I have a hospital with approx 3500 admissions a year and i want to examine patient outcomes after discharge. I want to administer a subjective generic health outcome assessment to a representative sample of patients at 2 time-points: On admission and 1 month after discharge. Obviously patients will present with a range of illnesses and it would be great to examine which illnesses show greater likelihood or degree of change after treatment and discharge but this must add greatly to the sample size needed and may even require assessment of all patients (if that were possible). However, if i were to just focus on patient outcomes in general instead of diagnosis, could anyone tell me what sort of sample size i should aim for? Also, if i were to try to examine outcomes for the 5 most prevelant types of illnesses to be admitted is there a way to get a representative sample from all admissions?
Determining sample size for patient outcomes depends on what sort of outcomes you are measuring (e.g., continuous, binary, polychotomous, ordinal) and whether those outcomes will be considered together (multivariate) or separately (univariate). Next you need, among other things, an estimate of the expected effect size and, if outcomes are continuous, estimates of variability. Once you have this information you need software that will estimate the power and sample size. A power and sample size estimation can be quite complicated and is best done with a statistical consultant.