Use of AI software led to improved outcomes for those with intracranial hemorrhage

Key takeaways:

• Study analyzed 587 adults with intracranial hemorrhage before and after artificial intelligence algorithm implementation.
• All-cause mortality was significantly reduced in the post-AI group.

Use of artificial intelligence triage software led to a significant reduction in 30- and 120-day all-cause mortality for those with intracranial hemorrhage compared with those treated without the algorithm, according to research.

“Artificial intelligence in health care is growing rapidly,” Dmitry Kotovich, of the Institute for Research in Military Medicine at The Hebrew University of Jerusalem in Tel Aviv, Israel, and colleagues wrote in the International Journal of Emergency Medicine. “Intracranial hemorrhage is one condition that is highly impacted by AI’s ability in prioritizing and triaging suspected findings, leading to earlier therapeutic interventions.”

Kotovich and fellow researchers sought to assess the effect of a commercial artificial intelligence (AI) solution in the ED on clinical outcomes in a single level 1 trauma center.

They conducted a retrospective cohort study for two time periods that analyzed 587 participants: pre-AI (January 2017 to January 2018; n = 289) and post-AI (January 2019 to January 2020; n = 298) in a level 1 trauma center, where an algorithm for intracranial hemorrhage (ICH) was applied to all individuals with a confirmed diagnosis of ICH on head CT upon ED admission.

Those admitted to the ED during the same time periods for other acute diagnoses such as ischemic stroke and myocardial infarction served as control groups. Primary outcomes were measurements of 30- and 120-day all-cause mortality. The secondary outcome was morbidity based on Modified Rankin Scale (mRS) score at time of discharge. Variables such as demographics, patient outcomes, and imaging data were part of the data set.

According to results, all-cause mortality at 30- and 120-day intervals were significantly reduced in the post-AI group compared with the pre-AI group (27.7% vs. 17.5% and 31.8% vs. 21.7%, respectively).

Data additionally showed that mRS scores at discharge were significantly reduced post-AI implementation (3.2 vs 2.8), but no significant differences between the two time periods were found with respect to demographics, comorbidities, type of ICH and length of hospital stay.

“By flagging a life-threatening, time-sensitive pathology, the AI solution may improve overall reader efficiency ... contributing to the timeliness in which radiologists can get to read scans .. and possibly have the potential to enhance radiologists’ accuracy,” Kotovich and colleagues wrote.