Risk Prediction Modeling Resources for Survival Data

[Amanda Markovitz]

Hi, I'm planning to run a survival risk prediction model (specifically, the Fine and Gray competing risk model using stcrreg) and looking at whether adding a variable improves model fit, using measures of discrimination (comparing c-statistics) and reclassification (IDI, NRI, continuous/category-free NRI). There are quite a few resources available to do this type of thing in R or SAS, but unfortunately my searches for packages in Stata have only returned those that work for logistic regression. I want to know if I'm missing anything out there that could help me so I don't just have to write everything from scratch or move to SAS/R.

[Amanda Markovitz]

To add to this, I did find an article published that described a user-written command called "adpredsurv" which supposedly calculates reclassification statistics for survival data. However, I couldn't find any code associated with the paper and it did not exist in ssc. If someone could help me in using this command (if it in fact exists) that would be very helpful. The original article is open access and is available at: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5192998/

[Andrew Lover]

Proably worth emailing the authors to ask about -adpredsurv-
And this might be useful? http://www.imperial.ac.uk/nhli/r.new...s/predsurv.pdf

[Amanda Markovitz]

Oddly enough, the email address for the corresponding author doesn't work but I've tried the journal editor and will post a follow-up if I learn more. Predsurv looks like it will be useful for comparing measures of discrimination so thank you.

[John Moran]

Hi Amanda

Andrew Vickers at "https://www.mskcc.org/departments/epidemiology-biostatistics/health-outcomes/decision-curve-analysis-01" has the Stata command "stcda" for survival analysis and net-benefit (see the discussion of this index in the "TRIPOD" statement: Moons et al. Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD): Explanation and Elaboration. Annals of Internal Medicine. 2015;162(1):W1-W73).

I guess until "adpredsurv" is located (I agree with you, it seems not to have been posted and the first author's e-mail is not functional), the R module "survIDINRI" by Hajime Uno will admirably serve the purpose.

john moran

[Charlotte Debiais]

Hi,

Did you find a way to assess the discrimination after a survival risk model with competing event (I use stcrreg)?
I am looking for the equivalent of "estat concordance" after a Cox model...

Thank you,
Charlotte