Article Summary
AI tools for detecting strokes grew in adoption between 2020 and 2023 but remained concentrated in larger hospitals with greater resources, suggesting that access to these technologies depends more on where patients are treated than on their clinical needs.
- AI stroke detection use peaked at 21% in 2022 among Medicare inpatients requiring thrombolytic treatment, growing from 2020 through 2023.
- Adoption was most common at comprehensive stroke centers and hospitals with 1,000 or more beds, indicating concentration in larger facilities.
- No disparities were found across patient demographics or stroke severity, meaning hospital location and resources—not patient need—determined access to AI tools.
- Selective access to AI at well-resourced hospitals may increase disparities in stroke evaluation and management for patients at smaller or less equipped facilities.
- Medicare's NTAP code for AI stroke detection launched in October 2020, with five vendors approved by end of 2021.
The use of AI software for detecting occlusions in stroke patients grew between 2020 and 2023 but remained concentrated in larger, specialized hospitals, according to a study published June 24 in the American Journal of Neuroradiology.
The finding comes from an analysis of 2,116 inpatient stroke episodes across 1,076 unique facilities from October 2020 through December 2023, the period during which Medicare's New Technology Add-On Payment (NTAP) code for stroke AI was available, noted lead author Casey Pelzl, of the Harvey L. Neiman Health Policy Institute in Reston, VA, and colleagues.
“AI tools have the potential to improve speed and accuracy in stroke detection, but our findings show that access to these technologies depends more on where a patient is treated than on their clinical needs," Pelzl said, in a news release from the institute.
Medicare issued the first NTAP code for an AI stroke detection tool in October 2020 and by the end of 2021, AI tools from five different vendors code could be billed under the same code. Large-scale data on uptake and the patient and facility factors associated with NTAP-billed AI use has not previously been examined, the authors noted.
To that end, the group analyzed Medicare data and identified NTAP ICD-10 code 4A03X5D use among inpatient stroke episodes requiring thrombolytic treatment. They assessed stroke severity and access to care using validated scales and the Area Deprivation Index, as well as evaluated associations between NTAP-billed AI use and patient, clinical, and facility characteristics using chi-square tests.
According to the results, among 2,116 AIS episodes across 1,076 facilities, NTAP-billed AI use among inpatients requiring thrombolytic treatment increased annually, peaking at 21% in 2022. NTAP-billed AI use was most common in episodes with CT imaging, at comprehensive stroke centers, and in facilities with ≥1000 beds.
Further, no disparities in NTAP-billed AI use were observed across patient demographics or stroke severity measures. In an adjusted logistic regression models, odds of NTAP-billed AI use were highest in 2022 (odds ratio [OR] 6.0), among beneficiaries in the Stroke Belt (OR 2.0), and at comprehensive stroke centers (OR 1.5), the researchers reported.
"This research affirms that AI for [large vessel occlusion] detection has found its footing at major stroke centers, where CT-based workflows and the infrastructure to support these tools already exist," added co-author Maria Sanmartin, PhD.
Ultimately, however, selective access to AI at hospitals with greater financial, technological, and training resources may further increase disparities in stroke evaluation and management in the inpatient setting, the researchers suggested.
“Future research should evaluate the impact of NTAP code approval on diffusion and uptake of AI use in stroke care and improvement of patient outcomes among [acute ischemic stroke] inpatients,” the group concluded.
The full study is available here.
