Article Text
PDF
Abstract
Objective To quantify the time between initial image acquisition (CT angiography (CTA)) and notification of the neuroendovascular surgery (NES) team, a potentially high yield time window to target for optimization of endovascular thrombectomy (ET) treatment times.
Methods We reviewed our multihospital database for all patients with a stroke with emergent large vessel occlusion treated with ET between January 1, 2017 and August 5, 2020. We dichotomized patients into rapid (≤20 min) and delayed (>20 min) notification times and analyzed treatment characteristics and outcomes.
Results Of 367 patients with ELVO undergoing ET for whom notification data were available, the median time from CTA to NES team notification was 24 min (IQR 12–47). The median total treatment time was 180 min (IQR 129–252). The median times from CTA to NES team notification for rapid (n=163) and delayed (n=204) cohorts were 11 (IQR 6–15) and 43 (IQR 30–80) min, respectively (p<0.001). The median overall times to reperfusion were 134 min (IQR 103–179) and 213 min (IQR 172–291), respectively (p<0.001). The delayed patients had a significantly lower National Institutes of Health Stroke Scale (NIHSS) score on presentation (15 (IQR 9–20) vs 16 (IQR 11–22), p=0.03), were younger (70 (IQR 60–79) vs 77 (IQR 64–85), p<0.001), and more often presented with posterior circulation occlusion (16.7% vs 7.4%, p<0.01). The group with rapid notification time had a statistically larger median improvement in NIHSS score from admission to discharge (6 (IQR 0.5–14) vs 5 (IQR 0.5–10), p=0.04).
Conclusions Time delays from initial CTA acquisition to NES team notification can prevent expedient treatment with ET. Process improvements and automated stroke detection on imaging with automated notification of the NES team may ultimately improve time to reperfusion.
- stroke
- CT angiography
- intervention
- standards
Data availability statement
Data are available from the corresponding author upon reasonable request.
Statistics from Altmetric.com
Introduction
Endovascular thrombectomy (ET) has become the accepted standard of care for patients with acute ischemic strokes caused by emergent large vessel occlusions (ELVOs).1 The clinical efficacy of ET relies on expedient treatment times.2 3 In the SWIFT PRIME trial, the probability of functional independence decreased by 10% over the first hour of treatment delay and by 20% with every subsequent hour.4 The SWIFT and STAR studies assessing ET showed that every 15 min reduction in time from onset to recanalization was associated with significantly improved 90-day clinical outcomes for 34 of every 1000 patients.5 Furthermore, expedient ET in appropriate patients with ELVO improves the cost effectiveness of the treatment.6 7
Given the importance of timeliness in ET, many clinical trials have found benefit is achieved when specific time windows in the stroke workflow are targeted for intervention.3 8 Targeting stroke diagnosis time led to the emergence of mobile stroke units, whereas triaging patients with probable ELVO to centers with endovascular capabilities has led to decreased transfer times. Based on analysis of time epochs related to ET treatment in our high-volume stroke care institution, we have observed that the time between initial image acquisition to neuroendovascular team notification is a potentially high yield time window to target for optimization of ELVO treatment times. In this study, we quantify this important time window and hypothesize that improving neuroendovascular surgery (NES) consultation times might lead to future improvements in stroke care. We hypothesize that patients with faster imaging to NES team notifications would have faster times to treatment with ET and quantify the relationship.
Methods
Prospective data collection
Retrospective analysis of a prospective quality improvement database was approved by our institutional review board. This prospectively maintained, multihospital (six primary stroke centers, four thrombectomy-capable stroke centers, one comprehensive stroke center included), single health system database was reviewed for all patients with stroke with ELVO treated with ET during the 43-month period between January 1, 2017 and August 5, 2020. Database variables of interest included workflow times (including time of last known well, presentation time, time of CT angiogram acquisition (CTA), NES consultation time, vascular access time), general patient demographics and presenting features (baseline modified Rankin Scale (mRS) score, occlusion location, presenting National Institutes of Health Stroke Scale (NIHSS) score), and outcome variables (including change in NIHSS score from presentation to discharge and 90-day mRS score). Patients were classified by mode of hospital arrival (arrival by emergency medical service (EMS) directly at an ET center, EMS then transfer to the ET center, transfer from an inpatient status, and walk-in). For patients who were transferred, time to NES team notification was calculated using the initial imaging at the initial hospital. In our institution, all NES consultations are determined for appropriateness by the stroke neurology team at each site. The decision when to consult the NES team (before hospital arrival, before imaging, or after imaging) is at the discretion of the stroke neurology team. Transfers are initiated if the patient meets the criteria for severity and is at a hospital location that is not thrombectomy-capable, within or outside of our health system. Team notification times were documented through time stamps on our institutional healthcare messaging mobile application. Imaging times were taken from timestamps on scans. The time interval between imaging acquisition and NES team notification was obtained by subtracting the two recorded times. Total treatment time was defined as the time interval between initial presentation to final reperfusion time.
Patient selection
In accordance with system-wide institutional stroke protocols, patients presenting with suspected acute ischemic stroke were typically initially evaluated with CT and CTA. A CT perfusion scan was added if a patient’s time from last known normal was greater than 6 hours. Appropriate patients with ELVO were selected for ET with a variety of techniques including stent retriever and direct aspiration per institutional standard of care based on factors such as baseline function, NIHSS score, and presence of penumbra. All consecutive patients who underwent ET at our health system between January 1, 2017 and August 5, 2020 were recorded in the database. Patients were excluded from our analysis if prespecified time points were unavailable (hospital arrival, initial CTA, neuroendovascular team notification, or Thrombolysis in Cerebral Infarction 2b+ times) or if the NES team was notified before CTA acquisition.
Statistical analysis
Our primary outcome was the time from image acquisition, defined as time of first CTA imaging, to NES team notification for each patient. The mean, SD, and median of the dataset were assessed. The data were dichotomized into rapid (≤20 min) and delayed (>20 min) notification times. This time point was based on American Heart Association/American Stroke Association (AHA/ASA) target: Stroke best practice guidelines suggesting a goal of 20 min from initial CTA scan to neuroradiological interpretation.9 If the NES team was notified of a patient before CTA was acquired, that patient was moved to the prenotification cohort. Statistical analyses were performed on Microsoft Excel and RStudio version 1.3. Data distributions were evaluated (online supplemental figure 1). Normally distributed, not normally distributed, and categorical variables were compared using t test, Kruskal-Wallis test, and χ2 test, respectively. P values <0.05 were considered significant. Matched propensity score analysis between the rapid and delayed groups was conducted with the MatchIt package in R.
Supplemental material
Results
Between January 1, 2017 and August 5, 2020, 621 patients underwent ET within our multihospital health system (figure 1). Of those cases, 147 were excluded because the NES team was notified before image acquisition (the prenotification cohort). Ninety-six patients were excluded owing to incomplete time data, and 11 patients were excluded owing to a delayed time course (total time window >24 hours). Of the remaining 367 patients meeting the inclusion criteria, the median time from CTA acquisition to endovascular team notification was 24 min (IQR 12–47). The median total treatment time was 180 min (IQR 129–252) (table 1). Patient demographics and arrival modes are shown in table 1.
Flow chart demonstrating patient selection and data acquisition. CTA, CT angiography; NES, neuroendovascular surgery.
Descriptive characteristics of rapid and delayed patients treated within a 24-hour window
Given the AHA/ASA guidelines targeting 20 min for neuroradiological interpretation following initial CTA imaging,9 10 we bisected the data into rapid (≤20 min) and delayed (>20 min) notification times and compared these cohorts. As shown in table 2, the median time from imaging to NES team notification for rapid (n=163) and delayed notification (n=204) were 11 (IQR 6–15) and 43 (IQR 30–80) min, respectively. We found that the median overall times to reperfusion were 134 min (IQR 103–179) and 213 min (IQR 172–291) in the two groups, respectively (p<0.001), leading to a time difference of 79 min between the two groups. The pre-stroke mRS scores were not statistically different between rapid and delayed notification groups. The delayed patients had a significantly lower NIHSS score on presentation (15 (IQR 9–20) vs 16 (IQR 11–22), p=0.03). Delayed patients were younger (70 (IQR 60–79) vs 77 (IQR 64–85), p<0.001) and more often presented with posterior circulation occlusion (16.7% vs 7.4%, p<0.01). There was no difference in the sex of the patients between cohorts (p=0.48). There was no significant difference in the time of presentation between cohorts (using workday definition of Monday to Friday 8:00-17:00, p=0.23). The rapid notification group had a statistically larger median change in NIHSS score from admission to discharge of one point (6 (IQR 1–14) vs 5 (IQR 1–10), p=0.04), although there was no significant difference in patients with a change in NIHSS score ≥4 (64% vs 54%, p=0.08). However, in a matched propensity score analysis of rapid and delayed cohorts (rapid matched=130, delayed matched=130), controlling for age, sex, pre-stroke mRS score, and presenting NIHSS score, there was no significant difference in the change in NIHSS score (p=0.35). For the 147 prenotification patients who were excluded because imaging was obtained after NES team notification, the median total treatment time was 124 min, 10 min faster than our rapid notification cohort (figure 2).
Notification delays lead to total treatment delays. Median time delay in notifying the neuroendovascular surgery (NES) team in relation to the median total times from presentation to final reperfusion. The prenotification cohort represents patients who had NES consultation prior to imaging. The rapid cohort represents patients for whom rapid communication was achieved (within 20 min), whereas the delayed cohort represents patients for whom a delayed notification (after 20 min) was relayed.
Rapid and delayed patient cohorts described by treatment times and stroke characteristics
We also looked at the difference in time cohorts based on mode of hospital arrival for all patients available in our database (rapid, delayed, and prenotification). As shown in table 3, we found that the median time from CTA to NES team notification in minutes was significantly faster for patients arriving to the hospital by EMS compared with transfer patients (11 (IQR 0–26) vs 20 (IQR 4–44), p<0.001). The total treatment time in minutes was also significantly faster for EMS patients than for transfer patients (114 (IQR 92–154) vs 201 (IQR 164–260), p<0.001). Furthermore, the proportion of patients in the rapid, delayed, or prenotification cohorts was statistically different in the EMS patient group compared with transfer patients (p<0.001). There was no significant difference between EMS and transfer pre-stroke mRS score. We found that the median presenting NIHSS score was statistically lower in EMS patients than in patients transferred (15 (IQR 11–20) vs 17 (IQR 12–21), p=0.045). The median change in NIHSS score was greater in EMS patients than in transfer patients (8 (IQR 2–15) vs 5 (IQR 0–11), p<0.001), and more EMS patients had a change in NIHSS score ≥4 (70% vs 53.7%, p<0.01). However, the median 90-day mRS scores were not significantly different by mode of arrival.
EMS and transfer patient cohort comparison
Discussion
Delays in image acquisition to endovascular team notification
In our population of patients who underwent thrombectomy for ELVO, the median time difference between initial CTA acquisition and endovascular team notification was 24 min. Delays in this time period led to proportionally greater delays in overall time to reperfusion (figure 2). These results indicate that time from radiological image acquisition to neuroendovascular team activation is an important period of time within the acute stroke workflow. These results demonstrate the importance of prompt intra-hospital team communication in the acute stroke response.
By dichotomizing the data into ≤20 min and >20 min from initial image acquisition to neuroradiological interpretation, we gain a better idea of optimal, rapid communication time benchmarks, which are recommended by the AHA/ASA.9 11 12 In cases with expedited communication to the endovascular team, the median notification time was 11 min (IQR 6–15), and final reperfusion after thrombectomy was obtained in a median of 134 min (IQR 103–179). In cases of delayed communication, the median notification time was significantly longer at 43 min (IQR 30–80). In delayed cases, reperfusion after thrombectomy was obtained in 213 min (IQR 172–291), over 1 hour longer than for cases in which communication was optimized. These time differences may be explained in part by the number of patients with severe disease in the rapid and delayed groups. The rapid group had a significantly higher median presenting NIHSS score (16 (IQR 11–22) vs 15 (IQR 9–20), p=0.032) and were significantly older (77 (IQR 64–85) vs 70 (IQR 60–79), p<0.001). These findings may suggest that clinical scoring tools are not sensitive enough to triage ELVOs with precision. Delayed patients more often presented with posterior circulation occlusion (16.7% vs 7.4%, p<0.01). This significant finding is in line with previous studies, which have found that treatment delays occur in patients with posterior circulation strokes compared to those with anterior circulation strokes. Posterior circulation strokes (vertebral, basilar, and posterior cerebral arteries, and their branches) can be associated with less obvious clinical presentation and lower presenting NIHSS scores.13–15 There was no difference in the sex of the patients between cohorts (p=0.48). In contrast to previous reports that time from CT to groin puncture was shorter during working hours (Monday to Friday 8:00-17:00), we found no significant difference between the number of working hour interventions done in the rapid and delayed groups.16 Other potential reasons for notification delay may include determination of eligibility (such as unknown time of onset) and acute concomitant medical conditions (such as hypertension or coagulopathies).17 Interestingly, NES team notification occurring before imaging (prenotification) yielded a final reperfusion time of 124 min, 10 min faster than the postimaging rapid notification group. The existence of this subgroup represents the results of implementing early communication policies, based on clinical suspicion of the primary stroke response team rather than imaging findings at some of the sites in the health system.
Given the well-established impacts of mode of arrival on treatment workflow,18 we also directly compared patients who arrived by EMS with patients who were transferred. We found that EMS patients had significantly shorter notification times (11 (IQR 0–26) vs 20 (IQR 4–44), p<0.001) and total treatment times (114 (IQR 92–154) vs 201 (IQR 164–260), p<0.001) than transfer patients. In our hospital system, transfers are initiated if the patient meets the criteria for severity and is at a hospital location that is not thrombectomy-capable. In addition to previously addressed delays, such as determination of eligibility and acute concomitant medical conditions, notification delays in the transfer cohort may be exacerbated by clerical or logistical barriers. While the distribution of patients with prenotification, rapid, and delayed NES team notification were significantly different between EMS and transfers, analysis of the EMS cohort shows that delayed notification is a significant problem (31.1%) independent of mode of arrival.
Past solutions to stroke workflow delays
Prior analyses of stroke care workflow have resulted in the identification of time windows at risk for delay. Targeting these for specific intervention has resulted in significant time reductions.19 20 These solutions have been compared with the detection of acute myocardial infarction by electrocardiograms due to similar needs for time sensitivity, center-specific care, and imaging.21 In particular, the importance of prehospital triage has been evaluated at multiple windows.22 Diagnostic accuracy by both the general public and EMS has been improved by implementing stroke field severity scales associated with a high probability of ELVO, including the three-Item Stroke Scale (3I-SS), the Los Angeles Motor Scale (LAMS), the Rapid Arterial Occlusion Evaluation (RACE) scale, the Cincinnati Pre-hospital Stroke Severity Scale (CPSSS), the LEGS (legs, eyes, gaze, speech) score, and the VAN (vision, aphasia, neglect) screening tool.19 These have further been prospectively validated in large trials.23–25
Prehospital triage has been optimized further by mobile imaging. Transcranial ultrasound has demonstrated a sensitivity of 78% and specificity of 98% for the diagnosis of middle cerebral artery or internal carotid artery occlusion compared with CTA/MRA performed at receiving hospitals.26 Mobile stroke units incorporating portable CT scanners have begun to enable accurate diagnosis in the field and reduction of onset to groin puncture times by triaging patients to appropriate, thrombectomy-capable stroke centers.27 Hospital transfer and triage protocols have been implemented in several regions of the country and recent early analyses have borne out reduced treatment times.28–32 Prehospital notification of patients with suspected stroke has also been found to reduce time to treatment.33
Benefits to clinical and cost outcomes
Since 1993, the phrase ‘time is brain’ has emphasized the emergent nature of stroke treatment.34 Faster time to endovascular treatment is specifically associated with better clinical outcomes. In patients experiencing a typical large vessel acute ischemic stroke, 1.9 million neurons, 14 billion synapses, and 12 km (7.5 miles) of myelinated fibers are lost each minute.35 An analysis of 390 patients who achieved substantial reperfusion with endovascular thrombectomy showed that each 1 hour delay to reperfusion was associated with increased degree of disability and decreased functional independence.28 Analysis of pooled data involving 1248 patients over a 10-year period showed that significant improvement has occurred in procedure times, without corresponding improvements in last known normal to puncture times. Prolonged last known normal to puncture times were significantly associated with a decreased chance of good outcome (OR=0.84, 95% CI 0.76 to 0.92; p=0.0004). Thus, rapid triage and neuroendovascular team notification is critical.36
Furthermore, shorter endovascular treatment times are associated with healthcare cost savings. Analysis of patient data from the seven-trial HERMES collaboration found that within the first 6 hours of stroke onset, every hour of thrombectomy delay increased the cost of quality-adjusted life years (QALYs) by US$6173/QALY in healthcare costs. These healthcare costs were coupled with a US$7597/QALY increase in societal costs.37
Potential solutions to expedite neuroendovascular consultation
As noted, one potential solution to expedite neuroendovascular notification is a notification based on high clinical suspicion for ELVO, which has been used in our institution. We found that prenotification was used in 28.6% of cases, but a recent multinational study found that prenotification may be used in 70% of cases.38 Although this may lead to the fastest times to treatment, the rate of false notification resulting in wasted resources, cost, and potential team burnout is significant.39–41 For example, Reddy et al found that as many as 73% of clinically detected patients later met the thrombectomy ineligibility criterion.40 The positive and negative predictive value of clinical scales varies between 0.6 and 0.7.22 In our cohort, notifying the neuroendovascular team of a potential ELVO led to nearly an hour faster time to reperfusion among all patients, illustrating the overall benefits of this philosophy. However, in comparison with patients for whom rapid imaging review was performed (rapid cohort), prenotification of the NES team increased the reperfusion time by only 10 minutes, confirming the significant advantage of a rapid triage.
Given these findings, solutions are certainly needed to expedite stroke triage and widespread notification throughout the comprehensive stroke team. Currently, interpretation of non-invasive imaging modalities, including Alberta Stroke Program Early CT Score on non-contrast CT head, CT angiogram, and CT perfusion, varies based on expertise, institution, and time delays.42 As recent research has demonstrated, quality improvement processes may reduce this notification period.3 Our present results show that identifying at-risk time periods may enhance the efficiency of the stroke team communications. Over the past decade, software solutions have emerged to deal specifically with image identification and rapid communication of results. Large vessel occlusion detection with artificial intelligence uses machine learning algorithms to quantify stroke core and penumbra size and mismatch, detect vascular branch occlusion, and predict complications.43 Current stroke detection software platforms include Viz.ai, Brainomix, General Electric, and iSchemaView. Viz.ai received de novo regulatory clearance from the Food and Drug Administration in February 2018 for the first computer-aided triage and notification platform to identify LVO strokes in CTA imaging.43 Such software platforms are currently in clinical use, targeting the window of time identified in this study, to reduce delays in patient selection for endovascular thrombectomy. Future studies will quantify their potential impact on stroke care workflow.
Limitations
For our analysis of time points in the acute stroke workflow, limitations must be discussed. First, we included only patients who ultimately underwent thrombectomy, as this is the patient population included in our institutional database. This necessarily eliminates patients who were otherwise not candidates for thrombectomy. However, analysis of this population still gives an idea of the workflow and where to target high-risk time windows. In addition, we were only able to include patients with all time points available in the database, which could create bias.
Conclusion
Time delays from initial CT acquisition to neuroendovascular team notification can prevent expedient treatment with endovascular thrombectomy. Process improvements and automated stroke and ELVO detection on imaging with automated notification of the NES team may ultimately improve time to reperfusion. Targeting this time window may improve patient outcomes and enhance cost efficacy of thrombectomy.
Data availability statement
Data are available from the corresponding author upon reasonable request.
Ethics statements
Patient consent for publication
Ethics approval
Approval was obtained from the institutional review board of the Icahn School of Medicine at Mount Sinai, approval #19–00956.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
Twitter @cprossitto, @JacquesJLR, @Shoirahz
Contributors KAY is guarantor for the manuscript. KAY and JTF devised the project. CPR, NFM, and JL-R collected the data. CPR performed statistical analysis. KAY and CPR wrote the manuscript with input from all authors. TL and TH assisted with data collection and manuscript preparation. HS and JM were involved in planning and provided critical feedback. JTF supervised the project. All authors discussed the results and commented on the manuscript.
Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.
Competing interests JM receives research funding from Stryker Neurovascular, Microvention, and Penumbra. He serves as a consultant for Imperative Care, Cerebrotech, Viseon, Endostream, Vastrax, RIST, Synchron, Viz.ai, Perflow, and CVAid. He is an investor in Imperative Care, Cerebrotech, Viseon, Endostream, Cardinal Consulting, RIST, Synchron, Viz.ai, BlinkTBI, Serenity, and Truvic.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.