NHS OUTSOURCING COST IMPLICATIONS OF AN AI-PRIORITISED WORKFLOW FOR REPORTING HEAD CT IN EMERGENCY CARE
Author(s)
Lucy Gregory, BSc, MSc.
Consultant Health Economist, Hardian Health, London, United Kingdom.
Consultant Health Economist, Hardian Health, London, United Kingdom.
OBJECTIVES: To develop an early economic model assessing the financial implications of a hypothetical AI-prioritised radiology workflow for reporting Head CT scans in out-hours NHS emergency departments (EDs).
METHODS: A decision model was developed in Microsoft Excel to estimate the 3-year budget impact of the hypothetical AI-driven triage pathway versus standard of care from an NHS radiology department perspective.The model evaluated adult (18≥years) non-contrast head CT (NCCT) scans in EDs, accounting for reporting via contracted NHS staff, insourcing, and teleradiology outsourcing across both in-hours (8 AM-8 PM) and out-of-hours (8 PM-8 AM) periods. The hypothetical AI system was assumed to label images as “prioritised” and “non-prioritised”, theoretically allowing less urgent scans to be deferred to in-hours workflows when outsourcing costs are lower. Modelling inputs were informed by published literature and clinical stakeholder opinion. A one-way sensitivity analysis was conducted to address parameter uncertainty.
RESULTS: For 2024-2025, an estimated 1,202,104 head NCCT scans were requested for reporting. The cost of adopting the hypothetical AI-triaged pathway was estimated at £4,534,734 in year 1, totalling £10,782,671 by year 3. Over three years, the modeled pathway would shift reporting of 2,101,311 images to in-hours, reducing total radiology spending by £52,206,235 (from £234,313,638 to £182,107,403). Sensitivity analysis identified the largest determinants of value to be the proportion of scans sent for reporting via outsourcing OOH, followed by the cost per scan, the number of scans performed out of hours and AI performance in reducing the number of scans requiring immediate reporting OOH.
CONCLUSIONS: While an AI-triaged reporting workflow demonstrates the potential to reduce NHS financial pressures and reliance on costly teleradiology, real-world implementation would depend on evidence of clinical safety. Additionally, multi-site prospective studies are required to validate economic modelling assumptions.
METHODS: A decision model was developed in Microsoft Excel to estimate the 3-year budget impact of the hypothetical AI-driven triage pathway versus standard of care from an NHS radiology department perspective.The model evaluated adult (18≥years) non-contrast head CT (NCCT) scans in EDs, accounting for reporting via contracted NHS staff, insourcing, and teleradiology outsourcing across both in-hours (8 AM-8 PM) and out-of-hours (8 PM-8 AM) periods. The hypothetical AI system was assumed to label images as “prioritised” and “non-prioritised”, theoretically allowing less urgent scans to be deferred to in-hours workflows when outsourcing costs are lower. Modelling inputs were informed by published literature and clinical stakeholder opinion. A one-way sensitivity analysis was conducted to address parameter uncertainty.
RESULTS: For 2024-2025, an estimated 1,202,104 head NCCT scans were requested for reporting. The cost of adopting the hypothetical AI-triaged pathway was estimated at £4,534,734 in year 1, totalling £10,782,671 by year 3. Over three years, the modeled pathway would shift reporting of 2,101,311 images to in-hours, reducing total radiology spending by £52,206,235 (from £234,313,638 to £182,107,403). Sensitivity analysis identified the largest determinants of value to be the proportion of scans sent for reporting via outsourcing OOH, followed by the cost per scan, the number of scans performed out of hours and AI performance in reducing the number of scans requiring immediate reporting OOH.
CONCLUSIONS: While an AI-triaged reporting workflow demonstrates the potential to reduce NHS financial pressures and reliance on costly teleradiology, real-world implementation would depend on evidence of clinical safety. Additionally, multi-site prospective studies are required to validate economic modelling assumptions.
Conference/Value in Health Info
2026-11, ISPOR Europe 2026, Vienna, Austria
Value in Health, Volume 29, Issue 12S
Code
MT31
Topic
Economic Evaluation, Medical Technologies
Topic Subcategory
Digital Health