Osteoporosis, Vertebral Fracture
Conditions
Keywords
vertebral fracture, osteoporosis, computed tomography, AI
Brief summary
The aim of this study is to evaluate clinical effectiveness and cost-effectiveness after implementing a medically approved and commercially available AI support system for opportunistic vertebral fracture screening in CT examinations within an adapted fracture care pathway (integrated care process) in clinical routine. Data will be compared to historical data (same period the previous year). The main question it aims to answer is: Does opportunistic AI-supported vertebral fracture screening in CT examinations integrated to a fracture care pathway increase the numbers of diagnosed vertebral fractures compared to usual care? CT scans in the clinical routine care will be opportunistically screened for vertebral fractures by the AI for 4 months. All positive findings will be confirmed by a radiologist and triaged by the FLS (Fracture Liaison Service).
Detailed description
Objective: This study aims to compare a 4-month period during which AI support was integrated into the clinical fracture pathway (AI-cohort) with a historical cohort from the same period one year earlier (control-cohort). The prevalence of vertebral fractures on CT scans is expected to remain constant during this period. The two cohorts include: 1. The AI-cohort where an AI algorithm was installed and integrated into the clinical workflow to automatically analyze CT images of the thoracic or lumbar spine for vertebral fractures, i.e. opportunistic screening. The AI algorithm operates on local regional servers, ensuring that no images are sent externally. A worklist of positive findings, indicating the presence of vertebral fractures, is automatically generated in the radiology PACS (Picture Archiving and Communication System). This list is reviewed and verified by dedicated radiologists to confirm the AI findings. Confirmed fractures are then linked to the fracture liaison service (FLS), where the fracture coordinator and osteoporosis specialist care for further patient management according to clinical routine, i.e. based on a medical review, writes a referral to primary care, which will be responsible for further patient investigation and treatment of osteoporosis (i.e. usual care in the region). 2. The control cohort (i.e. routine care), includes historical data from the same period the year before the AI-screening was introduced. In clinical routine, radiologists are expected to report incidental findings, such as vertebral fractures. The referring physician, who sends the referral to radiology, is then expected to act on the incidental finding of a vertebral fracture and is responsible for ensuring that the patient is managed further according to clinical routine for osteoporosis investigation and treatment. Timeframe: Data from the AI-intervention period, spanning 4 months (20 October 2024-19 February 2025), will be analyzed. The same timeframe will be used for the historical cohort from the previous year (2023-2024).
Interventions
DEVICEAI vertebral fracture diagnostics (Flamingo)
The AI (name: Flamingo from Image Biopsy Lab, IBL, Vienna, Austria) screens thoracic and abdominal CT scans for vertebral fractures.
Sponsors
University Hospital, Linkoeping
Study design
Observational model
COHORT
Time perspective
RETROSPECTIVE
Eligibility
Sex/Gender
ALL
Age
50 Years to No maximum
Healthy volunteers
No
Inclusion criteria
* abdominal or thoracic CT scans performed in clinical routine from patients living in the catchment area * women and men * age 50 or higher
Exclusion criteria
-CT scans using spine-only protocol with exclusively skeletal queries
Design outcomes
Primary
| Measure | Time frame | Description |
|---|---|---|
| Vertebral fractures (ICD coded) | Within 3 months from CT scan | Numbers of ICD-coded vertebral fractures within 3 months from the CT scan. |
Secondary
| Measure | Time frame | Description |
|---|---|---|
| Reported vertebral fractures (radiology report) | Within 2 months after the CT scan | Numbers of patients with vertebral fractures that are mentioned in the radiology report after the CT scan. |
| DXA screened patients | Within 12 months after the CT scan | Numbers of patients screened with DXA (osteoporosis investigation) within 1 year after the CT scan. |
| Medical treatment (anti-osteoporotic) | Within 12 months after the CT scan | Numbers of patients treated with antiosteoporotic drugs 1 year after CT scan. |
| Cost-effectiveness | Two time frames: Within 12 months after the CT scan and long-term using a decision-analytic model to assess outcome in 10 years | Incremental cost per outcome compared to no AI guidance. |
Other
| Measure | Time frame | Description |
|---|---|---|
| Performance of AI | Baseline | Performance (true positive, false positive, true negative and false negative) of AI detection of vertebral fractures compared to radiologist (ground truth). Only in the cohort AI screening aligned with FLS |
| Patients requiring further action according to clinical triage (FLS) | 1 month | Numbers of patients with vertebral fracture that needs further action according to the FLS. Only in the cohort AI screening aligned with FLS. |
| Vertebral fractures (radiologist confirmed) | Baseline | Numbers of vertebral fractures that are detected by AI and confirmed by radiologist in an opportunistic clinical setting. Only in the cohort AI screening aligned with FLS. |
| New or worsened vertebral fractures | Baseline | Numbers of radiologist verified vertebral fractures that were new (not previously visualised), or worsened since previous CT (minimum 3 months old). Only in the cohort AI screening aligned with FLS |
| Undiagnosed vertebral fracture | Baseline | Numbers of old previously visualized vertebral fractures that were previously undiagnosed (ICD-code). Only in the cohort AI screening aligned with FLS |
| Untreated patients with vertebral fracture | Baseline | Numbers of patients with vertebral fracture that were untreated at the time of the CT scan. Only in the cohort AI screening aligned with FLS. |
Countries
Sweden
Outcome results
None listed