I'm hoping for general advice on how to tackle a problem. There is an existing 10-item severity score.
I have data on 1000 patients; specifically, I have the score for each item, the overall score (calculated by summing the score for each item) & some outcome data (e.g. death). W
What I'm trying to do is simplify the score, i.e. work out which of the 10-items is least useful. Part of that's by looking at each item & working out (from a clinical perspective) which item(s) seem redundant (e.g. HR & RR often correlate), but I also want to explore this statistically. How can I do this?
My initial thought was to work out correlation coefficients between each item, and between the overall score. The idea is either that some items will so closely correlate with each other it's not necessary to have both, or some items will not change much with the overall score so they can be excluded anyway.
Does that make sense?
