Sorry, my message was abruptly sent! I meant to say that I might have a unique perspective: I recently had surgery at a major hospital, and that hospital, in trying to improve itself, sent me a survey questionnaire, with questions probably very similar to those posted by the person posting the initial call for research help. As a program evaluator and survey designer, I responded to the hospital's survey with a number of insights. Therefore, having just had the same kind of experience, probably being the same kind of patient who had to rate nurses, doctors and hospital processes on a number of survey items, I suggest a few approaches be considered:
-- Better specify better (we who have responded don't know enough about your setting or survey topics) to better achieve what I consider a predictive model, to include factors that could relate to patient ratings of nursing effectiveness. I'm thinking of: the unit or division of the hospital where the patient stayed (mentioned before, I believe), the nature or severity of the operation or treatment accounting for the patient's being in the hospital, the length of time that patient stayed there, and whether or not the nurse was on the day or evening shift. (Some of these are dummy/categorical variables.)
-- I don't consider this as a simple bivariate correlational model, with all multiple ratings being averaged or collapsed to form a linear composite of some form. The variance of each, their interactons, and their unique contribution to predicting the Y score, are of extreme diagnostic significance in planning decisions on nursing training, practices, policies, etc. I do like the earlier comment about possibly using a principal components strategy to consider the underlying response dimensions of the multiple survey items, and using those components or factor scores in some form of linear regression model (now with more reliable measures of the topics).
-- And that's where I would go most: to consider the multiple survey items as each measuring an attitudinal perception (not an ordinal or categorical one) of different aspects of how a given nurse worked with these patients. Obtaining beta values on each survey item topic (assuming there's a sufficient level of model fit or R squared), could provide a good deal of payoff to hospital administrators, nursing department heads, unit supervisors and of course, individual nurses, in terms of focusing on what to strengthen in working with patients.
-------------------------------------------
Milton Goldsamt
Consulting Research Psychologist and Survey Statistician
-------------------------------------------