Off the top of my head I can think of the Journal of Epidemiology study from the late 90's. Compared all singleton births in the US and grouped the data by midwife vs. physician. Found that stats for VLBW, PTL and infant mortality were all lower with CNM's.
There is also the BMJ article from 6/18/2005 which showed CPM stats OOH as good as in hospital.
Both showed be available free on-line.
Also check the CfM website and CIMS- CIMS and the childbirth connection just came out with a new study showing the evidence for CIMS, so might find some thing there.