There was an article a few years ago about the fall of the medical autopsy rate. Because it's unclear how we fund it, and most people aren't in a position of knowledge to request it (or in a financial position to pay for it), the rates of autopsy to find a considered cause of death have been dropping precipitously.
Atul Gawande talks about this a lot in "Complications" [1]. A great book about medical procedure, and specifically about the process of learning, trial and error and the whole feedback loop of working with disease.
When my grandfather died with cancer and fibrosis, one of the main complications involved in getting compensation from EEOICPA was the fact that his county doesn't have a coroner. They have a "medical examiner" who is based 70 miles away and never visits, plus some nurses he employs. As a result, death certificates in that county are only signed in extraordinary circumstances. The federal bureaucracy doesn't accept unsigned certificates for the purpose of establishing a cause of death. The county medical examiner is paid a flat yearly fee, so guess how many autopsies he performs?
We shelled out for a private company for an autopsy, but post-mortem, lots of things happen to the body that make identifying a source of an infection really hard – basically, it didn't tell us anything new :(
