DOES ARTIFICIAL INTELLIGENCE IMPROVE OUTCOMES IN ACUTE ISCHEMIC STROKE CARE?
Author(s)
Maria X. Sanmartin, PhD1, Jeffrey M. Katz, Dr.2, Jason J. Wang, PhD3, Elizabeth Rula, PhD4, Casey Pelzl, MS4, Eric Christensen, PhD4, Mir Ali, PhD5, Pina C. Sanelli, Dr.2;
1Northwell Health, Department of Radiology, Hempstead, NY, USA, 2Northwell Health, Manhasset, NY, USA, 3Northwell Health, Glen Cove, NY, USA, 4The Harvey L. Neiman Health Policy Institute, Reston, VA, Reston, VA, USA, 5University of Maryland, College Park, Maryland, MD, USA
1Northwell Health, Department of Radiology, Hempstead, NY, USA, 2Northwell Health, Manhasset, NY, USA, 3Northwell Health, Glen Cove, NY, USA, 4The Harvey L. Neiman Health Policy Institute, Reston, VA, Reston, VA, USA, 5University of Maryland, College Park, Maryland, MD, USA
OBJECTIVES: Despite uncertainty about the clinical effectiveness of artificial intelligence (AI) tools, they have been widely adopted, particularly for time-sensitive diseases like stroke. However, the proliferation of numerous AI tools raises skepticism about this costly, disruptive healthcare innovation, for which insufficient evidence of outcome exists. Our purpose is to assess differences in stroke outcomes before and after the implementation of AI-tools in NYS’s largest health system, while addressing sociodemographic disparities.
METHODS: We conducted a retrospective observational study of consecutive patients with acute ischemic stroke (AIS) who presented to a single comprehensive stroke center in the United States, from 2014 to 2024. Data collection included: demographic, clinical characteristics, and early clinical outcomes, and discharge disposition calculated from Electronic Health Record (EHR) timestamps. We use propensity score analysis to mitigate potential bias from confounding factors between the AI-exposed group and the control group. We performed an interrupted time series (ITS) analysis to estimate changes in outcomes associated with the implementation of AI-tools.
RESULTS: Of the 882 patients with AIS included, 50% were female. After controlling for all covariates, there was no statistically significant immediate level change in mRS 0-2 at the start of 2021, nor was there a significant difference in the trend between the pre- and post-AI implementation periods. The adjusted ITS model also shows discharge disposition outcomes. After controlling for all covariates, there was no statistically significant immediate level change in mRS 0-2 at the start of 2021 for home, rehab facility, nursing home, and other.
CONCLUSIONS: We found no significant difference in functional outcomes and discharge disposition in AIS patients between the pre- and post-AI implementation periods. Even though research has shown that AI tools expedite stroke diagnosis, streamline workflows, and enhance collaboration among medical teams, these findings warrant further investigations to determine whether AI-tools have a significant impact on patient outcomes.
METHODS: We conducted a retrospective observational study of consecutive patients with acute ischemic stroke (AIS) who presented to a single comprehensive stroke center in the United States, from 2014 to 2024. Data collection included: demographic, clinical characteristics, and early clinical outcomes, and discharge disposition calculated from Electronic Health Record (EHR) timestamps. We use propensity score analysis to mitigate potential bias from confounding factors between the AI-exposed group and the control group. We performed an interrupted time series (ITS) analysis to estimate changes in outcomes associated with the implementation of AI-tools.
RESULTS: Of the 882 patients with AIS included, 50% were female. After controlling for all covariates, there was no statistically significant immediate level change in mRS 0-2 at the start of 2021, nor was there a significant difference in the trend between the pre- and post-AI implementation periods. The adjusted ITS model also shows discharge disposition outcomes. After controlling for all covariates, there was no statistically significant immediate level change in mRS 0-2 at the start of 2021 for home, rehab facility, nursing home, and other.
CONCLUSIONS: We found no significant difference in functional outcomes and discharge disposition in AIS patients between the pre- and post-AI implementation periods. Even though research has shown that AI tools expedite stroke diagnosis, streamline workflows, and enhance collaboration among medical teams, these findings warrant further investigations to determine whether AI-tools have a significant impact on patient outcomes.
Conference/Value in Health Info
2026-05, ISPOR 2026, Philadelphia, PA, USA
Value in Health, Volume 29, Issue S6
Code
CO83
Topic
Clinical Outcomes
Topic Subcategory
Clinical Outcomes Assessment
Disease
SDC: Cardiovascular Disorders (including MI, Stroke, Circulatory)