(1) as logicallee has mentioned, we're getting better at treating patients and preventing death due to disease. The better we get, the more different treatments patients get, the higher is the risk of a medical error happening.
(2) it is very difficult to know exactly what proportion of "medical errors" are preventable. For instance, the seemingly "simple" case of healthcare-associated infection is far from simple, and except for a few cases (infection associated with elective surgery, and so on), there are NO known intervention to reduce risk further. I will give you an example: a patient needs a central catheter in order to survive. If we don't put it in, and he dies, he'll be classified as a "death due to primary disease X". If we do put a central line, and he acquires an infection (and a certain baseline rate of infection IS inevitable, because the catheter is indwelling and gets colonized by the patient's own flora), and dies 2 months later due to sepsis, it will be codified as "medical error". Adverse drug reactions are also codified as medical errors, even if the drug was administered in a life-or-death situation, when even the patient's ID was not known, even less his full list of medications (and thus, the error could not be prevented in any way). I could go on and on.
That being said, I also see stupid preventable errors like quietly giving the drugs to the wrong patient going unreported. Hospital stay is certainly dangerous.