Brier Score.
The model’s overall performance is quantified by considering the
distance between the actual outcome and the predicted outcome with
better models have smaller distances5. The Brier score
is used to calculate the model’s overall performance and is measured by
calculating the squared differences between actual binary outcomes and
predictions calculated by the model. The range of values that the Brier
score of a model can take lies between 0 and 0.25 with 0 indicating a
perfect model and 0.25 indicating a non-informative model with only a
50% incidence of the outcome2, 5. Brier score for
survival outcome is not possible to calculate directly because of
censoring. However, it is possible to calculate it indirectly defining a
weight function that considers the conditional probability of being
uncensored during the time. One disadvantage of the Brier score is that
its interpretation depends on the incidence of the outcome with lower
(higher) incidence corresponds to lower (higher) Brier
score2.