[AVBC 2025] Reducing Time to Clinical Decision (TCD) after Chest X-Ray in NHGP using Imaging AI
Care Process & Redesign
Technology
Appropriate & Value-based Care Conference
National Healthcare Group
10 November 2025
To reduce Time to Clinical Decision (TCD) after Chest X-Ray in NHGP using Imaging AI. Safe integration of a CXR AI model into a clinical workflow reduces TCD and has the potential to reduce overall patient wait times.
Year Submitted: 2025
Published Date: 10 November 2025
Tags: Technology, Care Process ＆ Redesign, Quality Improvement, Workflow Redesign, Productivity, Time Saving, Cost Saving, Clinical Practice Improvement, Operational Management, Resource Allocation, Digitalisation, Automation, Artificial Intelligence
About this Content
Aims
To reduce Time to Clinical Decision (TCD) after Chest X-Ray in NHGP using Imaging AI.
Background
The intersection of CXRs at primary care and scarce radiologist resources represents a resource allocation challenge that impacts TCD. Integrating AI into the workflow allows faster TCD by letting primary care physicians review patients before formal reporting.
Methods
A commercial HSA-approved CXR AI model LUNIT INSIGHT version 3.5.1.3 was evaluated and adapted for co-triage use in the local primary care setting. Implementation was achieved through technical phases overlapping clinical workflow phases.
Results
Overall mean TCD was reduced by 12 mins, a 36.2% improvement. Normal cases saw a TCD decrease of 21.7 mins, a 64.8% improvement. Abnormal cases had a 5.6 min reduction, a 17.1% improvement. AI performance showed sensitivity of 90.9%, specificity 66.9%, PPV 75.0%, and NPV 87.1%. No observed increase in CXR recall rate post-implementation.
Conclusion
Safe integration of a CXR AI model into a clinical workflow reduces TCD and has the potential to reduce overall patient wait times. Clinical standard in terms of recall was maintained.
Lessons Learnt
The AI model has sufficient sensitivity to support a triage function at point of care, effectively differentiating studies into those with high versus low probability of radiological abnormalities. However, the AI-triaged worklist is not yet default for radiologists, which may limit the full benefit of the prioritization workflow.
Keywords
AI, CXR, Imaging, Workflow, Triage, Clinical Decision
Innovators' Details
Innovators' Details
Healthcare Cluster(s) | National Healthcare Group |
Organization(s) Involved | National Healthcare Group Polyclinics, Tan Tock Seng Hospital, Lee Kong Chian School of Medicine, Nanyang Technological University, NHG Diagnostics |
Platform(s) | Appropriate & Value-based Care Conference |
Healthcare Professional Group(s) | Medical |
Applicable Specialty or Discipline | Diagnostic Radiography |
Project Lead(s) | Fong Qi Wei |
Project Member(s) | Soon AYQ |
Connect with this contributor!
Fong Qi Wei - qi_wei_fong@nhgp.com.sg