Talk:RAND Health Insurance Experiment

Updated the article to mention the new 2013 results, arising from more complete data from the Oregon Experiment. The initial claims had to be a bit deflated.

---

what do y'all think?

Pretty good. It's an important study. I added the Keeler summary article. I like the way you linked to the Normand review.

But the RAND health insurance experiment is very controversial -- the Cato Instititute cites it (also in BMJ) to support their claim that co-payments have no effect on health care, and Physicians for a National Health Program cites it (in NEJM) to support their claim that co-payments cost lives. One of the problems with the RAND study (in my interpretation) is that it didn't have enough statistical power to measure actual loss of life. It did demonstrate poorer control of blood pressure in the low income group, from which people could reasonably infer higher mortality.

I'd like to link to some other commentary (I think BMJ has discussed it repeatedly), and summarize some of the discussion in NEJM and JAMA. Some of this work has actually been replicated by insurance companies who did studies with historical controls to see what happened when their clients raised co-payments, and they also found that patients used fewer necessary as well as unnecessary drugs.

Don't fall into the trap of saying, "This side says A, this side says B, here are their arguments," without some kind of evaluation to see if you can tell how valid their arguments are.

It's going to take some work. Nbauman 17:10, 23 October 2006 (UTC)

I don't understand the meaning of the phrase, "There were as many in the higher cost-sharing group needing care as not". Can you cite the text in the sources that you're summarizing? The literal meaning would be that, say, 50% of the subjects in the high-cost group needed care, and 50% did not, or more generally that the number of subjects needing care was not statistically different from the number of subjects who did. I don't think it says that, though if anyone can cite a page where it does, I'll change my mind.

As I recall, the study found higher copayment rates reduced spending because people did not seek care as frequently -- and it reduced spending for necessary care as well as unnecessary care.

I've seen that stated repeatedly in secondary sources, like NEJM and JAMA, and I'm trying to find an explicit statement in the primary RAND publications themselves.

Keeler says on p. 318, "4. People with cost sharing don't just cut out nonessentials," which seems to be a way of saying that people with cost sharing cut out essential treatments.

Keeler also says, in the abstract, "Cost sharing led to poorer blood pressure control."

So I think it would be to say that the HIE found that "People with higher deductibles were more likely to cut out essential treatments, such as medication for blood pressure."

I'd also add that the effect on blood pressure was statistically significant, but the final results on strokes and death from high blood pressure was not, because the study didn't have a large enough number of subjects to measure strokes and death.

(I think that when the study says they "found no difference" between certain outcomes, what they mean is that there was no statistically significant (p>.95) difference that they could measure. It doesn't mean that there was no difference, or that someone wouldn't find a difference with a larger number of subjects.)

Any objections? Any comments? Anybody there? Nbauman 16:59, 10 December 2006 (UTC)

The original statement you can see under history. I'll quote it here: "the care which was not consumed by the higher cost-sharing group was equally necessary and unnecessary care." It confused me thus my change to: "There were as many in the higher cost-sharing group needing care as not". I consider more readable. But it could still be improved. Cakeandicecream 10:42, 18 December 2006 (UTC)

"the only study which can give definitive evidence as to the causal effects of different health insurance plans." This seems to me to be highly misleading, and also the statement about relevance. There is no control for profit-based HMO's, as the only one studied was a non-profit, and the sample size of HMO's studied is exactly one. As the sole basis for a whole system, this study seems woefully inadequate. Turbotaco 18:11, 20 June 2007 (UTC)