My comment is based on my experience in quality improvement research. We usually start with studying "usual clinical practice" and related outcomes, then look for opportunities to improve patient care. In a pragmatic clinical trial, we might propose a method to improve usual care and implement it using cluster randomization. The study question is typically chosen because we find something wrong with usual care in the setting in which we see patients. For instance, we have identified a high incidence of fluid overload in mechanically-ventilated patients in our healthcare system. We have two years of data and have generated statistical process control charting that shows the process is stable. But it's clearly suboptimal. As such, we would not be able to argue the rationale behind it, or justify its risks/benefits. And I don't think we should be required to in a pragmatic clinical trial. I think you can say "usual care is what it is," so long as observational data can be used to describe it. This may cause issues with generalizability, but it would then behoove other institutions to determine if their usual care is comparable to the usual care comparator used in the study.


Minor observation: The abbreviation CER is used without definition.

Robert Raschke
University of Arizona College of Medicine-Phoenix, Division of Clinical Data Analysis and Decision Support
Clinician
Representing my own views