Frequently Asked Questions
What is the Limb Viability Model?
It is a Bayesian Network-based system providing outcome probabilities for limb revascularisation following lower-limb vascular injury.
Why was the model developed?
It was developed as part of a larger project to build clinical decision support tools for trauma clinicians caring for injured patients.
How was the model developed?
The model was developed using a stepwise process: Step 1) identifying factors that influence limb outcome following severe limb injury. This was done through analysis of published evidence. Step 2) mapping the interactions between these factors into a Bayesian Network, through expert review and application of knowledge. Step 3) training the Network with anonymised trauma registry data. Step 4) validate the models performance. Where it performed less well, to understand why this was and to fine tune the Network accordingly. The final step was to test performance on an external population, unrelated to the data used in the earlier stages.
Who developed the model?
This is a collaborative effort by computer scientists and trauma and vascular surgeons from Queen Mary University London, Barts Health NHS Trust, UK Defence Medical Services, and the US Department of Defense. Funding sources included a US federal grant (W81XWH - Exploitation of Bayesian Networks for Clinical Decision Support on the Battlefield).
How effective is the model?
The model demonstrates high predictive accuracy – with an Area under the Receiver Operating Characteristic curve (AUROC) of 0.95-0.97 in the validation populations. It is also well calibrated, meaning that the model’s prediction outputs are aligned to the actual real-world observed outcomes, across the entirety of their distribution.
Does the model predict amputation?
No, it predicts the potential failure of surgical revascularisation, which is a direct measure of limb viability, not the surgical decision to amputate.
Why did the developers choose this outcome to model?
Predicting viability offers a direct assessment of the limb’s true condition post-surgery, unlike models predicting amputation, which may be influenced by subjective surgical decisions. While a non-viable limb is a valid reason to perform amputation, there are also other indications for amputation of a viable limb, such as poor function or chronic pain, that this model does not predict.
Why is the model focussed on vascular injury?
Vascular injury is a critical determinant of early limb viability. While including other relevant factors, the model is specifically tailored for cases requiring vascular surgery intervention.
Is the model approved for use in real-time decision making?
Currently, the model is not regulated by any regulatory body and, as yet, is not approved for real-time decision-making or clinical decision-support.
Can this model support clinical audit and quality improvement?
Yes, it can be used in conjunction with other audit tools to review patient care as part of a defined quality improvement programme. It can also be used by individual practitioners to explore outcomes in different clinical scenarios, and for the purpose of training.