Explainable AI (XAI) Models
XAI models help predict health outcomes by using machine learning methods that show how decisions are made, making it easier for doctors and patients to understand and trust the results.
At a glance
Use when
Evaluating AI-driven predictions in clinical settings where transparency is essential, such as personalized risk assessment
Avoid when
When only black-box predictions are acceptable without need for interpretation or when data are insufficient or biased
Inputs
Clinical and demographic data from patients with T1D or RRMS
Outputs
Predicted health outcomes with explanations of key contributing factors
How it works
Explainable AI (XAI) models such as random forest, logistic regression, AdaBoost, LightGBM, XGBoost, and CatBoost are used to predict health outcomes in conditions like Type 1 Diabetes (T1D) and Relapsing-Remitting Multiple Sclerosis (RRMS). These models provide interpretable outputs by highlighting feature importance and decision pathways.
- Project
- HTx
- Funding
- Horizon 2020
- Project status
- Completed 2024
- HTA domains
- Clinical Effectiveness
- Categories
- Predictive ModellingML/AI
- Technology
- Medicines
- Assumptions
- Model inputs are representative of real-world patient populations; features used are clinically relevant and available
- Strengths
- Provides transparency in predictions; supports clinician decision-making; handles complex, non-linear relationships in data
- Limitations
- Performance depends on data quality and representativeness; some models may still be complex to interpret without visualization tools
- Geographic & clinical scope
- T1D, RRMS
- Also known as
- XAI, Explainable Machine Learning
Questions this answers
- › Can this model predict my risk of complications from T1D?
- › How does the model decide which patients are at higher risk?
- › Why should I trust this AI prediction?
- › Can doctors understand how the AI reached its conclusion?
- › How is this different from other AI models that act like black boxes?
- › Does the model work equally well for all patient groups?
Similar by meaning
Beta record. Based on the original catalogue summary; primary-source enrichment pending.