Initial ideas about performance measures using this paper as guidance: Assessing the Performance of Prediction Models A Framework for Traditional and Novel Measures
Ewout W. Steyerberg (link)
Initial ideas about selection of outcomes for upcoming predictive modeling work
Steyerberg is great. One short-coming is little information on non-binary outcomes. Altman has a couple good papers on model performance for Cox models (perhaps the hardest since there’s no direct risk estimate). Here is my favorite: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3667097/