Bringing doctors together

or

A free, private network for verified physicians

Leading voices

Dr. Alpesh Amin

Hospital Medicine · UCI

View

Dr. Akua Ampadu

Hospital Medicine · NYP Weill Cornell

View

Dr. Reza Manesh

Hospital Medicine · Greater Los Angeles VA

View

Dr. Michael Calavano

Hospital Medicine · NYP Weill Cornell

View

Dr. Nasim Afsar

Hospital Medicine

View

Dr. Sara Murray

Hospital Medicine · Johns Hopkins Medicine

View

Dr. Robert Korom

Hospital Medicine · Penda Health

View

Dr. Sunil Kripalani

Hospital Medicine · VUMC

View

Dr. Ronald Greeno

Hospital Medicine · Society of Hospital Medicine

View

Dr. Moises Auron

Hospital Medicine · Cleveland Clinic

View
  • Discover your community

  • Stay current

  • A profile that represents you

  • Find your people

About RoonPrivacyTerms

    Overview

    Association of Mobile Stroke Unit Care and Spending, Utilization, and Death in New York City.

    Jan 16, 2025DOI: 10.1161/jaha.124.036784Journal ArticleAHA

    Written by: Dhruv KhullarBabak B NaviSaad MirMatthew E FinkLuisa TinapayGlenn AsaedaRainu KaushalEshani PareekAmelia M Bond

    Find these authors on Roon

    DK
    Dr. Dhruv Khullar
    Hospital Medicine · WCM

    Methods

    This study was approved by the Biomedical Research Alliance of New York, and informed consent was waived because only historical data with minimal risk of harm were used. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology reporting guidelines.13We cannot share our data due to data use agreements. Traditional Medicare claims data can be obtained through a Data Use Agreement from the Centers for Medicare and Medicaid Services. New York City MSU Registry data and electronic health records from 2 large hospitals contains personal health information and cannot be shared publicly.The NewYork‐Presbyterian MSU was launched in 2016, in collaboration with Weill Cornell Medicine, Columbia University Irving Medical Center, and the Fire Department of New York, making it the first such emergency vehicle on the East Coast. It is dispatched through New York City's 911 system from 9:00 am to 5:00 pm on Mondays through Fridays and is intended to offer timely care to patients with stroke through several notable features, including a portable computed tomography scanner, medications such as tPA, trained paramedics and nurses, and in‐person or telemedicine consultation with a stroke neurologist.The main data sources included the New York City MSU registry, 100% 2016 to 2020 Medicare fee‐for‐service claims, and the 2019 5‐year American Community Survey, as well as electronic medical records from 2 NewYork‐Presbyterian Manhattan hospitals through the Tripartite Request Assessment Committee network.14 Using Medicare claims, we identified all admissions between October 2016 and December 2019 for patients aged ≥65 years occurring in a hospital serviced by a New York MSU with an admitting or primary discharge diagnosis of acute stroke or a relevant stroke mimic. To identify our case population, we next used the MSU registry to identify all patients transported via the New York MSU, which serves the boroughs of Manhattan, Queens, and Brooklyn, for suspected acute stroke between October 2016 and December 2019. We then linked the MSU transports of patients aged ≥65 years to the set of acute stroke or stroke mimic inpatient Medicare claims using age, sex, date of transport, and admitting hospital with a multistep approach (Data S1). Further, we used the American Community Survey to link patients' zip code–level demographic information to Medicare claims. Finally, we used Tripartite Request Assessment Committee data, which contained data for a secondary clinical outcome, to link patients with a hospitalization at 2 NewYork‐Presbyterian Manhattan hospitals between October 2016 and December 2019 to hospitalizations in Medicare claims using unique patient identifiers, including Social Security number, health insurance claim number, or Medicare Beneficiary Identifier.
    Admissions with a length of stay <1 day were excluded because such patients are unlikely to have had a clinically meaningful stroke. For patients with multiple admissions during the time frame, only the index admission was included to avoid having a single patient counted as an observation multiple times or for certain events to be counted multiple times.
    Our primary outcomes were short‐term hospital outcomes as well as longer‐term health, usage, and spending outcomes. Short‐term hospital outcomes included the length of stay of the index hospitalization and whether a patient was discharged home. Longer‐term usage measures included whether a patient had any inpatient or emergency department visits within 1 year of discharge. We also examined total spending within 1 year of discharge as well as total spending by broad categories including inpatient, outpatient, post–acute care, and hospice. Finally, we measured 1 long‐term health outcome: death within 1 year after discharge.
    The secondary outcome available for a subset of patients admitted to 2 Manhattan hospitals was a modified Rankin Scale (mRS) in electronic medical records collected at the time of discharge from an index admission. The mRS is a commonly used score that measures the degree of disability or dependence people experience after a stroke or other neurological event.15 It grades patients' function on a scale from 0 (indicating no symptoms) to 6 (indicating death).
    Patient demographic information included age, race, and sex. Two residential demographic variables were added from the American Community Survey on the basis of a patient's zip code: proportion with high school diploma and median household income.
    Clinical characteristics before hospitalization included residence at a care facility and Hierarchical Condition Category risk score as well as indicators for whether a patient had atrial fibrillation, diabetes, or hypertension. Residence at a care facility immediately before an index admission was constructed using MDS Nursing Home Assessment data and identified on the basis of an assessment within 90 days before stroke hospitalization with no discharge to another facility. The Hierarchical Condition Category risk score was constructed using Medicare claims data the year before a patient's index admission. Indicators for chronic conditions were obtained through the Chronic Conditions segment of the Master Beneficiary Summary File the year before a patient's index admission. Based on the admitting or primary discharge diagnosis from the index hospitalization claim, hospitalizations were categorized into stroke types as ischemic stroke/transient ischemic attack, intracerebral hemorrhage, subarachnoid hemorrhage, or other diagnosis, which included stroke mimics. For the MSU group, the working diagnosis of the treating attending neurologist at the time of MSU transport is also provided.
    For a sensitivity analysis, we constructed 3 measures of a patient's medication use up to 1 year before admission using Medicare Part D claims. Measures included any use of statins, anticoagulants, and insulin. For a secondary outcome (mRS) analysis, patients' National Institutes of Health Stroke Scale was also collected from electronic medical records at the time of admission during the index hospitalization.
    Data on intravenous tPA use were collected. For patients transported via an MSU ambulance, tPA administration was documented in the registry. For patients not transported via an MSU ambulance, tPA administration was determined by emergency department claims data in the MedPAR or outpatient file, using codes identified in previous literature.16, 17
    MSU and control admissions were propensity score matched to account for potentially confounding variables. The propensity score was constructed using logit regressions, where MSU treatment was the dependent variable. Propensity score–matched controls were identified on a 1:20 basis using Mahalanobis distance matching. The 1:20 match rate was based on a prespecified power calculation indicating 80% power to detect a difference in proportions for discharge to home of ≈14%. The study was powered on the basis of discharge disposition because it is a commonly used proxy for functional outcomes.18 Controls were matched by patient age, sex, Hierarchical Condition Category risk score, chronic condition indicators, and 2 residential demographic variables, as well as year and quarter of admission. Due to the sample size of patients with mRS, the match rate was 1:5 for this secondary outcome. All controls in the main analysis were used in the secondary analysis with the additional National Institutes of Health Stroke Scale adjustment. Admissions were separately matched for admitting hospital, stroke type, and whether a patient was transported from a care facility and then combined for analysis. We tested the balance of the sample by comparing covariates across treatment and control admissions using t tests or tests of proportion. We also compared outcomes across treatment and control admissions using t tests or tests of proportion. Administration of tPA was not used for matching because it is thought to be a key mechanism through which outcomes may improve for MSU versus non‐MSU patients.
    As tests of sensitivity, we performed 3 more analyses to ensure robustness of our results to matching criteria and patient identification strategies. First, we limited the set of MSU treatment admissions to those that satisfied a stricter set of match criteria between the MSU registry and Medicare claims (Data S1). Second, focusing only on patients with Part D, we propensity score matched using 3 additional controls that measured a patient's medication use before admission. Third, we added race as a covariate in the propensity for matching.
    All statistical analyses were performed using Stata version 17.0 (StataCorp, College Station, TX), and 2‐sided P<0.05 was considered statistically significant.

    Results

    Between October 2016 and December 2019, 294 patients aged ≥65 years were transported by a New York City MSU, including 167 Medicare patients who met the study's inclusion criteria (Figure). The mean age was 79.9 years, 56.3% of patients were women, and 61.1% had a stroke type of ischemic stroke/transient ischemic attack. MSU patients and propensity score–matched control patients (N=2518) were similar on a broad range of clinical and demographic characteristics, including age, sex, comorbidities, and type of stroke (Table 1). MSU patients were significantly more likely to receive tPA relative to matched patients (49.9% versus 9.4%; difference, 37.5 percentage points [95% CI, 30.2–45.6]; P<0.001). For the MSU patients, the working diagnoses of the attending neurologist at the time of MSU transport were ischemic stroke (n=130 [77.8%]), acute encephalopathy (n=17 [10.1%]), intracerebral hemorrhage (n=15 [9.0%]), transient ischemic attack (n=2 [0.6%]), subarachnoid hemorrhage (n=1 [0.3%]), seizure (n=1 [0.03%]), and recrudescence of old stroke (n=1 [0.03%]).
    LOS indicates length of stay; mRs, modified Rankin Scale; and MSU, mobile stroke unit.
    Demographic and Clinical Characteristics by MSU Treatment
    MSU indicates mobile stroke unit; and tPA, tissue‐type plasminogen activator.
    Comparisons reflect mean‐comparison (t) test or 2‐group test of proportions between indicated subgroups.
    Values suppressed due to low sample size (n<11).
    Presence of chronic conditions were obtained using the Chronic Conditions segment of the Master Beneficiary Summary File the year before a patient's index admission.
    Hierarchical Condition Category risk score constructed using Medicare claims data the year before a patient's index admission.
    Neighborhood demographic variables added from American Community Survey on the basis of a patient's zip code.
    Table 2 presents outcomes during the index hospitalization. Compared with patients transported by traditional ambulances, MSU patients had similar lengths of stay (5.9 versus 6.7 days; difference, −0.8 days [95% CI, −1.9 to 0.2]; P=0.13) and were discharged to skilled nursing facilities at similar rates (15.6% versus 15.1%; difference, 0.5 percentage points [95% CI, −5.2 to 6.2]; P=0.86).
    Clinical, Usage, and Spending Outcomes by MSU Treatment
    MSU indicates mobile stroke unit.
    Comparisons reflect mean‐comparison (t) test or 2‐group test of proportions between indicated subgroups.
    Spending does not include cost of index admission.
    Table 2 also presents mortality, utilization, and spending outcomes at 1 year after hospital discharge. Compared with patients transported by traditional ambulances, MSU patients had numerically lower mortality rates, although the difference did not reach statistical significance (21.6% versus 28.4%; difference, −6.8 percentage points [95% CI, −13.3 to 0.3]; P=0.058). They were similarly likely to experience a repeat hospitalization within 1 year of discharge (24.0% versus 23.2%; difference, 0.8 percentage points [95% CI, −5.9 to 7.4]; P=0.82), and had similar likelihood of emergency department visits without inpatient admission (13.8% versus 11.6%; difference, 2.1 percentage points [95% CI, −3.2 to 7.5]; P=0.41). Overall spending at 1‐year after discharge was also similar between MSU and control patients ($33 526 versus $32 191; difference, $1335 [95% CI, −6527 to 9197]; P=0.74), and there were no significant differences by type of spending (inpatient, outpatient, postacute, or hospice).
    Results were broadly similar in sensitivity analyses. However, in analysis with a more conservative set of treatment observations (Data S1 and Table S1), MSU patients had significantly lower lengths of stay during the index hospitalization, and the mortality difference, while large, was not statistically significant (Table S2). In an analysis of patients with Medicare Part D and controlling for prior medication use, MSU patients had significantly more outpatient spending, and the mortality difference was also large but not statistically significant (Tables S3 and S4). Finally, in analysis using race as an additional demographic covariate, the mortality difference between MSU and control patients was large but not statistically significant (Tables S5 and S6).
    In a secondary analysis examining patients for whom mRS scores were available at time of discharge, we found that MSU patients had similar mRS scores relative to non‐MSU patients (2.20 versus 2.48; difference, −0.28 [95% CI, −1.22 to 0.66]; P=0.56; Tables S7 and S8).

    Discussion

    In this retrospective, observational study of patients presenting with suspected acute stroke who were transported by MSUs versus traditional transport in New York City, MSU care was associated with a trend toward a lower mortality rate at 1 year, although this finding did not reach statistical significance, possibly owing to the sample size. MSU patients experienced shorter lengths of stay during the index hospitalization, although this finding did not reach statistical significance. Other outcomes, including usage of inpatient and emergency department care, as well as inpatient, outpatient, postacute, and hospice spending at 1 year, did not differ between the 2 groups. These results suggest that MSU care may result in meaningful reductions in the 1‐year mortality rate among Medicare beneficiaries presenting with acute stroke symptoms, without increased health care usage and costs, although more research is needed to confirm these findings. It is possible that more frequent and timely administration of thrombolytics, optimized blood pressure management, or improved prehospital triage to appropriate stroke centers could contribute to better short‐ and long‐term outcomes for patients experiencing acute stroke.19
    Our study adds to the existing literature in several ways. First, it examines longer‐term health, usage, and spending outcomes for patients who received MSU care; most prior work focuses on outcomes during the index hospitalization and functional outcomes at 3 months. For example, a recent study found that the mortality rate was somewhat lower at 90 days for MSU patients versus those transported by traditional emergency medical services4; other studies have found no significant difference in the 7‐day mortality rate.7, 20 Second, this study is, to our knowledge, the first to link MSU registry data to claims data from the Medicare program, which covers >65 million Americans.21 Finally, the study expands the evidence base for MSU to a major metropolitan area, New York City, which is the largest and densest city in the United States. Ambulance time and stroke care are known to differ across geographic areas, and it is important to understand how the clinical and financial impact of MSUs may vary accordingly.
    Establishing an MSU involves significant up‐front capital expenditures. Many hospitals have small operating margins and may hesitate to invest in the creation and maintenance of MSUs.22 Meanwhile, many payers do not currently reimburse for MSU services, and may resist doing so, without a clearer understanding of long‐term health and financial outcomes.23 To our knowledge, no research has examined 1‐year mortality differences for patients transported by MSUs versus traditional ambulances. With regard to spending, a recent study found that MSU care in Berlin was associated with higher costs relative to traditional care at 3 months24; another analysis of patients mostly in the Houston area that has not yet been peer reviewed, reported similar costs across groups at 12 months.25 Both studies found that MSU care met commonly used cost‐effectiveness thresholds. The findings of our study may reassure health care leaders and policymakers concerned about the possibility of increased long‐term costs or usage associated with MSU programs.
    Our study has limitations. First, this research examined outcomes for Medicare beneficiaries in a dense urban environment and may not be generalizable to other populations, such as younger patients and patients in rural areas. Second, we were not able to directly examine the care that MSU and control patients received during hospitalization, such as acute blood control and antithrombotic and statin medications. Third, patients in the MSU registry were linked to Medicare claims through a series of steps, and it is possible that some patients were misidentified; however, a sensitivity analysis with different matching criteria produced broadly similar results. Fourth, although MSU and control patients were well balanced on various clinical and demographic characteristics, our results are subject to residual confounding. For instance, because of our reliance on Medicare claims data for our control group, we were unable to adjust for patients' last known well time, which affects the odds of acute reperfusion therapies and therefore could have impacted clinical outcomes. However, as the MSU generally transports only patients deemed eligible for acute reperfusion therapies, and more severe strokes tend to present earlier by emergency medical services transport, we would expect such a selection bias to favor worse outcomes in the MSU group. Furthermore, because the study was observational, we cannot make causal claims about the impact of MSU care on patient outcomes. Additionally, our long‐term outcomes were assessed at 1 year after patients' index event, and this time lapse could have mitigated any potential benefits of MSUs.
    Nonetheless, this study presents novel evidence on long‐term outcomes for patients transported by MSU versus traditional ambulances in the largest city in the United States. We found that MSU care was associated with a trend toward lower 1‐year mortality rates for Medicare patients suspected to have acute stroke at presentation. Further research is needed to confirm these findings in other patient populations and regions of the country.

    Sources of Funding

    This work was supported by NewYork‐Presbyterian. NewYork‐Presbyterian had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

    Disclosures

    Dr Khullar reported receiving grants from the Agency for Healthcare Research and Quality, the Physicians Foundation, and the National Institutes of Health, outside the submitted work. Dr Navi reported receiving consulting fees for serving on an adjudication committee for MindRhythm Inc. and receiving grants from the Patient‐Centered Outcomes Research Institute for serving as site principal investigator for the BEST‐MSU (Benefits of Stroke Treatment Delivered Using a Mobile Stroke Unit) trial. Dr Mir reported receiving grants from the National Institute of Neurological Disorders and Stroke and the National Heart, Lung, and Blood Institute, outside this work. Dr Fink reported receiving grants from the National Institute of Neurological Disorders and Stroke and providing professional services to Relias Media LLC, outside this work. Dr Kaushal reported receiving grants from the Patient‐Centered Outcomes Research Institute, the Task Force for Global Health, Inc., Novartis, and the National Institute of Mental Health and serving on the Advisory Board of and receiving stock from Curai Health, outside this work. Dr Bond reported receiving grants from the Agency for Healthcare Research and Quality, Commonwealth Fund, and Defense Health Agency, as well as a speaking honorarium from Brown University, outside this work. The remaining authors have no disclosures to report.

    Supporting information

    Data S1
    DKDr. Dhruv Khullar

    Dr. Dhruv Khullar

    Internal Medicine · Hospital Medicine
    Weill Cornell Medicine
    dhruv