Is a comparison of patients between 2 clinics using propensity score matching a matched case control study? I'm helping my boss design a study that will look at the effects of a clinic level intervention on a patient level outcome. 
I would like to look at the effect of our intervention on one clinic (after 3 years) when compared to other patients in the same population that attended a clinic without the intervention using a propensity score matching to account for selection bias, since the patients cannot be randomly selected.  
My boss is calling it a matched case control study.  Is this accurate?  One of my coworkers does not think so, and I'm having a hard time explaining things.
 A: If you are gathering "exposure" data at the outset of a study and then comparing outcomes that develop at a later time, then you have a cohort study rather than a case-control study. I'm a retired physician. This error in understanding the differences in study design seemed especially common among surgeons in my experience, although internists and family physicians were by no means immune. Many physicians rise to positions of authority without good statistical training. Trying to correct their errors in terminology can be a delicate surgical procedure.
You can use propensity scores to do matching or adjustment with either a cohort design or a case control design. That aspect does not determine the nature of the study. Case-control studies are characterized by gathering subjects based on their outcomes rather than on the basis of their exposures. Your primary exposure of interest is attendance at a particular clinic, and you are following two cohorts of patients with subsequent outcomes. A contrary example would be gathering all cancer cases and comparing to non-cancerous controls. Your study analysis does not suggest any matching has been done, so you would be using statistical adjustment with propensity scores to substitute for matching. Matching is another aspect often misunderstood (and incorrectly assumed to be some sort of preferred design) by statistically untrained supervisors. There is an excellent section on the perils of matching (and the superiority of regression adjustment) in Breslow and Day's two-volume book "Statistical Methods in Cancer Research". See chapter 7 of volume 2.
It is possible to perform a case-within-cohort study as well. In such a design you would have some data features (perhaps sex and age) on all of a cohort, but only gather particular data (say a full chart review)  on cases and a subset of controls. This can be useful when the time or cost to gather the interesting data features would prevent a universal assessment. 
