Deconstruction of Interhospital Transfer Workflow in Large Vessel Occlusion
Real-World Data in the Thrombectomy Era
Background and Purpose—Interhospital transfer is a critical component in the treatment of acute anterior circulation large vessel occlusive stroke transferred for mechanical thrombectomy. Real-world data for benchmarking and theoretical modeling are limited. We sought to characterize transfer workflow from primary stroke center (PSC) to comprehensive stroke center after the publication of positive thrombectomy trials.
Methods—Consecutive patients transferred from 3 high-volume PSCs to a single comprehensive stroke center between January 2015 and August 2016 were included in a retrospective study. Factors associated with key time metrics were analyzed with emphasis on PSC intrahospital workflow.
Results—Sixty-seven patients were identified. Median age was 74 years (interquartile range [IQR], 63.5–78) and National Institutes of Health Stroke Scale 17 (IQR, 12–21). Median transfer time measured by PSC-door-to-comprehensive stroke center-door was 128 minutes (IQR, 107–164), of which 82.8% was spent at PSCs (door-in-door-out [DIDO]; 106 minutes; IQR, 86–143). The lengthiest component of DIDO was computed-tomography-to-retrieval-request (median 59.5 minutes; IQR, 44–83). The 37.3% had DIDO exceeding 120 minutes. DIDO times differed significantly between PSCs (P=0.01). In multivariate analyses, rerecruiting the initial ambulance crew for transfer (P<0.01) and presentation during working hours (P=0.04) were associated with shorter DIDO times.
Conclusions—In a metropolitan hub-and-spoke network, PSC-door-to-comprehensive stroke center-door and DIDO times are long even in high-volume PSCs. Improving PSC workflow represents a major opportunity to expedite mechanical thrombectomy and improve patient outcomes.
Mechanical thrombectomy in acute anterior circulation large vessel occlusion has been heralded as a new era in modern stroke care.1 Currently, patients are assessed at primary stroke centers (PSCs) with urgent interhospital transfers to comprehensive stroke centers (CSCs) when required.
Efficacy of thrombectomy is critically time dependent.2 Earlier studies suggested that interhospital transfers are major contributors to prolonged ischemic time and poor outcomes.3 To date, data on transfer workflow in the evidence-based thrombectomy era are limited. The paucity of real-world data essential for quality improvement benchmarking and theoretical modeling may limit development of thrombectomy service models. We sought to characterize interhospital transfer workflow in a metropolitan hub-and-spoke system in Australia.
Consecutive patients presenting to 3 PSCs with acute anterior circulation large vessel occlusion and transferred to a single CSC for thrombectomy, between January 2015 and August 2016, were included. Data were collected from prospective departmental databases and transfer records from the state-wide ambulance service. Ethical approval was granted by the respective institutional review boards.
The study was conducted in Melbourne where thrombectomy support for PSCs was provided by a sole designated CSC via a hub-and-spoke model. Retrieval requests to ambulance for urgent transfers were initiated after approval from the CSC. The closest available ambulance crew was recruited for road transport (Figure I in the online-only Data Supplement).
Analysis was performed to determine clinical characteristics and process-related factors associated with the following key time metrics: (1) PSC-door-to-CSC-door (D2D) and (2) PSC-door-in-door-out (DIDO). Secondary analyses were performed on transport time and the deconstructed serial components of DIDO, including (1) door-to-computed-tomography (CT); (2) CT-to-retrieval-request; (3) retrieval-request-to-ambulance-arrival; and (4) ambulance-arrival-to-PSC-departure.
Time metrics were analyzed using Mann–Whitney U and Kruskal–Wallis tests as appropriate (SPSS 20.0). Variables with P<0.05 were included in the multivariate linear regression model with log-transformed (normally distributed) DIDO and D2D as dependent variables. P<0.05 was considered significant in the final model.
Of 2776 acute ischemic stroke admissions, 86 patients (3.1%) had interhospital transfers. Patients with inpatient onsets (n=3), symptoms onset beyond 4.5 hour (n=1), incomplete data (n=2), and posterior circulation large vessel occlusion (n=13) were excluded. Sixty-seven patients were analyzed.
The median age was 74 years (interquartile range [IQR], 63.5–78), median National Institutes of Health Stroke Scale was 17 (IQR, 12–21), and 52.2% (n=36) presented after hours (Table I in the online-only Data Supplement). Sixty-two patients (92.5%) received intravenous thrombolysis. Median door-to-needle time was 56.5 minutes (IQR, 42–73). Overall, 97% of patients met trial inclusion criteria for MR CLEAN (Multicenter Randomized Clinical Trial of Endovascular Treatment for Acute Ischemic Stroke in the Netherlands).4 Among those undergoing CT perfusion (CTP) and thrombolysis (n=44), 75% met EXTEND-IA (Extending the Time for Thrombolysis in Emergency Neurological Deficits–Intra-Arterial) inclusion criteria.5
Interhospital Transfer Time Metrics
Median transfer workflow as measured by D2D was 128 minutes (IQR, 107–164) and encompassed 64% of the overall treatment workflow from first point of medical contact to arterial puncture (200 minutes [IQR, 168–213]; Figure; Table II in the online-only Data Supplement). The 82.8% of transfer workflow was composed of DIDO (median 106 minutes; IQR, 86–143). The longest component of DIDO was CT-to-retrieval-request (59.5 minutes, IQR, 44–83). The 17.9% (n=12), 37.3% (n=25), and 16.4% (n=11) had DIDO of <75, >120, and >150 minutes, respectively. The fastest DIDO was 51 minutes.
DIDO differed significantly between PSCs (P<0.01). On univariate analyses, same crew (P<0.01) and presentation during working hours (P<0.04) were associated with shorter DIDO. In a multivariate linear regression model, all 3 factors remained independently associated with DIDO (Table II in the online-only Data Supplement). No other clinical characteristics or process-related factors were associated with DIDO. Same crew and working hours were associated with shorter D2D on univariate analysis, but only same crew was independently associated with D2D on multivariate analysis.
We found interhospital transfers exceed 2 hours even in high-volume PSCs in the evidence-based thrombectomy era. DIDO occupied a substantial proportion of interhospital transfer time and is associated with modifiable factors. Previous literature on stroke interhospital transfer before the MR CLEAN trial was performed without the use of consistent time metrics3 while recent positive thrombectomy trials have only indirectly assessed transfer workflow by comparing onset-to-groin and reperfusion times among transferred and nontransferred patients.6
We assessed transfer workflow through D2D and DIDO, 2 well-established metrics used in transfers for ST-segment–elevation myocardial infarction. These metrics capture workflow processes unique to interhospital transfers without incorporating CSC intrahospital processes. In our cohort, D2D and DIDO were responsible for 64% and 53% of the overall treatment workflow, respectively.
The longest component of DIDO was CT-to-retrieval-request (56.1% of DIDO) where tasks performed include (1) imaging acquisition, reconstruction, and interpretation; (2) acute treatment decision making; (3) imaging transfer to CSC; (4) CSC referral; and (5) transfer retrieval request. Potential inherent inefficiencies exist in all these processes but are not quantified further in this study.
DIDO was associated with multiple process-related factors. Consistent with previous studies, intrahospital workflow slowed after hours, which is likely related to reduced staffing, both in numbers and seniority. Majority of our cohort presented after hours with 25.3% presenting during weekdays 5 pm to 10 pm. Extending on-site stroke team presence to 10 pm may improve efficacy.
Patients transported by the same ambulance crew had faster transfers likely through having continuity of care regardless whether the crew remained on-site while awaiting transfer decision or departed the PSC and subsequently returned. Crews who remained on-site had additional advantages of avoiding return travel time.
DIDO times differed significantly between PSCs, suggesting that inherent intrahospital practice variations are important contributors to workflow delays. Streamlined and unified protocols may reduce process inefficiencies. Finally, in view of recent debate about the role of CTP and thrombolysis in patients with large vessel occlusion, we found neither process prolonging workflow.
Major implications for stroke service organization emerge from our findings. First, major effort should be directed toward optimizing PSC workflow. Although protracted DIDO times support bypassing PSCs for direct CSC presentation, the low rate (3.1%) of transfers emphasizes the role of PSCs in selecting thrombectomy candidates and, more importantly, in providing timely management for the remaining majority of patients. A proficient hub-and-spoke system requires proactive collaboration of all stakeholders. Formal quality improvement initiatives at PSCs with periodic monitoring and interhospital transfer benchmarking will aid stroke care delivery. The unforgiving physiology of cerebral ischemia requires aggressive goal setting. Based on our data, we propose DIDO as a key performance index for PSC workflow with 75 minutes (top 15th centile) as the target time.
Our findings suggest that this ambitious target can be readily achieved via 2 strategies. First, door-to-retrieval-request time (78 minutes) may be reduced to the door-to-needle time (56.5 minutes) if PSCs initiate transfers without awaiting CSC approval in patients clearly meeting well-defined criteria. Vast majority of our cohort met inclusion criteria of recent thrombectomy trials, the current American Heart Association/American Stroke Association (80.6%),7 and European Stroke Organization (92.5%)8 selection guidelines for thrombectomy. A streamlined interhospital transfer referral process would eliminate redundant decision-making time. Accordingly, efforts in improving door-to-needle time will also directly improve transfer efficiency. Second, enlisting ambulance crews to remain at PSCs until definitive transfer decisions will eliminate crew travel time. In combination, our data suggest at least a 30-minute reduction in DIDO which translates to 32 less disabled patients per thousand treated.2
Second, our findings emphasize the value of local real-world patient-level data when using theoretical modeling for stroke service development. The considerable differences in all key metrics between data from this study and the modeling estimates used in recent mathematical models based on North American data9 highlight the importance of up-to-date region-specific data when applying modeling in different jurisdictions.
This study has several limitations. Our data set from 1 hub-and-spoke system may not be generalizable to other networks. Second, we were unable to analyze patient outcomes. Finally, more granular parameters, such as CT acquisition and processing and CSC referral times, were not available.
We provide the first systematic analysis of interhospital transfers in a metropolitan hub-and-spoke model. PSC workflow efficiency is an underappreciated major determinant of interhospital transfer and overall treatment time. Optimizing PSC and transfer workflow is equally, if not more important, than parallel efforts in improving CSC metrics. Further studies and quality improvement efforts incorporating DIDO evaluations are recommended.
The authors acknowledge the contribution of the Neurointerventional Service at Royal Melbourne Hospital (Prof Peter Mitchell, A/Prof Rick Dowling, A/Prof Bernard Yan, Dr Steve Bush) who performed all mechanical thrombectomy procedures. We thank Jason Muller for map graphics and Dr Derek Law for statistical advice.
The online-only Data Supplement is available with this article at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.117.017235/-/DC1.
- Received March 5, 2017.
- Revision received March 29, 2017.
- Accepted March 31, 2017.
- © 2017 American Heart Association, Inc.
- Alberts MJ,
- Shang T,
- Magadan A
- Asif KS,
- Lazzaro MA,
- Zaidat O
- Menon BK,
- Sajobi TT,
- Zhang Y,
- Rempel JL,
- Shuaib A,
- Thornton J,
- et al
- Powers WJ,
- Derdeyn CP,
- Biller J,
- Coffey CS,
- Hoh BL,
- Jauch EC,
- et al
- Fiehler J,
- Cognard C,
- Gallitelli M,
- Jansen O,
- Kobayashi A,
- Mattle HP,
- et al
- Holodinsky JK,
- Williamson TS,
- Kamal N,
- Mayank D,
- Hill MD,
- Goyal M