Background Patients with an acute ischemic stroke (AIS) due to large vessel occlusion often require transfer to an endovascular center for treatment.
Objective To assess the effect of hospital transfer on outcomes after endovascular revascularization.
Methods Outcomes of endovascular revascularization were compared between directly admitted and transferred patients using data from a national database and our own institution.
Results 118 institutions within the database reported outcomes of 8533 inpatient admissions for endovascular treatment of AIS. Mortality rate (14.9% vs 18.6%; p=0.049) and mortality index (1.1 vs 1.6; p=0.048) were significantly lower among directly admitted patients than among transferred patients. Within our institutional cohort of 140 patients who underwent endovascular therapy, directly admitted patients had a significantly faster time to revascularization than transferred patients (277.4 vs 420.4 min; p≤0.0001). Among transferred patients, an increasing distance of transferred hospital to our home institution was associated with an increasing risk of mortality (unit OR=1.26, 95% CI 1.07 to 1.54; p=0.0061).
Conclusions Outcomes of revascularization may improve with methods to identify patients with large vessel occlusion before hospital admission, thus increasing the likelihood of initial triage to a comprehensive stroke center for patients eligible for endovascular intervention.
Statistics from Altmetric.com
Recent clinical trials have validated endovascular revascularization as an effective therapy for acute ischemic stroke (AIS) secondary to large vessel occlusion.1–5 Though a large majority of the US population reside within 6 hours of an endovascular center,6 endovascular revascularization is performed in a minority of facilities across the country,7–8 often necessitating transfer of patients to a facility with endovascular capabilities. While hospital transfer has been shown to delay revascularization,9 the effect of this delay on patient outcomes has not been thoroughly investigated. Herein, we compared outcomes after endovascular treatment of AIS between patients who presented initially to the hospital at which they received endovascular intervention and those who were transferred from an outside facility using data from a national database and from our own institution.
National database study
The Vizient, formerly known as University HealthSystem Consortium, Clinical Database/Resource Manager (CDB/RM) is a database of hospital-reported outcomes that is updated quarterly with reports from over 270 institutions around the country. Outcomes of inpatient admissions are reported voluntarily to the CDB/RM. Patients of interest are identified through International Classification of Diseases-9th and 10th Edition diagnosis and procedural codes. Reported outcomes include hospital-level standard outcomes, such as length of stay, incidence of complications, and mortality, as well as more specific complications tracked by the CDB/RM. Outcomes are reported as mean values for each institution —for example, mean length of stay of patients admitted to a particular institution. Patients admitted to a hospital for treatment of AIS between October 2012 and June 2016 were identified using variations of the ICD-9 codes 433, 434, 436, and 437 and ICD-10 code I63. Patients who subsequently were treated for endovascular revascularization were identified using the ICD-9 code 3974 and variations of the ICD-10 code 03C. Hospitals were included for analysis if they reported outcomes from at least five inpatient admissions for endovascular treatment of AIS.
Patient characteristics and outcomes
Information on patient age, sex, and admission source was collected. Administration of IV thrombolysis (IV tissue plasminogen activator (tPA)) was determined using the ICD-9 code 9910 and ICD-10 codes 3E03317 and 3E04317. The primary outcome of interest was mortality during admission. The CDB/RM reports both observed and expected mortality, allowing the calculation of a mortality index, or the ratio of observed to expected mortality. Expected mortality is adjusted for patient demographics and comorbidities using validated risk prediction models.10–13 Secondary outcomes included the rate of intracerebral hemorrhage and discharge to home. Patients who had an intracerebral hemorrhage were identified using the ICD-9 code 431 and variations of the ICD-10 code I61. Patient characteristics and outcomes were determined for all patients and then separately for directly admitted and transferred patients.
After institutional review board approval we retrospectively reviewed outcomes of patients who had undergone endovascular revascularization at our institution between the years of 2009 and 2016. Patients initially deemed inappropriate for endovascular therapy on admission who subsequently declined and underwent intervention were excluded (n=12/184, 6.5%). Patients who experienced an AIS while already admitted to our hospital were also excluded (n=25/184, 13.6%), as were patients who underwent diagnostic angiography only (n=7/184, 3.8%).
Institutional protocol for determining eligibility for endovascular therapy
The protocol for determining eligibility for endovascular therapy for patients presenting to our institution is as follows: upon detection of focal neurologic symptoms consistent with a stroke syndrome, a stroke code is activated leading to prompt evaluation of the patient by the on-call neurologist. Vascular imaging is obtained on the basis of the severity of the patient's neurologic symptoms and the discretion of the evaluating neurologist. If a large vessel occlusion amenable to endovascular recanalization is visualized, the decision to proceed with endovascular therapy is determined primarily by the time from symptom onset. For patients presenting beyond the 6-hour treatment window, endovascular therapy is pursued only if there is no evidence of an established infarct on CT imaging or if there is evidence of an ischemic penumbra on CT perfusion studies. The protocol is similar for transferred patients, who are typically received in the emergency department, thus facilitating rapid neurologic examination and subsequent imaging studies.
Patient characteristics and outcomes
We collected information on patient age, sex, admission source, distance in miles between ZIP codes of the transferring hospital and our institution, National Institutes of Health Stroke Scale (NIHSS) score on presentation, location of large vessel occlusion, dominance of the hemisphere affected, administration of an IV thrombolysis agent (IV tPA), time to endovascular revascularization in minutes, Alberta Stroke Program Early CT score (ASPECTS), recanalization score, incidence of intracerebral hemorrhage, incidence of in-hospital mortality, rate of discharge to home, and 90-day functional outcome as measured by the modified Rankin Scale (mRS). NIHSS was determined from the neurologic examination performed by the admitting neurologist in our emergency room for both directly admitted and transferred patients. Time to revascularization was considered as the time from symptom onset to the time of initial recanalization. For patients with an unknown time of onset (n=41/140, 29.3%), the time at which the patient was last normal was considered the time of onset. ASPECTS was determined from the final CT scan obtained before revascularization using previously described methodology.14 The proportion of patients with ASPECTS <9 was also determined. Recanalization score was described according to the Thrombolysis in Cerebral Infarction (TICI) criteria. The presence of intracerebral hemorrhage was determined by review of CT scans obtained after revascularization. Hemorrhages were classified as hemorrhagic infarcts versus parenchymal hematomas (PHs). PHs were further classified as PH1 or PH2 hemorrhages, with criteria for PH2 consisting of hemorrhage occupying >30% of the area of infarction with significant mass effect, as previously described.15 Patients were categorized as having an intracerebral hemorrhage if the hemorrhage met criteria for a PH2 hemorrhage. For functional outcome, patients were categorized as having an mRS score of ≤2 or >2. mRS was determined at roughly 90 days after revascularization and was based on the evaluation of a stroke neurologist, physical medicine and rehabilitation physician specializing in brain rehabilitation, or an allied health staff member trained to determine the mRS during a clinic visit or telephone encounter.
Statistical analysis for national database and institutional data
Descriptive statistics for continuous and categorical variables were reported as a mean (SD) and frequency (percentage), respectively. For the database study, outcomes for directly admitted and transferred patients were compared using the paired two-sample Student's t test. Comparison of patient characteristics and outcomes between low, medium, and high volume centers was performed using one-way analysis of variance and the unpaired Student's t test.
For our institutional data, outcomes for directly admitted and transferred patients were compared with the unpaired Student's t test and Pearson's χ2 test, when appropriate. Predictors of intracerebral hemorrhage, in-hospital mortality and functional outcome at 90 days were identified using univariate logistic regression analysis. Independent predictors of patient outcomes were identified using multivariate logistic regression analysis. Owing to the small number of intraparenchymal hemorrhages and deaths within our institutional dataset, transfer from an outside institution was included in a 1:1 fashion, with other variables in regression models determining independent predictors of these two outcomes. Transfer was included with multiple other variables in models identifying predictors of poor functional outcome. The relationship between distance of transferring hospital and patient outcomes was also investigated using univariate and multivariate logistic regression analysis. Similarly to transfer, distance was included in a 1:1 fashion, with other variables in multivariable models determining predictors of intracerebral hemorrhage and mortality. All statistical tests were two-sided with an α level for statistical significance set at 0.05. All analyses were performed using commercially available software (JMP V.10.0.0, 2012 SAS Institute Inc).
National database study
One hundred and eighteen institutions reported outcomes from 8533 inpatient admissions for the endovascular treatment of AIS that met our inclusion criteria for analysis. The mean number of cases reported by each institution was 72.3. The percentage of reporting institutions that were AAMC teaching hospitals was 89.0% (105/118), while 72.9% (86/118) of institutions were certified as stroke centers by the Joint Commission. Reporting hospitals were roughly evenly distributed across the mid-Atlantic (23.7%), mid-continent (16.1%), Midwestern (20.3%), New England (10.2%), southeastern (17.8%) and western (11.9%) regions of the USA. The mean percentage of patients aged ≥65 years was 61.1%, and 49.1% of patients were female. The mean percentage of patients who were transferred from another facility was 36.1%, while the mean percentage of patients who received IV tPA while admitted to the endovascular center was 35.9%. The mean rate of intracerebral hemorrhage was 18.7%. The mean mortality rate was 15.6%, while the mean mortality index was 1.2. Finally, the mean percentage of patients who were ultimately discharged to home was 17.2% (table 1).
Outcomes were then compared between directly admitted and transferred patients at each hospital. The percentage of directly admitted patients who received IV tPA at the endovascular center was significantly greater than the percentage of transferred patients (47.6% vs 17.1%; p<0.001). The rate of mortality (14.9% vs 18.6%; p=0.049) and mortality index (1.1 vs 1.6; p=0.048) were significantly lower among directly admitted patients than among transferred patients. There was no difference in the rate of intracerebral hemorrhage (18.7% vs 17.2%; p=0.314) or discharge to home (17.7% vs 16.7%; p=0.589) between directly admitted and transferred patients (table 2).
One hundred and forty patients met our inclusion criteria for analysis. A majority of patients (78/140, 55.7%) were transferred from another hospital. The NIHSS score at presentation to our institution was similar between directly admitted and transferred patients (17.8 vs 18.5; p=0.472), as was the location of large vessel occlusion (p=0.121). Time to revascularization was significantly faster in directly admitted relative to transferred patients (277.4 vs 420.4 min; p<0.001). This difference remained even after patients with an unknown time of onset were excluded from the analysis (217.6 vs 316.4 min; p<0.001). There was a trend towards greater ASPECTS among directly admitted patients than among transferred patients (9.2 vs 8.8; p=0.055). The proportion of patients with an ASPECTS <9 was significantly lower among directly admitted patients (19.4% vs 39.7%; p=0.008). There was no difference in the success of revascularization as judged by the TICI score between directly admitted and transferred patients (p=0.665). We did not observe significant differences in the incidence of PH2 intracerebral hemorrhages (3.2% vs 10.3%; p=0.109), mortality (12.9% vs 19.2%; p=0.316), or proportion of patients with an mRS >2 at 90 days (61.7% vs 66.2%; p=0.580) between directly admitted and transferred patients (table 3).
Predictors of PH2 intracerebral hemorrhages, death, and mRS >2 among patients treated with endovascular revascularization were then identified by univariate logistic regression analysis. Increasing time to revascularization (unit OR=1.03, 95% CI 1.00 to 1.05; p=0.025) and a TICI revascularization grade of ≥2b (OR=3.34, 95% CI 1.81 to 9.42; p=0.011) were found to be significantly associated with an increased rate of intracerebral hemorrhage. Increasing age (unit OR=1.07, 95% CI 1.02 to 1.12; p=0.001), NIHSS (unit OR=1.12, 95% CI 1.03 to 1.24; p=0.011), and basilar occlusion (OR=6.53, 95% CI 1.85 to 23.25; p=0.004) were associated with an increased likelihood of death. A TICI revascularization grade of ≥2b was associated with a decreased likelihood of death (OR=0.31, 95% CI 0.12 to 0.79; p=0.015; table 4). Finally, increasing age (unit OR=1.03, 95% CI 1.00 to 1.06; p=0.025) and NIHSS (unit OR=1.12, 95% CI 1.05 to 1.21; p<0.001) were associated with an increased likelihood of an mRS >2 at 90 days. A recanalization grade of ≥2b was associated with a decreased likelihood of mRS >2 (OR=0.40, 95% CI 0.16 to 0.92; p=0.031). The complete results of the univariate analysis are shown in table 4. To further assess the effect of hospital transfer on patient outcomes, transfer was included as a variable in separate multivariate logistic regression models for intracerebral hemorrhage, death, and mRS >2. We found no independent relationship between transfer and these outcomes on multivariate analysis (table 5).
For transferred patients we also determined the distance of the transferring hospital to our home institution. The average distance in miles was 61.2 (SD 31.9), with a range of 16.0–196.9 miles. We observed no linear correlation between distance traveled and time to revascularization (r2<0.001). The relationship between distance traveled and intracerebral hemorrhage, death, and 90-day mRS >2 was then assessed using univariate logistic regression analysis. Distance was not found to be associated with risk of intracerebral hemorrhage (unit OR=1.07, 95% CI 0.85 to 1.29; p=0.500; table 6). On multivariate analysis, distance was again not associated with intracerebral hemorrhage, though after adjusting for a TICI revascularization grade of ≥2b, there was a slight trend towards an increased incidence of hemorrhage with increasing distance (unit OR=1.15, 95% CI 0.98 to 1.36; p=0.77; table 7). Distance traveled was found to be significantly associated with an increased risk of death among transferred patients (unit OR=1.26, 95% CI 1.07 to 1.54; p=0.006; table 6). On multivariate analysis, distance remained associated with death after controlling for patient age (unit OR=1.29, 95% CI 1.08 to 1.61; p=0.004), NIHSS (unit OR=1.23, 95% CI 1.03 to 1.51; p=0.017), basilar occlusion (unit OR=1.27, 95% CI 1.07 to 1.56; p=0.005), administration of IV tPA (unit OR=1.02, 95% CI 1.01 to 1.05; p=0.014), time to intervention (unit OR=1.26, 95% CI 1.07 to 1.54; p=0.006), ASPECTS (unit OR=1.29, 95% CI 1.08 to 1.59; p=0.004), and TICI revascularization grade of ≥2b (unit OR=1.25, 95% CI 1.06 to 1.54; p=0.009; table 7). There was a trend towards an association of distance with mRS >2 at 90 days on univariate analysis (unit OR=1.19, 95% CI 0.99 to 1.48; p=0.061; table 6), though this association was not significant. Distance was not found to be associated with mRS >2 on multivariate analysis (unit OR=1.09, 95% CI 0.89 to 1.41; p=0.424; table 7).
Using data from 8533 admissions from 118 endovascular centers collected within a national database, we observed a 46% increase in adjusted mortality (mortality index: 1.6 vs 1.1) after endovascular therapy for AIS in transferred patients relative to directly admitted patients. Within our institutional cohort, transferred patients had a significantly longer time from symptom onset to revascularization, with an average time to revascularization beyond the 6-hour time window during which endovascular intervention has been shown to provide greatest therapeutic benefit.1–5 Transferred patients also had lower ASPECTS. In addition, increasing distance between the transferring hospital and our institution was associated with an increased odds of mortality.
The mechanism by which hospital transfer leads to increased mortality is not immediately clear from our results, mainly as we did not observe a difference in the rate of intracerebral hemorrhage between directly admitted and transferred patients in the multi-institutional or single-center datasets. It is possible that transferred patients were more likely to remain neurologically devastated after endovascular therapy, leading to an increased frequency of withdrawal of care. An additional possibility is that patients in the transferred group were more likely to develop malignant cerebral edema, given the associated delay in treatment. Whatever the mechanism, the detrimental effect of transfer is probably related to a delay in treatment and establishment of infarction, as shown by the increased time to revascularization and lower ASPECTS, respectively, among transferred patients. Further investigation into how the delay associated with transfer contributes to increased mortality is warranted.
Although our findings need to be confirmed, we believe a negative effect of hospital transfer on patient outcomes would have important implications for the organization and planning of endovascular treatment networks and stroke care in general. Despite the increase in endovascular therapy for AIS over the past decade,16 the number of healthcare institutions with endovascular capabilities remains relatively small,6–8 particularly in light of the number of potentially eligible patients.17 Arguments for the centralization of care at specialized centers are based primarily on the correlation between increasing procedural volume and improved patient outcomes.7 ,18 Our results suggest, however, that the need for hospital transfer associated with centralization may blunt the beneficial effect of centralization, especially if patients require transfer from increasingly remote areas, particularly as the protocols for accepting transferred patients become maximally streamlined. Ultimately, careful planning of endovascular treatment networks will be needed to maximize the benefit of centralized care while mitigating the effects of hospital transfer.
Current American Heart Association/American Stroke Association guidelines state that upon manifestation of stroke symptoms patients should be transferred to the nearest primary or comprehensive stroke center,19 a practice that has undoubtedly increased the number of patients receiving IV tPA within the 4.5-hour window in which it has been shown to provide therapeutic benefit.20 Our results suggest, however, that patients with large vessel occlusion may benefit from initial transportation to an endovascular center, at least those patients without the possibility of transfer to a high volume center. The difficulty with this practice, of course, is the identification of patients with large vessel occlusion before diagnostic vascular imaging. We would caution against the indiscriminate bypass of local primary stroke centers, as it would probably increase the number of patients who would not receive IV tPA owing to arrival at a stroke center beyond the treatment window, including some who would inevitably be ineligible for endovascular intervention. Ultimately, protocols at primary centers for the identification of endovascular candidates and prompt transfer to an endovascular center should continue to be streamlined. In addition, efforts facilitating the accurate diagnosis of large vessel occlusion outside a hospital should be pursued.
Our analysis of national outcomes was limited by the lack of information on preprocedural NIHSS score, which has been shown to influence the likelihood of successful revascularization and subsequent good functional outcome.21–23 We also observed significant differences in the rate of IV tPA administration in the national database between directly admitted and transferred patients. We suspect this is explained, at least in part, by the administration of IV tPA to many of the transferred patients before transport. However, it is possible that differences in IV tPA administration affected outcomes between directly admitted and transferred patients. Alternatively, directly admitted and transferred patients might have had differing baseline characteristics, leading to a greater frequency of contraindications for IV tPA administration in the transferred group. Irrespective of the differences between directly admitted and transferred patients, the overall rates of IV tPA administration were 31.8% and 58.6% within the multi-institutional and single-center datasets, respectively, which are much greater than previous estimates placing the national rate of IV tPA administration to eligible patients at 3.4–5.2%.24 This discrepancy may reflect differences between our study population and the population at large, potentially limiting the generalizability of our results. On the other hand, the rate of IV tPA administration may be higher among patients with AIS owing to large vessel occlusion, as shown by the high rate of tPA administration (75.3–89.0%) to patients eligible for revascularization before randomization seen in recent clinical trials.1 ,3 Contemporary data on average rates of IV tPA administration to patients undergoing endovascular revascularization are needed to determine the generalizability of our results.
The frequency of intracerebral hemorrhage was markedly elevated within the multi-institutional database relative to patients within our institutional cohort (tables 1⇑ and 3). It is likely that clinically insignificant hemorrhages were counted as intracerebral hemorrhages within the national dataset, which might have limited our ability to detect an effect of transfer on meaningful complications. Outcomes to the CDB/RM are self-reported by affiliated institutions and not subject to mandatory reporting policies, thus it is possible that selection bias affected our results. In addition, the mean number of cases reported by each institution over the study period was 72.3, translating into only 19.7 cases a year, a rather small number. The range of the number of cases reported was broad (5–621), however, reflecting the reality that a minority of specialized center perform a large proportion of endovascular procedures for AIS around the country.6–8 Nevertheless, it is possible that the small annual number of cases is in part explained by institutions not reporting the outcomes of a significant number of patients treated at their facility, potentially skewing our results.
In this study, we present evidence that transfer from another hospital is associated with increased mortality among patients treated with endovascular revascularization for AIS. Our results highlight the need to identify endovascular candidates before initial hospital presentation so that they can be more promptly directed towards endovascular centers.
Contributors LR: study design, data collection, statistical analysis, drafting of manuscript, critical revision; WB, AAR: study design, critical revision; BAM: statistical analysis, critical revision; MB, HC, DFK: critical revision.
Competing interests None declared.
Ethics approval Institutional review board.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement The data used for this article were obtained partially from a national database; these data are available to members. Data from our institution are available to the authors only.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.