Mudassir Farooqui
Santiago Ortega-Gutierrez

GEO-SPATIAL MODELING FOR STROKE CARE (GMSC)

Time is of essence in stroke care. There are many systems which facilitate timely stroke therapy and encompass prehospital triage, use of transfer networks and telemedicine, and a tiered system for quick diagnosis. [1-3] The advent of endovascular therapy (EVT) has led to the recognition of more sophisticated and time sensitive models for early recognition of stroke. [4,5] While many regional (rural) hospitals can provide intravenous thrombolysis for stroke, there are very few hospitals with EVT capabilities. In a recent trial, transfer patients were delayed to the tertiary hospital which rendered them ineligible for EVT. [6] This urban-rural disparity in stroke management has a significant impact on the stroke care and its outcomes. [7] While some would argue to bypass the regional (rural) hospitals in favor of comprehensive stroke centers, others have proposed to use the data from the clinical trials to build a model for predicting the value of regionalization in patients with large vessel occlusion (LVO). [8-10] However, to date there is no such data available in the literature.
The overall objective of this proposal is to refine a model of stroke care in the U.S. health system that can be optimized to maximally benefit all stroke patients. Our central hypothesis is that clinical trials data provides the assumptions necessary to estimate the differential effects for the prevalence of LVO, time-effects of tPA on neurologic outcome, and the distribution of times to tPA in U.S. hospitals. These assumptions will form the basis of a Bayesian model to predict optimal EMS destination selection given multiple inputs. Further, a detailed geographic analysis would permit patient-level regionalization recommendations. With 19% of Americans living outside metropolitan areas and triage decisions leaving over 80% of qualifying patients without endovascular therapy, there is a critical need to refine a quantitative approach to prehospital stroke triage.
The rationale for our proposal is that regional care recommendations remain a significant weak link in the stroke chain of survival. The competing interests of timely intravenous therapy and timely endovascular therapy require a population-based approach to maximize overall outcomes. A clinical trial to answer this question is not feasible because of the impracticalities of informed consent, generalizability in geography and health systems, and complex rural EMS systems. Because of that, the best strategy for optimizing stroke outcomes is through formal decision analysis and simulation.

Hypothesis 1;
‘Increasing time to EVT is associated with worse neurological outcomes.’


Hypothesis 2.
‘Longer door-to-needle time is associated with later EVT administration.’

Hypothesis 3. ‘Longer symptom onset-to-hospital arrival time is associated with later EVT administration.’

We will use the VISTA data to estimate effect size and the uncertainty for application in a subsequent Bayesian model. Part of our Bayesian modeling strategy will be using Monte Carlo simulation drawing from a distribution of times. Quantifying and estimating this distribution accurately in U.S. hospitals will be important to predict response to treatment.

ANALYSES PLAN AND VARIABLES:

The study population is all patients with ischemic stroke who have NIHSS and times documented. We will use the following analyses plan:
To develop an algorithm to optimize hospital selection for emergency medical services (EMS) using a detailed geospatial model of the United States and assumptions of the time-varying efficacy of stroke care from clinical trials data.
We will use geographic analysis of all U.S. addresses and hospitals based on stroke therapy availability to calculate transport times for patients with acute stroke. We will develop a Markov decision analysis model to use conditional probabilities to estimate the population-based optimal triage of stroke patients from every location in the continental U.S., accounting for the probability of LVO and good neurologic outcomes contingent upon treatment assignment. We will use clinical trial data, epidemiology (GWTG), and we will incorporate prehospital risk stratification tools to predict LVO. The data provided for this analysis will be based on the distributions and assumptions from this data set and will be built into a Bayesian prediction model.

To simulate the utility of a revised triage system on predicted patient outcomes, using Medicare claims data.
We will use 1 year of Medicare claims data to identify a cohort of stroke patients and measure the health care that they received and their clinical outcomes. We will use Monte Carlo simulation to predict patients with large vessel occlusion, the distribution of times prior to accessing the health care system, delays in inter-hospital transfer. We will compare simulated scenarios adopting the optimal triage algorithm in Aim 1 to the actual care delivered in Medicare beneficiaries to estimate the impact of revised triage on neurologic outcomes, long-term disability, hospital utilization, and cost-effectiveness. We will also simulate the financial impact on health systems and the impact on patients and their families.


Process of Care Variables
• Time to hospital arrival
• Door-to-needle time
• Door-to-thrombectomy time
• Location where stroke was discovered
• Method of transport to hospital


Hospital level (to identify process of care variables that we may include in our modeling if relevant)
• Hospital size
• Region
• Teaching status
• Rural location
• Certified primary stroke center status
• Average number of patients treated with tPA annually
• Average number of patients treated with IA therapy annually
• Average number of annual stroke discharges.


Patient level
• Age
• Sex
• Medical history
• Ambulatory status
• Initial physical examination
• Stroke severity
• Time/day of arrival


Outcome
• Long term outcome (mRS at 3 months)

-

1. Benjamin EJ, Blaha MJ, Chiuve SE, Cushman M, Das SR, Deo R, et al.: Heart Disease and Stroke Statistics-2017 Update: A Report From the American Heart Association. Circulation 135:e146-e603, 2017

2. Prevalence and most common causes of disability among adults--United States, 2005. MMWR Morb Mortal Wkly Rep 58:421-426, 2009

3. Berkhemer OA, Fransen PSS, Beumer D, van den Berg LA, Lingsma HF, Yoo AJ, et al.: A Randomized Trial of Intraarterial Treatment for Acute Ischemic Stroke. New England Journal of Medicine 372:11-20, 2015

4. Alberts MJ, Range J, Spencer W, Cantwell V, Hampel M: Availability of endovascular therapies for cerebrovascular disease at primary stroke centers. Interventional Neuroradiology 23:64-68, 2017

5. Burton TM: A Breakthrough Stroke Treatment Can Save Lives—If It’s Available: Wall Street Journal. New York, Dow Jones & Company, Inc., 2018

6. Froehler MT, Saver JL, Zaidat OO, Jahan R, Aziz-Sultan MA, Klucznik RP, et al.: Interhospital Transfer Before Thrombectomy Is Associated With Delayed Treatment and Worse Outcome in the STRATIS Registry (Systematic Evaluation of Patients Treated With Neurothrombectomy Devices for Acute Ischemic Stroke). Circulation 136:2311-2321, 2017

7. Leira EC, Hess DC, Torner JC, Adams HP, Jr: Rural-urban differences in acute stroke management practices: A modifiable disparity. Archives of Neurology 65:887-891, 2008

8. Southerland AM, Johnston KC, Molina CA, Selim MH, Kamal N, Goyal M: Suspected Large Vessel Occlusion: Should Emergency Medical Services Transport to the Nearest Primary Stroke Center or Bypass to a Comprehensive Stroke Center With Endovascular Capabilities? Stroke 47:1965-1967, 2016

9. Holodinsky Jessalyn K, Williamson Tyler S, Kamal N, Mayank D, Hill Michael D, Goyal M: Drip and Ship Versus Direct to Comprehensive Stroke Center. Stroke 48:233-238, 2017

10. Urban and Rural Classification and Urban Areas Criteria. Suitland, Maryland, US Census Bureau, 2010