Using AI-Driven Triaging to Optimise Clinical Workflows in Non-Emergency Outpatient Settings A Real-World Case Study Concerning the Screening of Tuberculosis

Growing diagnostic imaging workloads threaten to undermine the sustainable delivery of healthcare imaging services, worsening clinical and productivity outcomes. Artificial intelligence (AI) technology promises to help address this issue through its triaging capabilities which can optimise clinical workflows and patient management. However, the impact and trade-offs of AI-driven triaging must be understood to determine whether it is a net positive and more beneficial than a first-in-first-out queueing approach. Existing research has focused on studying the impacts of AI-driven triaging in emergency department contexts and has given less attention to outpatient settings. This research in progress paper presents the preliminary results of an industry-academia collaboration study exploring the real-world impact of AI-driven triaging in the diagnosis of tuberculosis (TB) for an outpatient setting. A mixed-methods approach is adopted to examine radiologist perspectives of its effect on daily clinical practice and to quantify its productivity impact on the time spent completing cases. Further investigation to establish the impact of AI-driven triaging on the diagnostic accuracy of clinicians and its relationship with task productivity is underway.