Commonly a mathematical norm for whatever category the physician practices in. The degree to which that computation sorts out “over prescribers” from the norm instead of excluding competent physicians and rewarding idiots is determined by how well the categories are drawn and how much sophistication is involved in making the judgements. I don’t have a lot of faith in the ability of the average insurer to make those calls properly.
All three of the circumstances you mention could wrongly put a PCP out of the running unless each individual physician’s practice is carefully examined. I haven’t seen an excess of “carefully” among the company folks I’ve met.