TY - JOUR
T1 - Designing a Complex Intervention Using Early Model-Based Cost-Effectiveness Analysis: A Case Study of an Artificial Intelligence-based Algorithm to Identify Vertebral Fragility Fractures
AU - Dalal, Garima
AU - Youn, Ji-Hee
AU - Kariki, Eleni
AU - Bromiley, Paul
AU - Luetchens, Shawn
AU - Cootes, Timothy
AU - Payne, Katherine
PY - 2021/6/1
Y1 - 2021/6/1
N2 - Objectives: Understand if, and how, using an artificial intelligence-based algorithm (AI) to identify vertebral fragility fractures (VFFs) from routine computed tomography scans (CT) should be used to maximise added value to patients. Methods: A de-novo decision-analytic model (lifetime horizon; NHS-England perspective) linking a bespoke decision-tree with a published discrete-event simulation (DES) was conceptualised and developed for a cohort of 400,000 individuals aged 70 years. The intervention (ASPIRETM) used AI to identify VFFs from existing CT with referral (to Fracture Liaison Service (FLS) or general practitioner) to start a bisphosphonate. The comparator was the current practice of radiologists identifying VFFs from CT and referral to start a bisphosphonate. Technical validation was completed using TECH-VER criteria. Model input parameters were identified from published literature and structured expert elicitation. For ASPIRETM and current practice the base-case analysis reported: number of VFFs identified; costs (£; 2014); quality-adjusted-life-years (QALYs). Uncertainty was quantified using one-way, two-way, scenario, threshold and probabilistic sensitivity analyses. Results: ASPIRETM identified 47,029 additional VFFs, costing an additional £8,681,804 (95% confidence interval (CI): £8,606,882 to £8,756,726) generating 139 (CI: 137 to 140) QALYs. The incremental cost-effectiveness ratio was £185 per additional VFF identified. All QALY gains (0.00035 per person) were derived from starting a bisphosphonate (the DES component). Threshold analysis showed increasing QALY gains to 0.00108 resulted in £20,000 per QALY gained. Key drivers of cost-effectiveness were: specificity; ASPIRETM unit cost; radiologists’ time averted by ASPIRETM and FLS cost. There was substantial uncertainty in the limited evidence available. Conclusions: Indicative cost-effectiveness analysis shows the importance of embedding ASPIRETM into a complex intervention directing people to effective bone management strategies (fall prevention, exercise and nutrition programmes) in addition to bisphosphonates and have data to show radiologists’ time averted. Importantly, decision-makers must be sufficiently certain in the model-estimated QALY gains from bisphosphonates to understand the potential value of ASPIRETM.
AB - Objectives: Understand if, and how, using an artificial intelligence-based algorithm (AI) to identify vertebral fragility fractures (VFFs) from routine computed tomography scans (CT) should be used to maximise added value to patients. Methods: A de-novo decision-analytic model (lifetime horizon; NHS-England perspective) linking a bespoke decision-tree with a published discrete-event simulation (DES) was conceptualised and developed for a cohort of 400,000 individuals aged 70 years. The intervention (ASPIRETM) used AI to identify VFFs from existing CT with referral (to Fracture Liaison Service (FLS) or general practitioner) to start a bisphosphonate. The comparator was the current practice of radiologists identifying VFFs from CT and referral to start a bisphosphonate. Technical validation was completed using TECH-VER criteria. Model input parameters were identified from published literature and structured expert elicitation. For ASPIRETM and current practice the base-case analysis reported: number of VFFs identified; costs (£; 2014); quality-adjusted-life-years (QALYs). Uncertainty was quantified using one-way, two-way, scenario, threshold and probabilistic sensitivity analyses. Results: ASPIRETM identified 47,029 additional VFFs, costing an additional £8,681,804 (95% confidence interval (CI): £8,606,882 to £8,756,726) generating 139 (CI: 137 to 140) QALYs. The incremental cost-effectiveness ratio was £185 per additional VFF identified. All QALY gains (0.00035 per person) were derived from starting a bisphosphonate (the DES component). Threshold analysis showed increasing QALY gains to 0.00108 resulted in £20,000 per QALY gained. Key drivers of cost-effectiveness were: specificity; ASPIRETM unit cost; radiologists’ time averted by ASPIRETM and FLS cost. There was substantial uncertainty in the limited evidence available. Conclusions: Indicative cost-effectiveness analysis shows the importance of embedding ASPIRETM into a complex intervention directing people to effective bone management strategies (fall prevention, exercise and nutrition programmes) in addition to bisphosphonates and have data to show radiologists’ time averted. Importantly, decision-makers must be sufficiently certain in the model-estimated QALY gains from bisphosphonates to understand the potential value of ASPIRETM.
U2 - https://doi.org/10.1016/j.jval.2021.04.603
DO - https://doi.org/10.1016/j.jval.2021.04.603
M3 - Meeting Abstract
VL - 24
SP - S123-S124
JO - Value in Health
JF - Value in Health
SN - 1098-3015
IS - S1
ER -