In the original Hofling hospital experiment (1966):
A person would telephone a nurse, saying that he was a doctor and giving a fictitious name, asking the nurse to administer 20 mg of a fictitious drug named "Astroten" to a patient, and that he/she would provide the required signature for the medication later. A bottle labelled "Astroten" had been placed in the drug cabinet, but there was no drug of that name on the approved list. The label clearly stated that 10 mg was the maximum daily dose.
There are several reasons why the nurse should refuse the request:
- They don't know the doctor
- They don't know the drug
- The requested dose is greater than the labelled maximum dose
- Paperwork was not provided
Nonetheless, 21 of 22 nurses complied. One of the (many) disturbing aspects of the experiment, is that even with so many reasons to refuse, the nurses tended to obey. The implication is that with fewer reasons to refuse, they would be even more likely to obey.
This turned out not to be quite true, as per a partial replication by Rank and Jacobson (1977), in which they eliminated one of the reasons: They used a drug known to the nurses (Valium) instead of a fictitious drug, and somewhat counterintuitively got much lower compliance rates.
So if you wanted to have a particular hospital patient "eliminated", it seems to me that Hofling provides a simple plan: Make a phone call from a payphone somewhere to the nurse's station at the hospital, pretending to be a doctor, and ask them to administer a lethal dose of some poison - the only caveat is that as Rank and Jacobson showed, you don't want to refer to a drug they are familiar with, so ideally a lethal toxin placed in the drug cabinet ahead of time (or equivalently, a mislabelled toxin).
Rank and Jacobson suggest that perhaps reduction in obedience may be partially due to cultural changes: Changes in the relationship between doctors and nurses, the self-esteem of nurses, changes in the legal landscape and fear of lawsuits, etc. Have there been any recent replications of the experiment that suggest how easy / hard it is today to have a patient anonymously murdered by a nurse?