I've build a competing risk regression model (Fine & Gray model) to predict time-to-nursing home admission in dementia patients (competing risk = death).

Now, I would like to investigate the added predictive value of a new predictor by calculating the NRI at a certain time point.

I have found the R package 'survIDINRI' which appeared to be exactly for this purpose.

Description: Performs inference for a class of measures to compare competing risk prediction models with censored survival data. The class includes the integrated discrimination improvement index (IDI) and category-less net reclassification index (NRI).

In the 'indata' argument, however, you can only specify two option (0=censored and 1=event). I have tried coding the events like this: 0=censor, 1=event of interest, 2=competing event, but this results in error messages. 

Therefore, I have to conclude that the package survIDINRI is not suitable for this purpose. In the phrase "compare competing risk prediction models" from the package description, the word "competing" likely refers to "competing prediction models" rather than "competing risks". 

Any advice on how to calculate the NRI in the presence of competing risks (e.g. suggestions for other packages/software) will be much appreciated!