Farooq et al outlined how to evaluate the performance of a prognostic score. This can help a clinician decide if the model is relevant to his or her patients. The authors are from Erasmus University in Rotterdam, Cardiolysis in Rotterdam and Hospital Clinic in Barcelona.
Factors to consider when evaluating the performance of a prognostic score:
(1) accuracy (overall performance score)
(2) discrimination between those who do or do not have the outcome in question (C-statistic)
(3) calibration of the score (goodness-of-fit such as Hosmer-Lemeshow)
(4) transportability and generalisability
Some things that may impact the performance:
(1) prevalence of the condition in the population
(2) spectrum of disease severity
(3) factors impacting the data (laboratory method, reference ranges, etc)
To read more or access our algorithms and calculators, please log in or register.
