I have a dataset containing a point score (an integer between 0-41) and mortality (0/1) for a sample.

I have performed a logistic regression analysis on these data (mortality~score) and extracted the parameter estimates (intercept and regression coefficient for the point score), and have used these to calculate the estimated risks for all observations in the dataset.

I have then validated the model using various performance measures, including Nagelkerke's R-squared (using the NagelkerkeR2 function from the fmsb package). So far, no problems.

Then I'd like to validate the model externally in a different sample without re-estimating the intercept and regression coefficient, that is, by using predicted risks calculated using the intercept and coefficient from the first sample.

I cannot figure out how I calculate Nagelkerke's R2 in this setting, without having to re-fit the model to the new data and thus changing the intercept and regression coefficient.

Help would be greatly appreciated (if possible with example code).

Thanks in advance and best regards,
Anders