부산대학교 도서관

Background:Several accrediting bodies certify the level of stroke care hospitals provide. The Joint Commission on Hospital Accreditation (JC) is the largest accrediting body in the United States. There is no open source Geographic Information Systems (GIS) dataset showing the distribution of JC accredited centers by ZIP code.Objective:to create a stroke center accessibility and stroke center desert system using geospatial analysis and machine learning which provides real-time assessment of stroke center availability, distribution and access to care.Method:Geospatial data layers of JC accredited stroke centers were compiled using data sources including U.S. Census Bureau and CDC. Map layers corresponding to the levels of JC accredited stroke hospitals geolocated using ZIP code were created as follows: 1) Acute Stroke Ready 2) Primary 3) Thrombectomy Capable 4) Comprehensive Stroke Center. A GIS dataset displaying stroke mortality by region was obtained from the ArcGIS Living Atlas. Stroke center deserts are analyzed using a 4.5 hour drive map along with population and diversity. Machine learning models were implemented to estimate stroke mortality as a function of distance to care centers and capability levels of the stroke centers.Results:Stroke centers are highly concentrated within large urban centers. There are geographic regions that have poor access to stroke centers. Such regions include the Gulf Coast States of Louisiana, Mississippi, and Alabama that have large areas with poor stroke center acces while having some of the highest stroke mortality in the country. (Figure 1)Conclusion:There are regional variations in stroke center availability. There are certain regions with high stroke mortality with very little stroke center access. Geospatial AI tools can be utilized to improve stroke systems of care.