Article Text

Original research
National trends in endovascular therapy for acute ischemic stroke: utilization and outcomes
1. Laura Stein1,
2. Stanley Tuhrim1,
3. Johanna Fifi2,
4. J Mocco3,
5. Mandip Dhamoon1
1. 1 Neurology, Icahn School of Medicine at Mount Sinai, New York City, New York, USA
2. 2 Neurology, Neurosurgery, and Radiology, Icahn School of Medicine at Mount Sinai, New York, NY, USA
3. 3 Neurosurgery, Icahn School of Medicine at Mount Sinai, New York City, New York, USA
1. Correspondence to Dr Laura Stein, Neurology, Icahn School of Medicine at Mount Sinai, New York City, NY 10029, USA; laura.stein{at}mssm.edu

## Abstract

Objective Following widespread acceptance of endovascular therapy (ET) for large vessel occlusion stroke in 2015, we assessed nationwide utilization of revascularization for acute ischemic stroke (AIS).

Methods We utilized the 2013–2016 Healthcare Cost and Utilization Project Nationwide Readmissions Database. We identified AIS admissions, treatment with intravenous thrombolysis (IVT), ET, and vascular risk factors using International Classification of Disease Clinical Modification codes. Main predictor of outcome was the time period of index admission (‘pre-endovascular era (pre-EA)’ January 2013–January 2015 and ‘endovascular era (EA)’ February 2015– December 2016). We calculated the proportion of AIS admissions in which, first, VT and second, ET was performed. Among patients treated with ET, we examined the association between era and discharge disposition, in-hospital mortality during index admission, and 30-day readmission.

Results There were 925 363 index AIS admissions before the EA and 857 347 during. A higher proportion of AIS patients received IVT (8.4% vs 7.8%) and ET (2.6% vs 1.3%) in the EA. Although length of stay (LOS) was shorter in the EA (5.70 vs 6.80 days), total charges were greater ($56 691 vs$53 878), and admissions were more often to a metropolitan hospital (65.2% vs 57.2%). Among those treated with ET, a smaller proportion received IVT (29.7% vs 44.9%), LOS was substantively shorter (9.75 vs 12.76 days), and patients had a lower odds of discharge home.

Conclusions The utilization of ET has doubled in the EA but ET remains underutilized. ET is predominantly provided at metropolitan teaching hospitals and associated with higher charges despite shorter LOS and unchanged in-hospital mortality.

• intervention
• stroke
• thrombectomy
• thrombolysis

## Introduction

Acute ischemic stroke (AIS) care was revolutionized in 2015 with the establishment of endovascular therapy (ET) as an effective treatment for eligible large vessel occlusion (LVO) stroke.1 In February 2015, five trials were published that demonstrated the superiority of ET in addition to intravenous thrombolysis (IVT) for eligible patients with confirmed LVO of the anterior circulation who present within 6 to 8 hours of symptom onset.2–7 More recently, the window of benefit for select patients with LVO was extended to 24 hours.8 9 ET now has the greatest magnitude of effect of all acute stroke therapies.

Before the advent of the ‘endovascular era,’ defined as the time from February 2015 onward, the minority of AIS patients in the United States received acute treatment even with reasonable access to treating facilities. One large study of Medicare patients demonstrated that 4% of AIS patients received IVT and 0.5% ET in 2011. At the time of this study, it was estimated that 56% of the United States' population had access to endovascular-capable hospitals within 60 min by ground transportation.10 By some estimates, as many as 20 000 of the 650 000 AIS patients in the United States annually may be eligible for ET based on the criteria from the DAWN and DEFUSE-3 trials whose results form the basis for the 2018 American Heart Association/American Stroke Association (AHA/ASA) extended window guidelines.11–13 This number may be even greater as we attempt to identify patients not included in the original randomized trials, including those with more distal LVO, who may benefit from ET.14 As the number of potentially eligible patients expands, the burden of meeting the demand for treatment likely continues to increase.15

The stroke community has called for rapid development of systems of care to meet the mandate for equal access to interventional therapy throughout the country. With an ever-expanding population of patients potentially eligible for treatment with highly-specialized and resource-intensive ET, it is not clear if acute stroke treatment rates have improved. More national data is needed to inform the continued development of stroke systems of care.12 15 16 We performed a study using nationwide data on AIS hospitalizations, comparing characteristics and outcomes of AIS patients treated with ET before and during the endovascular era. We believe a better understanding of current utilization can help inform continued expansion of ET-related resources nationally, such that all potentially eligible patients can have access to this revolutionary new standard of care.

## Methods

We utilized the Nationwide Readmissions Database (NRD) of the Healthcare Cost and Utilization Project (HCUP), including the years 2013–2016. The NRD contains nationwide data on approximately half of US admissions for all payers and the uninsured, excluding rehabilitation and long-term acute-care hospitalizations but with nationally representative sampling for all other acute hospital and geographic types. Each individual in the NRD has an anonymized, verified linkage identifier that allows for analysis of readmissions. A variable is also provided that allows calculation of days between admissions for an individual. Due to the structure of the NRD, it is not possible to track individuals or readmissions across years, but only within a single calendar year. The IRB approved this project and waived the need for human consent. All analyses comply with the HCUP data use agreement.

For hospitalizations from January 2013 through September 2015, we identified index acute ischemic stroke (AIS) admissions using International Classification of Disease, Ninth Revision, Clinical Modification (ICD-9) codes previously validated in the literature (433.x1, 434.x1, 436). Utilization of these validated ICD-9 codes for stroke as the primary discharge diagnosis has a sensitivity of 74%, specificity of 95%, and positive predictive value (PPV) of 88%.17 18 We used standard ICD-9 codes to identify the following vascular risk factors: diabetes, hypertension, hyperlipidemia, atrial fibrillation/flutter, and tobacco use. We used the ICD-9 procedure code 99.1 to identify IVT and 39.74 to identify ET. The coding of medical conditions and procedures was changed nationally from the ICD-9 to ICD-10 system in October 2015. Hence, for hospitalizations from October 2015 through December 2016, we identified index AIS admissions using the previously validated ICD-10 codes of I63.x in the primary diagnosis position, which has a PPV of ≥82%0.19 We used ICD-10 procedure codes to identify IVT (3E03317 and 3E04317) and ET (03CG3ZZ, 03CH3ZZ, 03CJ3ZZ, 03CK3ZZ, 03CL3ZZ, 03CP3ZZ, 03CQ3ZZ). To translate ICD-10 codes to ICD-9 codes for comorbidities, we used the General Equivalence Mapping database for index admissions from October 2015 through December 2016, which provides almost 100% concordance for cardiovascular conditions.20

We described characteristics of the index hospitalization as defined by HCUP, including hospital bed size (small, medium, or large), teaching hospital status (metropolitan non-teaching, metropolitan teaching, and non-metropolitan hospital), income quartile of patient’s zip code, and National Center for Health Statistics urban-rural location classification (‘central’ counties of metro areas of>=1 million population, ‘fringe’ counties of metro areas of>=1 million population, counties in metro areas of 250 000–999 999 population, counties in metro areas of 50 000–249 999 population, micropolitan counties (population 10 000–49  999), and not metropolitan or micropolitan counties). Patients in the NRD are categorized into All Patient Refined Diagnosis Related Groups (APR-DRGs) using 3M Health Information Systems software, which classifies patients into 25 major diagnostic categories. Patients are further subdivided into four severity of illness subclasses according to degree of loss of function (minor, moderate, major, extreme) and four risk of mortality subclasses (minor, moderate, major, extreme). These two subclasses are calculated independently and may differ from one another. The APR-DRG has been adapted to the VA health system, and analysis of the APR-DRG has shown a correlation between mortality rates and increasing APR-DRG risk of mortality scores.21 22 Additionally, HCUP characterizes discharge disposition by one the following descriptors: routine; transfer to short-term hospital; transfer to skilled nursing facility, intermediate care facility, or other facility; home healthcare; against medical advice;  died, or discharged alive, destination unknown.

The main predictor of outcome was the time period, or ‘era’ of index admission. We defined the ‘pre-endovascular era’ as including admissions from January 2013 through January 2015, and the ‘endovascular era’ as including admissions from February 2015 through December 2016.

### Statistical analysis

We first calculated baseline characteristics of index AIS admissions, stratified by era (pre-endovascular era versus endovascular era). Using population weights provided by HCUP, we calculated means and SD for continuous variables and frequencies and percentages for categorical variables to arrive at nationally representative estimates. Next, we calculated baseline characteristics of AIS admissions in which ET was performed, stratified by era.

We then calculated the proportion of all AIS admissions in which, first,  IVT and second, ET was performed, per month, from January 2013 through December 2016, and depicted the proportions graphically.

Among AIS patients treated with ET, we then examined the association between the primary predictor of era (endovascular era versus pre-endovascular era as the referent group) and three main outcomes: discharge disposition (defined as discharge home versus other location): in-hospital mortality during the index admission; and 30-day all-cause readmission. We estimated odds ratios (OR) and 95% confidence intervals (CI) using logistic regression, and we ran three models for each outcome: unadjusted, adjusted for APRDRG Risk of Mortality measures, and adjusted for APRDRG Severity measures. Analyses were performed in SAS version 9.4.

### Data availability

Because HCUP data is publicly available, we will not provide the data, analytic methods, and study materials to other researchers for purposes of reproducing the results or replicating the procedure. The authors are further limited to release data by the data use agreement.

## Results

Table 1 summarizes demographics, comorbidities, and hospitalization characteristics of index admissions for acute ischemic stroke (AIS), stratified by era (before or during the endovascular era). Using weights to arrive at nationwide estimates, there were 925 363 index AIS admissions before the endovascular era and 857 347 during. Demographics, medical comorbidities, discharge disposition, and hospitalization characteristics were similar for patients admitted with AIS in the two eras, except that a higher proportion received IVT (8.4% vs 7.8%) and ET (2.6% vs 1.3%) in the endovascular era compared with the pre-endovascular era. Also, although LOS was shorter in the endovascular era (5.70 vs 6.80 days), total charges of hospitalization were greater ($56 691 versus$53 878), and admissions were more often to a metropolitan hospital (65.2% vs 57.2%).

Table 1

Baseline characteristics of index ischemic stroke admission, stratified by pre- or post-endovascular era

Next, we examined characteristics of AIS patients treated with ET, comparing the two eras (table 2). A smaller proportion of those treated with ET in the endovascular era received IVT (29.7% vs 44.9%), and LOS was substantively shorter (9.75 vs 12.76 days), although the distribution of discharge disposition locations was similar. Characteristics of treating hospitals were similar in the two eras, including a predominance of endovascular-treated AIS patients at large hospitals (>80%), in large metropolitan areas (almost 70%), at metropolitan teaching hospitals (>85%).

Table 2

Baseline characteristics of index ischemic stroke admissions treated with endovascular therapy, stratified by pre- or post-endovascular era

We examined time trends in the proportion of all AIS index admissions treated with IVT and ET, over the 2 years examined in the pre-endovascular era, and in the first 2 years of the endovascular era (figure 1). As expected, there was an increase in the proportion treated with ET after January 2015, from 1.09% in March 2013, to 2.27% in March 2015, to 3.35% in December 2016. There was a slight increase in the proportion treated with IVT, from 7.27% in March 2013, to 8.47% in March 2015, to 9.33% in December 2016. While there were similar transient increases in IVT utilization at the end of 2014 and 2015,>8% of all AIS patients were treated with IVT in 2016 and >9% for the last 3 months of the year.

Figure 1

Trends in prevalence of intravenous thrombolysis and endovascular therapy before and during the endovascular era.

When we examined outcomes of admissions for AIS treated with ET, comparing the two eras (table 3), we found that there were reduced odds of discharge home in the endovascular era (adjusted OR 0.85) compared with the pre-endovascular era. There was no significant association between era and in-hospital mortality, but there was a trend for a positive association between treatment in the endovascular era and 30-day all-cause readmission.

Table 3

Outcomes after endovascular therapy, comparing the pre-endovascular era to the endovascular era

## Discussion

This study describes the progressive increase in acute treatment of ischemic stroke with the advent of the ‘endovascular era’ but also highlights the persistent underutilization. Not only has treatment with ET increased after the demonstration in February 2015 of the efficacy of ET (from 1.09% of all AIS admissions in March 2013, to 2.27% in March 2015, to 3.35% in December 2016), but treatment with IVT has also increased (from 7.27%, to 8.47%, to 9.33% over the same time period). Even so, the minority of AIS patients are treated with revascularization therapy, and such treatment occurs predominantly at large metropolitan teaching hospitals. At the same time that acute treatment rates have increased, cost has increased and length of stay has decreased. This study is novel in its use of nationally representative data that is not reliant on quality registry participation, which allowed us to estimate the state of real-world utilization in the US of the latest treatment for AIS. Additionally, it sheds light on the impact of ET on metrics for which hospitals are increasingly held accountable, including cost and length of stay. Our findings highlight that despite the significant increase, there is a pressing need for continued expansion and improvement in our AIS systems of care so that we can treat even more eligible patients with ET throughout the country. Given the persistently low treatment rates and predominance at large metropolitan teaching hospitals, it calls into question whether hospitals and proceduralists will be able to meet the lofty volume requirements established by the Joint Commission for certification as comprehensive and thrombectomy capable stroke center.23 Additionally, one must question whether access to ET should be expanded in the same ways throughout the country.

The implications of AIS treatment in the endovascular era were recently assessed in Get With the Guidelines (GWTG) stroke patients. A study of ET from April 2003 to June 2016 demonstrated a significant increase in the utilization of ET after January 2015 despite a gradual increase in the preceding months. The overall rate of increase went from 0.13% to 1.33% per year, with a higher increase of 2.12% in ET-capable hospitals. In the third quarter of 2016, 3.3% of all patients at all GWTG hospitals and 7.5% at GWTG ET-capable hospitals received ET, estimated to comprise 15.1% and 27.3% of all eligible patients.24 We, too, found a significant increase in ET, coinciding with the timing of published trials. We also saw a progressive increase sustained through December 2016, such that 3.3% of all AIS were treated with ET in our study as well. While our study does not include data on treating state, the predominance of ET at urban, metropolitan teaching hospitals suggests wide variation in access throughout the country. Less than half of all US stroke admissions are represented in the GWTG registry, and large, urban, teaching hospitals tend to have disproportionately greater representation.25 Our study adds data that is more representative of trends throughout the entire country by including data from hospitals that chose not to participate in GWTG as well as data on concurrent treatment with IVT, in-hospital mortality, discharge disposition, readmissions, cost, and length of stay.

A smaller study of over 13 000 patients in the state hospital association database also utilized ICD-9 and ICD-10 codes for AIS to examine the change in utilization of ET by comparing treatment rates in 2014 and 2015. This study demonstrated a similarly progressive increase in ET beginning in January 2015 and found slightly higher rates of treatment with ET of 4% in the fourth quarter of 2015.26 Unlike the GWTG study and similar to ours, this study assessed rates of utilization of IVT. In the last quarter of 2015, 5.7% of all AIS patients received IVT, but rates of utilization did not change significantly throughout the study. In contrast, we found a slight increase in the percentage of patients treated with IVT, from 7.27% in March 2013 to 9.33% in December 2016. Differences may be due to our larger, nationally representative sample. Furthermore, we suspect that optimization of systems of care, as well as increased public education about acute stroke therapies, resulted in the slight increase in IVT in the endovascular era. This study also assessed cost, length of stay, discharge disposition, and mortality. Similar to ours, they found an increase in hospital charges. Overall hospital charges increased from 2014 to 2015, and treatment with IVT and ET led to a two- and four-fold increase in charges respectively. Median length of stay did not change from 2014 to 2015 and was 5 days for patients who received ET and 3 days for patients who received IVT. There was no change in discharge home or mortality between 2014 and 2015.26 We, too, found an increase in overall charges in the endovascular era. However, at the same time, we found that overall LOS was shorter in the endovascular area. Like these authors, we did not find a difference in discharge home or mortality in the endovascular era, which is perhaps not surprising since the benefit for ET was demonstrated at 90 days and the greatest improvements were seen in functional status rather than mortality with the positive endovascular trials.

Despite the increase in ET utilization during the endovascular era, AIS patients treated with ET tend to be treated at large teaching hospitals in large metropolitan areas. This demonstrates unequal access to ET throughout the country, and greater efforts must be made to understand and address geographic treatment disparities. It is notable that there was minimal change in cost for patients treated with ET in the two eras but increase in cost of almost \$3000 among all AIS patients. It is likely most of the increase in the cost for AIS patients during the endovascular era is unrelated to ET or stroke care. We suspect that some of the change in LOS during the endovascular era is due to the evolving healthcare system and greater accountability for metrics such as LOS. As more patients are treated with ET during the endovascular era, there will likely be greater comfort in caring for these complicated patients and more streamlined efforts at disposition with higher treatment volumes. We acknowledge that at the same time LOS decreased, odds of discharge home were lower during the endovascular era. While endovascular techniques and tools have improved and resulted in better recanalization in randomized clinical trials, as centers treat more patients with ET, selection criteria are likely not as strict, and patients with poorer baseline function than in the randomized clinical trials are being offered this life-saving therapy. Additionally, it is not clear to what extent factors such as proceduralist training, experience, and case volume impact metrics such as discharge disposition.

Interestingly, our study demonstrated lower rates of IVT in patients treated with ET in the endovascular era. This may reflect presentation and treatment of AIS patients eligible for ET>4.5 hours from last known well. Alternatively, it is possible that fewer patients are receiving IVT before ET because of a perceived lack of benefit of IVT with LVO, with the low likelihood of LVO clot lysis with IVT shown in the ET trials.3–5 13 Such choices would be counter to current AHA/ASA guidelines, and further national studies are needed to explore the reasons for this trend.

Our study likely reflects the real-world state of ET in the endovascular era, but we acknowledge several limitations. We have used ICD-9 and ICD-10 code clusters previously validated in the literature, and translated codes using the General Equivalence Mapping database, but there may be misclassification of diagnoses based on ICD-9 and 10 codes. The NRD does not provide details on time metrics or imaging characteristics to allow us to assess the percentage of eligible patients treated as well as severity of presentation and ultimate mechanism of stroke. Additionally, we were unable to capture the role of transfers in acute stroke care in the endovascular era.

In summary, we have demonstrated that the endovascular era has resulted in a progressive increase in utilization of ET for AIS, with a doubling of the total proportion of AIS patients treated from January 2013 through January 2015 compared with February 2015 through December 2016. At the same time, there has been a slight increase in utilization of IVT. However, the minority of AIS patients are treated with ET, and treatment occurs predominantly at large metropolitan teaching hospitals. Significant work remains to be done to better understand why we are not yet seeing higher treatment rates and the improvement in outcomes such as discharge disposition and in-hospital mortality that we expect to see based on the results of ET randomized clinical trials. Additionally, more work is needed to understand geographic disparities and ensure that patients throughout rural and urban areas of the United States have timely access to health systems capable of providing ET.

## Footnotes

• Contributors LS: Study design, interpretation of data, drafting the work. ST: Critical revision for important intellectual content. JF: Critical revision for important intellectual content. JM: Critical revision for important intellectual content. MD: Study design, analysis and interpretation of the data, critical revision for important intellectual content.

• 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 reports the following disclosures: Research Support: Stryker, Penumbra, Medtronic, Microvention; Consultant/Ownership Interest: Imperative Care, Cerebrotech, Viseon, Endostream, Rebound Therapeutics, Vastrax; Investor/Stockholder/Owner: BlinkTBI, Serenity, NTI, Neurvana, Cardinal Consulting.

• Provenance and peer review Not commissioned; externally peer reviewed.

• Data sharing statement Because HCUP data is publicly available, we will not provide the data, analytic methods, and study materials to other researchers for purposes of reproducing the results or replicating the procedure; the authors are further limited to release data by the data use agreement.

• Patient consent for publication Not required.

## Request Permissions

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.