0

I've recently developed a risk prediction model to predict risk of coronary heart disease. The model was developed in a cohort of adults (>45 yrs old)followed for a median of 15 years. It was developed using all observations in the cohort.

Now, I want to externally validate this model. However, the validation cohort, which I'm thinking to use,was followed for a median of 8 years. Since I can't find any other cohort followed for a similar length of time, can I use the cohort which I mentioned it to validate my model, at least to check how my model performs in a population followed for a shorter duration? I thank you all, in advance. Cheers, John

amoeba
  • 104,745
user109294
  • 11

1 Answers1

0

Technically you can use the 8-year cohort to validate the model, but in fact the dependet variable would have another meaning. I suppose your dependent variable is simply the probability of survival, so the probability of surviving 15 years is clearly different from the probability of surviving 8 years and some factors may influence both probabilities in different ways.

Probably a better idea would be to split your first cohort into a training and a test set. Then you can use the training set for fitting and the test set for validating the model.

bratwoorst711
  • 96