Embedded Files
How do prediction models work?
How do prediction models work?
Prediction models use data and statistical and machine learning tools to build a model that estimates what might happen in the future.
For example, given person's age and medical history, what are the chances to develop a heart disease in the next 10 years? This is exactly what QRISK tool is built to estimate. Doctors use it to decide who should be offered treatment to help prevent heart problems.
More and more of these models are being created as larger datasets and better tools become available.
Prediction modelling lifecycle.
Behind any model there is training data: based on this data, the model learns the relationships in the training data to predict risk of a health condition.
1 Data to Model
Both data and the prediction algorithm play an important role in getting accurate predictions:
High quality data with sufficient number of participants that represent the population well, rich in related to the health condition information, would be a great start for developing a good model.
Prediction algorithms can be more or less suited for specific data. AI models may be too overly flexible for small datasets, but be of great help for big data.
2 Model to Prediction
Once the model is trained, it can be used to predict the risk for new patients. A person's information is given to the model, and it can return the chances of a health condition being developed in the future.
Ideally, a model should be well calibrated, and its risk estimates can be interpreted in light of the underlying training data. For example, if the estimate is 15%, it implies that among the people with a similar risk score, 15% have developed the health condition over a year.
Page updated
Google Sites
Report abuse