There's another factor too, which the author doesn't really touch on. The likelihood of software being bad increases along with the complexity of the business logic. That's why enterprise software projects are so hard to get right. My understanding is that state regulations vary so much that any kind of medical software is fraught with byzantine complications just to do the most basic thing. So it seems to me the most logical thing we can do to lower the cost and increase the quality of medical software is to adopt some standards. Now how much of a dent that can put in the overall costs is still up for debate, but I bet it could be significant.
This article makes a bunch of very valid points, but the real issue is that health care in the US is simply much too expensive per treatment, and that the cost to society of things like obesity are enormous. That is not going to be cured with 'software' of any kind but with a lifestyle change.