AI helps improve care quality in stroke patients

Use of artificial intelligence (AI) as part of the support system can greatly improve the quality of care in patients who just had a stroke and may even prevent the occurrence of various vascular events in the future, according to a study presented at ISC 2024.

Adoption of the AI-based clinical decision support system (AI-CDSS), a promising approach to improve healthcare delivery, results in significantly fewer vascular events at 3 months and leads to improvements in stroke care quality when compared with usual care among patients with acute ischaemic stroke (AIS).

“We conducted a multicentre, cluster-randomized clinical trial among 77 hospitals in China from January 2021 to June 2023,” said the researchers led by Xinmiao Zhang from Beijing Tiantan Hospital, Capital Medical University, Beijing, China.

Zhang and colleagues determined the efficacy of an AI-CDSS on stroke care quality and clinical outcomes among AIS patients. AI-CDSS support including AI-assisted imaging analysis, stroke aetiology classification, and evidence-based treatment recommendations was adopted by 38 hospitals in the intervention group, while usual care was provided by 39 hospitals in the control group.

A new composite vascular event (ie, ischaemic stroke, haemorrhagic stroke, myocardial infarction, or vascular death) 3 months after stroke onset served as the primary endpoint. A composite AIS quality score, defined as the total number of 13 preset performance measures carried out, divided by the total number of performance measures for which a given patient was eligible, was the secondary endpoint.

In total, 21,603 patients with AIS (mean age 67.0 years, 35.5 percent women) were enrolled in this study, of whom 11,054 were included in the intervention group and 10,549 in the control group. Nearly all participants (21,579, 99.9 percent) completed the 3-month follow-up. [ISC 2024, abstract LB15]

AIS patients in the intervention group had significantly fewer new composite vascular events at 3 months compared with their counterparts in the control group (2.9 percent vs 3.9 percent; adjusted hazard ratio, 0.75, 95 percent confidence interval [CI], 0.59‒0.95; p=0.02).

In addition, participants in the intervention group had a greater chance of getting a higher composite score of evidence-based performance measures than those in the control group (91.4 percent vs 89.7 percent; adjusted odds ratio [aOR], 1.20, 95 percent CI, 1.16‒1.25; p<0.001).

On the other hand, no significant between-group differences were observed in the modified Rankin Scale Score 3‒6 at 3 months (11.8 percent vs 9.6 percent; aOR, 1.24, 95 percent CI, 0.97‒1.59; p=0.08).

Diagnosis

A recent systematic review of 30 eligible studies with machine learning (ML) algorithms showed the potential of AI in the diagnosis and triage of patients with AIS.

Based on the pooled results, ML techniques accurately identified large vessel occlusions on computed tomography. ML algorithms also demonstrated accuracy in predicting clinical and angiographic outcomes as well as related factors. [World Neurosurg 2022:159:207-220.e1]

“However, the role of AI in the management and prognostication remains limited and warrants further research to help in decision support,” according to the investigators. [World Neurosurg 2022:159:207-220.e1]