UPmémoires

English

Impact on emergency patient workflow of an artificial intelligence for fracture recognition

Background: In most countries, emergency departments (ED) are subject to a constant increase of admissions for traumatic injuries leading to organizational difficulties.

Purpose: to assess the impact of AI limb bone fracture diagnosis software (Ais) on patient workflow in an emergency department.

Materials and Methods: a retrospective study was conducted in 2 phases, the first without Ais between the first of January 2020 and the 30th of June 2020 (period 1), the second, with Ais (Gleamer, Paris, France) during the same time interval in 2021 (period 2). Ais was introduced on the first of November 2020 and training performed during 2 months. Patients who were hospitalized or had undergone a CT scan were excluded. Discrepancy rate of limb bone fracture (difference between patient's final diagnosis at discharge and radiologist's report), clinically relevant discrepancy (defined as significant changes in patient management linked to the radiologist's report), length of stay in the ED, number of specialist opinions and number of orthopedic appointments within a month after discharge were compared for the 2 periods.

Results : 3720 patients, 1780 for period 1 and 1940 for period 2, were included (mean age of 42 years, Standard deviation 20). During period 2, the number of discrepancies (- 1.58%, p= 0.04), length of stay (at mean -9 minutes, p = 0.03), the number of specialist opinions (-1%, p=0.38) and clinically significant discrepancies decreased (- 1.3%, p = 0.99).

Conclusion : Ais can improve patient workflow by reducing discrepancy rate (-1,58%) and length of stay (- 9 minutes) in an ED.

Keywords : Radiology, fractures, bone, AI Artificial intelligence, emergency, humans, retrospective study.