Machine Learning and Improved Quality Metrics in Acute Intracranial Hemorrhage by Noncontrast Computed Tomography

Area of Science:

• Radiology and Medical Imaging
• Artificial Intelligence in Healthcare
• Clinical Workflow Optimization

Background:

• Timely reporting of critical radiology results is vital for patient outcomes.
• Artificial intelligence (AI) offers potential to enhance clinical radiology workflows.
• Machine learning (ML) algorithms can identify high-probability intracranial hemorrhage (ICH) on CT head scans.

Purpose of the Study:

• To evaluate the impact of an FDA-approved ML algorithm on healthcare system metrics.
• To determine if ML reduces report turnaround time (RTAT) and length of stay (LOS).
• To assess ML's effect on emergency department (ED) and inpatient populations.

Main Methods:

• An ML algorithm was implemented across CT scanners in January 2018.
• RTAT and LOS data were collected for periods before (July-December 2017) and after (January-June 2018) ML implementation.
• A total of 50,654 ED and inpatient cases were analyzed.

Main Results:

• Overall RTAT decreased from 75 to 69 minutes (P <0.001).
• ED LOS decreased by 11.5% for patients without ICH (P <0.001).
• Inpatient LOS decreased significantly for both patients with and without ICH.

Conclusions:

• ML utilization was associated with a statistically significant reduction in RTAT.
• ML led to a significant decrease in LOS for non-ICH ED patients.
• Further research is needed to fully understand ML's impact on patient care and outcomes.