Franklin Valier wrote: > In science this type of study only has value as to its scientifically > agreed upon use. Its ability to be relied upon to make reliable > conclusions from the methodology has to be taken into perspective when > reading the study. It has value, but in science you don't take it too > seriously. We rely on empirical studies for serious evaluation of a > phenomena. If they haven't been done, all you can say is this is all > have and this is all we know right now. Not much. I wouldn't get too > upset about this.
I think that you are being overly dismissive of observational studies. Controlled experiments are great, but a) they can be hard to arrange when the thing being tested is a hospital-wide information system which costs tens of millions of dollars to implement and b) controlled trials can introduce their own sets of biases and limit generalisability due to overly tight selection criteria. And how practical is it to randomise whole hospitals to "get teh computer system" or "stay with paper"? OPolitically that is rather hard to do. Certainly in the case of evaluations of implementations of hospital and other clinical infromations systems it is best to use a before-and-after study design, in which the hospital acts as its own matched control, and the same survey instruments and methods are used before and after the implementation of the system. It is easy to say that in retrospect, but getting money from management to commission an expensive evaluation study of a new information system BEFORE the system has even begun to be installed can be a challenge, I suspect. Tim C