I am planning a retrospective chart review study using data abstraction form.

My questions are:

1) How many number of cases/percentage of cases to be selected for inter-rater reliability assessment?

2) Are the cases selected for the assessment to be included in final results analysis? If it is, there are more than one sets of data from the same case, which one to be included?

3) From my reading, the assessment is done on selected important variables in the data abstraction form, is that correct?

4) Continue from question number 3, if a few variables were compared, how to perform the inter-rater realiability test? The test is perform on each individual variables, or collectively on individual cases? (In some example I read, the assessment compare the outcome of a case, e.g. has or doesn't have the disease, in stead of multiple elements in a case)

5) Can the assessment perform in a pilot test, in stead of in the actual study?

6) Are there any references outline the steps to carry out such assessment in retrospective chart review study?

Thanks