemic attacks, regardless of age, comorbidities, or eligibility for rehabilitation. Patients are not admitted to the stroke unit if they require intensive care unit, if they are diagnosed with aneurysmal subarachnoid hemorrhage, or if their initial goals of care are not compatible with active treatment (e.g., palliative goals of care).
Approval from our local research ethics board (Hamilton Integrated Research Ethics Boards) was obtained prior to initiating this study, and the board waived the need for patient consent. We were fully compliant with the Declaration of Helsinki.
Study Population
Subjects in this study included all consecutive patients with confirmed acute ischemic or hemorrhagic stroke diagnosed at the three hospital sites, who were subsequently transferred to the stroke unit, and whose date of stroke unit admission occurred between January 1, 2016 and December 31, 2018. Subjects were excluded if they were admitted to the intensive care unit prior to the stroke unit, if their primary reason for admission was not acute stroke (e.g., stroke occurring during hospital stay), or if they were diagnosed with aneurysmal subarachnoid hemorrhage, transient ischemic attack, or stroke mimic.
Data Collection
The following data were collected for all subjects: demographics, stroke onset, initial stroke severity as measured by the National Institute of Health Stroke Scale (NIHSS), ED arrival time (weekday daytime [8 AM–5 PM] vs weekend or nighttime, ED length of stay), initial brain imaging time, acute stroke treatment (administration of tissue plasminogen activator [tPA], onset-to-needle time, door-to-needle time, and EVT), and delay between ED arrival and arrival to the stroke unit.
The primary outcomes were delays (intervals) between ED arrival and stroke investigations (vascular imaging, echocardiography, and Holter monitoring), discharge location from the stroke unit, total length of stay in hospital (from ED arrival to discharge from stroke unit), modified Rankin scale (mRS) at 3 months poststroke, Reference Bruno, Shah and Lin5 death at 3 months poststroke, and nondisabled functional status at 3 months poststroke defined by mRS of 0–2. The mRS was assessed by the stroke neurologist either via in-person visit or via telephone if in-person visit was not possible.
Statistical Methods
Multivariable logistic regression analysis was used to determine whether delay to stroke unit admission was associated with each of the primary outcomes, adjusted for the following variables: age, sex, comorbidities (coronary artery disease, hypertension, diabetes mellitus, dyslipidemia, atrial fibrillation or atrial flutter, and chronic kidney disease), stroke type, initial NIHSS, administration of tPA, and/or EVT. Of note, death at 3 months poststroke was considered as an outcome only in a sensitivity analysis. Subgroup analysis was stratified according to age (<65 years, 65–75 years and >75 years), sex, stroke type (ischemic, hemorrhagic), and NIHSS (mild 0–5, moderate 6–11, severe 12, and above). Subgroup analysis stratified according to the source of patient admission (admitted directly from CSC, transferred from PSCs) was also performed as a sensitivity analysis.
Multivariable analysis evaluating the correlation between potential determinants of delay (age, sex, comorbidities, stroke type, initial NIHSS, hospital of origin, admission from an alternative service, ED arrival during weekday daytime, stroke onset within 6 h of ED arrival, delay to brain imaging, administration of tPA, and EVT) and delay to stroke unit admission was also performed. Of note, stroke onset within 6 h of ED arrival and delay to brain imaging were included as covariates in the sensitivity analysis.
All comparisons were carried out using a two-sided test at a significance level of 0.05. Missing values were accounted for using simple imputation by mean for continuous characteristics and imputation by mode for categorical variables.
Results
This study included 1435 patients, 663 (46.2%) female, with a mean (SD) age of 72.9 (13.5) years. Ischemic stroke patients comprised of 92.4% of the cohort (N = 1326), while the remaining 109 (7.6%) patients had hemorrhagic stroke. The median initial NIHSS was 5 (interquartile range [IQR] 2–10). tPA was administered in 299 patients (20.8%), EVT was performed on 93 patients (6.5%), and 62 patients (4.3%) received both tPA and EVT. The median delay to stroke unit admission was 7.7 h (IQR 3.8–15.1), and the range of delay was between 4 and 31,745 min (22.0 days). Two hundred and eight (14.5%) patients were admitted to the stroke unit with delays of at least 1 day. See Tables 1 and 2. Most patients were admitted directly from ED at the CSC to the stroke unit (N = 1208, 84.2%). Patients’ locations of origin are illustrated in Figure 1.
Figure 1: Patient journey prior to stroke unit admission.
Table 1: Baseline characteristics (N = 1435 unless otherwise specified)
CSC = comprehensive stroke center; NIHSS = National Institute of Health Stroke Scale; PSC = primary stroke center; SD = standard deviation.
* Missing data: initial NIHSS = 7.
Table 2: Treatment characteristics and delays (N = 1435 unless otherwise specified)
IQR = interquartile range.
* Missing data: emergency department arrival during weekday daytime = 4, delay between emergency department arrival and initial brain imaging = 3, onset-to-needle time = 2, door-to-needle time = 2, emergency department length of stay = 2, delay between arrival to and departure from an alternative service = 7, delay between emergency department arrival and arrival to the stroke unit = 12.
† N = 299.
‡ N = 161.
The median delays to stroke investigations were vascular imaging, 11.8 h (IQR 0.2–41.8), transthoracic echocardiogram, 65.5 h (IQR 43.4–98.9), and Holter monitoring, 95.0 h (IQR 60.5–168.2). The median length of hospitalization was 10.0 days (IQR 4.0–26.9), and 545 (38.0%) patients were discharged home directly from the stroke unit. The median mRS at 3 months poststroke was 2 (IQR 1–4), 175 (12.2%) patients were deceased at 3 months, and 801 (55.8%) patients were nondisabled at 3 months with mRS 0–2. Full details related to patient and system outcomes are presented in Table 3.
Table 3: Primary study outcomes (N = 1435 unless otherwise specified)
IQR = interquartile range; mRS = modified Rankin Scale.
* Missing data: vascular imaging = 3, transthoracic echocardiogram = 3, Holter monitoring = 3, total length of stay in hospital = 62, discharge location from the stroke unit = 60, mRS at 3 months poststroke = 2, nondisabled functional status at 3 months poststroke = 2.
† N = 1231.
‡ N = 1153.
§ N = 829.
After adjusting for covariates, the effects of delay to stroke unit admission on primary outcomes are in Table 4. For each additional day of delay, we noted 2.0 days of increase in total length of stay in hospital (95% confidence interval [CI] 0.8–3.2, p = 0.001), 11.5 h of additional delay to initial vascular imaging (95% CI 9.6–13.4, p < 0.0001), 24.2 h of additional delay to initial Holter monitoring (95% CI 7.9–40.6, p = 0.004), and 1.02 time greater odds of discharge to home from the stroke unit (95% CI 1.01–1.04, p = 0.01). Each additional day of delay in stroke unit admission was also associated with a reduction in the odds of nondisabled functional status at 3 months (odds ratio 0.98, 95% CI 0.96–1.00, p = 0.01).
Table 4: Effect of 1 day of additional delay to stroke unit admission on primary outcomes (N = 1435)
mRS = modified Rankin Scale.
The most significant factor affecting delay was stroke onset within 6 h of ED arrival (605.9 min decrease in delay, 95% CI 407.9–803.9, p < 0.0001), delay to brain imaging (59.4 min increase in delay for each additional hour, 95% CI 48.0–71.4, p < 0.0001), admission from an alternative service (3918.7 min increase in delay, 95% CI 3621.2–4079.9, p < 0.0001), transfer from a PSC (740.2 min increase in delay, 95% CI 456.2–1019.9, p < 0.0001), and older age (6.2 min increase in delay for each additional year, 95% CI 0.1–12.4, p = 0.046). See Table 5.
Table 5: Effect of potential predictors on delay to stroke unit admission (N = 1435)
N/A = not applicable; NIHSS = National Institute of Health Stroke Scale.
* Positive value indicates increased delay. Negative value indicates decreased delay.
† In sensitivity analysis, two additional predictors were added: (1) stroke onset within 6 h of emergency department arrival and (2) 1 h of additional delay to brain imaging.
Discussion
We found that delay to stroke unit admission was associated with longer hospital stay and longer delays to stroke investigations. It was also associated with poorer functional outcomes in stroke patients, especially among those aged 75 years and above (see Table 6). Arrival to ED within 6 h of stroke onset and shorter delays to brain imaging predicted shorter delays to stroke unit admission, while transfer from a PSC or from an alternative service, as well as older patient age, predicted longer delays.
Table 6: Subgroup analysis of the effect of 1 day of additional delay to stroke unit admission on nondisabled functional status at 3 months poststroke (modified Rankin Scale of 0–2)
Compared to the Ontario Stroke Registry, Reference Kapral, Fang and Alibhai6 our cohort showed similar age, gender distribution, and stroke severity. However, our study population reported higher proportion of ischemic stroke (92.4% vs 86.0%), higher prevalence of cardiovascular risk factors including hypertension (76.4% vs 67.0%), diabetes (33.6% vs 24.4%), dyslipidemia (61.3% vs 36.7%), atrial fibrillation (22.9% vs 15.8%), and renal disease (8.8% vs 7.9%), as well as higher rate of thrombolysis among ischemic stroke patients (22.5% vs 14.7%).
Longer delays to stroke unit admission were associated with increased length of hospital stay and longer delays to stroke investigations, especially vascular imaging and Holter monitoring. There were two additional days of hospitalization for every one additional day of delay to stroke unit admission. Shorter length of stay may reduce stroke-related healthcare costs, considering that a hospitalized stroke patient cost $4700 USD per day in the United States in 2007. Reference Feigin, Norrving and Mensah7–Reference Yacoub, Al-Qudah, Khan, Farhad, Ji and Souayah9 Although shorter delays to stroke unit admission may be associated with shorter hospitalization, it is worth noting that efforts specifically targeted at reducing delays to investigations may also lead to shorter length of stay. Reference Khalifa10
Longer delays to stroke unit admission were associated with lower likelihood of nondisabled functional outcome as defined by mRS 0–2, and older patients aged 75 years and over may have worse outcomes with longer delays in the subgroup analysis. Although stroke units have been found to reduce dependency or death (mRS 3–6) in a meta-analysis and earlier admission may be associated with lower mortality, it was unclear whether stroke unit admissions were time-sensitive in regard to functional outcome, Reference Turner, Barber and Dodds4,11 a knowledge gap which we aimed to fill. Shorter delays to stroke unit admission were, however, not associated with a leftward shift in the mRS distribution or lower mortality. Reducing delays to stroke unit admission may not result in benefits as clinically significant as reducing delays to tPA administration or EVT, given that most clinically meaningful therapeutic strategies occur in the hyperacute stage. Reference Moussaddy, Demchuk and Hill12 Interestingly, longer delays to stroke unit admission were associated with increased likelihood of discharge to home. This would contradict previous studies, Reference Langhorne and Ramachandra2 and likely represented reverse causation, as patients who were more likely to be discharged home may have received less priority to stroke unit admission as the initial event was mild.
Arrival to ED within 6 h of stroke onset and shorter delays to brain imaging predicted shorter delays to stroke unit admission. Given the high risk of clinical deterioration and recurrent stroke in the hyperacute stage, in addition to possible complications following reperfusion therapies, patients assessed in the context of a “code stroke” often receive priority admission when bed resources are limited on the stroke unit, compared to patients who present to ED in the subacute stage. Reference Amatangelo and Thomas13 Furthermore, given that radiological evidence of an acute stroke is often required for stroke unit admission and that earlier brain imaging may be a marker for higher stroke acuity, shorter delays to brain imaging may also lead to faster stroke unit admission. Transfer of patients from PSCs and alternative service predicted longer delays to stroke unit admission. However, ED arrival on weekends and during nighttime did not predict delay to stroke unit admission in this study, in contrast to previous studies showing longer delays to admission during these times. Reference Kim14,Reference Mueller, Fiskio and Schnipper15 This may reflect specific aspects of local practices not necessarily generalizable to other institutions. Lastly, older patient age predicted longer delays to stroke unit admission, in keeping with previous literature reporting longer delays in accessing stroke care among older patients. Reference Birnbaum, Rodriguez and Topel16 This is relevant especially given that older patients may be at increased risk of delirium related to prolonged ED stay and that these patients may benefit more from faster stroke unit admission according to our subgroup analysis. Reference Emond, Grenier and Morin17 Time spent on consultation with other services regarding most appropriate admission service may also lead to longer delays to stroke unit admission.
This study had several limitations. This was a single-center observational study, which reported a patient population that may differ from the regional stroke patient profile. When compared to the Ontario Stroke Registry, Reference Kapral, Fang and Alibhai6 patients in our study showed a higher proportion of ischemic stroke, a more severe cardiovascular risk profile, and were more likely to receive thrombolysis. In addition, the finding of poorer functional outcomes with longer delays should be interpreted with caution as the upper limit of the 95% CI overlapped with the null hypothesis despite p < 0.05. Moreover, patients transferred from PSCs were associated with longer delays to stroke unit admission, in contrast to patients admitted directly to the CSC. These two groups of patients may have different stroke characteristics. However, if patients transferred from PSCs were excluded, the effects of delays on functional outcomes would become statistically insignificant due to the narrower range of delays among patients admitted directly from the CSC. Our results may not be generalizable to systems where the workflow does not include admission to PSCs prior to transfer to a regional stroke unit. Furthermore, this study excluded any patients who required intensive care prior to stroke unit admission, among which patients with hemorrhagic stroke were likely more prevalent than those with ischemic stroke. Reference Wang, Hsieh, Huang, Chen, Hsieh and Chao18 Lastly, 10 other hospital sites not mentioned in this study transferred patients to the CSC only for the purpose of EVT. These patients were excluded as they returned to their hospitals of origin shortly after EVT. Results from this study may not be generalizable to other comprehensive stroke networks and need to be replicated. Further studies using multicentered data registries are needed to clarify whether shorter delays to stroke unit admission may lead to better outcomes. While we have accounted for relevant factors in the analysis, reverse causation cannot be completely eliminated in observational data.
This study showed better functional outcomes among patients with a shorter delay to stroke unit admission. Moreover, shorter delays were associated with better system-level outcomes including shorter length of stay and shorter delays to stroke investigations, which may reduce healthcare costs. If future additional studies successfully replicate these key benefits and provide further evidence to the time-sensitive nature of stroke unit admission, a strong argument could be made in favor of optimizing healthcare resources in order to reduce delays to stroke unit admission. Beyond allocating more resources into funding additional beds and personnel on the stroke unit, the creation of a “stroke navigator” role may to lead to faster stroke unit admission by optimizing patient flow and facilitating discharges from the stroke unit. Reference Fulbrook, Jessup and Kinnear19
Acknowledgements
The authors would like to thank Susie Bojko for her assistance with data collection. The authors also wish to thank Danielle Jerome, Linda Gould, and Stefan Pagliuso for their assistance with study design.
Funding
This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.
Conflict of Interests
The authors declare that there is no conflict of interest.
Statement of Authorship
B.Y.C.: study design, data collection, and manuscript preparation; H.G.: data collection and manuscript revision; C.Y.: data analysis; H.E.S.: data collection: C.L.: data collection; A.B.: data collection; J.J.: data collection; Y.S.: data collection; V.S.: data collection; S.C.: data collection; I.Z.: data collection; Y.F.M.: data collection; S.M.: data collection; M.S.: study design, manuscript revision, and study supervision.
Data Availability
Requests for data from this study will be considered 1 year after the date of publication.
Open access
