Optimizing Prehospital Stroke Systems of Care-Reacting to Changing Paradigms (OPUS-REACH): a pragmatic registry of large vessel occlusion stroke patients to create evidence-based stroke systems of care and eliminate disparities in access to stroke care

The Optimizing Prehospital Stroke Systems of Care-Reacting to Changing Paradigms (OPUS-REACH) registry will include information from nine health systems across the Northeastern United States. The registry includes data from EMS dispatch to 90 day outcome.

The OPUS-REACH consortium consists of nine health systems with the shared goal of studying and optimizing prehospital systems of care for LVO stroke patients (Table 1). These nine health systems consist of 46 acute care hospitals including 10 ESCs and span systems span urban, suburban, and rural areas (Figs. 2 and 3). The nine health systems account for over 1.7 million ED visits, over 10,000 stroke admissions, and approximately 850 mechanical thrombectomies annually (Supplement 1). The lead hospital in the OPUS-REACH network, Temple University, is an innovator in using Geographical Information Systems (GIS) to assess systems of healthcare delivery.

Our LVO stroke registry will include all patients from the nine health systems who have undergone EVT for LVO stroke between 2015 and 2020. Patients will be excluded if they were not transported to either the ESC or non-ESC by EMS or the patient does not have a 90 day functional outcome assessed.

Each spoke in the OPUS-REACH consortium has a site investigator (SI) responsible for oversight of data collection. These SI are primarily emergency physicians, but also include neurologists. Data collection for the OPUS-REACH registry will be extracted at each site individually and then submitted to a central registry housed at Temple University [18]. The REDCAP database is approved for the storage of protected health information. Only study personnel approved by the Institutional Review Board and designated by the principal investigator (DI) will have access to protected health information. All investigators listed on this manuscript will have access to de-identified data for the purpose of writing manuscripts for publication. Authorship will be determined based on the contributions of each investigator to the manuscript. Data and results will be disseminated via peer-reviewed journals in the fields of neurology, emergency medicine, and public health.

The main outcome variable is functional status at 90 days as measured by the Modified Rankin Scale (mRS) [19‐22]. Scores of 0 to 2 are considered good outcomes while scores of 3 to 6 are considered poor outcomes. In the cases where a 90-day mRS may not be available, the SI from each site will review the electronic medical record for clinic notes, emergency department visits, or physical therapy notes, that would describe the patient’s functional outcome at 90 days. They will then assign a score of “good” or “poor functional outcome.” This approach has been documented as an accurate way to estimate a patient’s functional outcome [22].

We will compare outcomes between patients transferred for EVT to those directly transported by EMS to an EVT center. We will stratify patients by eligibility for IVT as well as onset of stroke. We will also stratify patients by drive times, based on modeling, to the EVT center.

Hospitals will identify LVO stroke patients treated with EVT at their facilities. For each patient who underwent an EVT, the SI will complete a standardized case report form (CRF). This CRF will include prehospital data, in-hospital data, and 90 day functional outcomes.

A prior study showed an 8% absolute difference in outcomes for LVO stroke patients between patients who were transported directly to an ESC versus those who were transferred from a primary stroke center: 40% in the direct transport group v. 32% in the transfer group [21]. Therefore, to detect an 8% difference between groups, with a 3:1 patient ratio of primary to secondary transports, 95% confidence and 80% power, we will need approximately 1600 patients.

The use of GIS and spatial analysis moves prehospital system design from expert opinion to a data-driven process. Geocoding patient data (e.g., pick up locations), integrating it into a GIS, and linking it with other clinical data will provide the necessary platform to examine the complex movements of patients in time and space. We will map the paths to care for LVOs stroke patients and calculate times based on time of day, day of week, and traffic patterns.

To examine geographic disparities in stroke care there are two methods we will apply. First, we will use spatial scan statistics. These methods generate circles (or ellipses) of various sizes and evaluate observed versus expected rate ratios (risk within circle compared to risk outside) to identify statistically significant “clusters” of outcomes including clustering over time. We will use this approach to find places in the study area where patients are statistically significantly more likely to not get transported directly to an ESC. Spatial scan statics will be estimated using the free software SaTScan. The second method we will use is a structured additive regression model based on a fully Bayesian approach via Markov Chain Monte Carlo (MCMC) simulations. These models can account for spatial and temporal random effects, spatial effects, and spatial-temporal interactions. These models provide geographically smoothed spatial parameter estimates that can be used to visualize disparities in stroke outcomes (e.g., places with a higher odd of not getting transported directly to an ESC) and provide specific location where outcomes are statistically significantly higher or lower (e.g., credible intervals in Bayesian methods) than the study areas average values.

When comparing the outcomes of LVO stroke patients who underwent primary versus secondary outcomes, we will use univariate analysis of the both the primary group and the subgroups. Groups will be compared by X² tests We will then perform multivariate analysis to control for individual stroke centers and initial NIHSS to verify our results. In addition, we will perform a regression analysis to determine the relationship of time to EVT to functional outcomes.

To look at the effect of socioeconomic factors on LVO stroke outcomes, we will use the United States Centers for Disease Control and Prevention (CDC) Social Vulnerability Index (SVI) [23, 24] (Fig. 4). SVI components include socioeconomic status, minority status, and percentage of population over 65 years of age. Because the SVI is available at the level of the census tract, we will plot the SVI components against drive time to ESCs and determine the differences between populations and their access to care.

OPUS-REACH is support by a grant from the American Heart Association. Temple University will serve as the lead site and the other eight health systems will function as spokes of Temple University. The design of the registry as well as the study protocols were developed by the site investigators at all nine health systems. The study was approved by the Institutional Review Boards (IRB) at each health system. Any alternations in the study protocol will be submitted to each individual IRB for review. As this is a registry study, no trial registration is required.