« Tapping! | Main | Is Voting Genetic? »
October 24, 2007
Return of the RAND?
via Jason Furman, RAND has a response (pdf) to claims of attrition. It's quite convincing on the subject of whether cost sharing reduces utilization (it does -- of course it does). So far as whether this fouls the experiment's data on health outcomes in the cost sharing group, the researchers tracked the health of most of those who dropped out of the study, and found that it didn't differ significantly from the experimental group, suggesting that the drop-outs weren't sicker, which would obviate the problem.
October 24, 2007 | Permalink
Comments
please visit
www.gabrielchristou.blogspot.com
you will see PHOTOS of WHO and WHERE Bin Laden and his NETWORKS ARE….
URGENT…PLEASE HELP…. I CANNOT FROM HERE….. I AM BLOCKED ALL AROUND
FORWARD THIS INFORMATION TO THE FBI.
Posted by: gabriel christou | Oct 24, 2007 7:26:32 PM
kudos for being intellectualy honest and open. Rare to find.
Posted by: richard | Oct 24, 2007 7:59:30 PM
I hope not so very rare, but I agree, richard.
At the end of the day I don't think the RAND study adds that much to the current discussion, since the ways in which our current healthcare system breaks down result in expenses a couple orders of magnitude more at least than the co-pays under study. It would be nice to brush it aside since the RAND study gets misused so much by opponents of Universal Healthcare, but it honestly doesn't make a difference if it stands.
Posted by: idlemind | Oct 24, 2007 8:30:13 PM
I'm not completely convinced. I saw a quotation from the report suggesting they had followed the health of some dropouts, and it made me thing that even if health outcomes "didn't differ significantly from the experimental group," that may not remove all doubts about the study. To wit:
1. It is almost certainly clear that the nature of the high cost plan drove people out -- otherwise there would have been similar drop out rates in both plans, when in fact the rate was about 500% higher in the high cost plan. Surely not chance.
2. The most plausible reason for getting out was to get cheaper health care, care that the participants wouldn't be able to afford, or didn't want to have to pay for, on the high cost plan.
3. This suggests that those who got out probably got more care, on average, than they would have, had they remained in the plan (i.e., if not allowed to drop out, some would certainly have had to, or would have chosen to, avail themselves of less care).
4. This raises the possibility that the reason the health outcomes of those who left the study remained equivalent to those who stayed in it was only because they were able to leave and get cheaper care. Had they been forced to remain in, their outcomes might well have been worse.
I grant that this is speculative. On the other hand, that 6.7% dropout rate, compared to the much smaller rate on the low cost plan (less than one percent, wasn't it?), really does create a problem. It was not a chance factor, and it might have been big enough to skew the results.
Posted by: David in NY | Oct 24, 2007 8:33:17 PM
On the other hand, the RAND folk seem to think they've been able to discount the factors I cite above (primarily the supposition that those who left would have been able to get more care, but also to a lesser degree the view that the difference in drop out rates was non-random). So maybe they've got a point.
Posted by: David in NY | Oct 24, 2007 8:42:31 PM
I think the point of the Rand study is that there is an area of diminishing returns (and one might add negative returns) in health care. Now, that doesn't mean that we shouldn't strive for universal healthcare. However, it does mean that any plan for universal health care should have cost sharing provisions (except for lower income folks). In the end the goal is not simply to get as much health care as possible to achieve better health outcomes but do so in a manner that is cost effective.
Posted by: richard | Oct 24, 2007 8:54:17 PM
In all of the discussions of regressivity in the health care debate I wish I would see one of the candidates to get rid of the health care tax exemption and use part of the increased revenues to subsidize health care for lower income folks. Obama gestures in that direction but in a very small way.
Posted by: richard | Oct 24, 2007 8:56:12 PM
At the end of the day I don't think the RAND study adds that much to the current discussion, since the ways in which our current healthcare system breaks down result in expenses a couple orders of magnitude more at least than the co-pays under study.
Incomprehensible. What expenses are you comparing to what other expenses, exactly? The clear implication of RAND, and subsequent research on the same issue, is that health insurance just doesn't make much difference to health "outcomes." That obviously greatly undermines the case for universal health insurance.
Posted by: JasonR | Oct 24, 2007 9:09:16 PM
I quote point 4 of the RAND response
4. Moreover, Nyman’s speculation about a high degree of non-random attrition
is contradicted in work that is unpublished but also posted on Newhouse’s
home page. In this work Manning, Duan, and Keeler carried out additional
analyses of those who did not complete the Experiment. They concluded that
there was in fact a modest amount of non-random attrition, but that its effects,
if accounted for, would have left our conclusion that cost sharing reduced use
unchanged.
the web page is here
Manning, Duan and Keeler is the third article on the page.
I quote from Manning, Duan and Keeler
"Dropouts were sicker on average at enrollment than those who stayed,"
That is they agree that the estimated health status of people in the plans with copays should biased up due to the loss of data on some of the dropouts.
The arguments concerning health outcomes in the RAND reply are based entirely on the fact that RAND tried to assess outcomes for dropouts and did so for 77% of people whose participation ended prematurely. That is, their claim is based on the assertion that 77% = 100 %
Not up the the level of mathematical expertise that I expect from RAND.
Posted by: Robert Waldmann | Oct 24, 2007 9:40:19 PM
"The clear implication of RAND, and subsequent research on the same issue, is that health insurance just doesn't make much difference to health 'outcomes.'" I don't think RAND claims that, and I don't think that it makes any sense at all. Everybody in the RAND program had health insurance. Some just paid more when they used health services. The study doesn't "undermine... the case for universal health insurance" because it doesn't address the problem of people with no insurance at all. Tday there are tens of millions of uninsured, because insurance is currently a function of one's employment situation, a problem that would be cured by universal health insurance.
Posted by: David in NY | Oct 24, 2007 9:41:22 PM
David,
The RAND study concluded, among other things:
"...free care had at most a small effect on any of the five general measures of health for the average enrollee."
Why do you think that "doesn't address the problem of people with no insurance at all?" It implies that even if they get free care, it probably won't make much difference to their health.
Posted by: JasonR | Oct 24, 2007 10:01:12 PM
RAND seems to claim that there was no advantage to sicker people in leaving the "Experiment" because they were somehow held harmless if they had big expenses. Raises two questions to me: 1) Why then did sicker people in the more expensive plan tend to drop out at dramatically higher rates?, and 2) Does this mean that the plan with co-pays didn't have co-pays if you really got sick? Maybe I just haven't understood the brilliance of their "Experiment" design, or taken the time to fully comprhend it, but their explanation of this point stumps me.
Posted by: David in NY | Oct 24, 2007 10:01:45 PM
The arguments concerning health outcomes in the RAND reply are based entirely on the fact that RAND tried to assess outcomes for dropouts and did so for 77% of people whose participation ended prematurely. That is, their claim is based on the assertion that 77% = 100 %
This is just utter nonsense. They're not asserting that 77% = 100%. They're saying they were able to collect health status data on 77% of the drop-outs (85% of those who survived) and incorprate that data into their findings. The remaining drop-outs whose health status could not be determined were too few to affect the conclusions of the study.
Posted by: JasonR | Oct 24, 2007 10:08:33 PM
Jason: Because, unless I'm completely misunderstanding the "Experiment", that sentence you quote is comparing the outcomes of those in the "free" plan to those in the plan with co-pays, that is to other insured people. I don't think there was any comparison at all of those with some insurance and those with none. Indeed, it's preposterous to think that the uninsured have equally good outcomes to those with insurance -- I mean, if that were true, nobody would bother with insurance. But we all know that, if we don't have insurance and we get really sick (say, accidents like Graeme Frost's family had), we would be totally fucked.
Posted by: David in NY | Oct 24, 2007 10:12:58 PM
Oh that's right Jason, I forgot. You don't need insurance. If you get sick, "you just go to an emergency room." GW Bush.
Posted by: David in NY | Oct 24, 2007 10:18:18 PM
Jason: Because, unless I'm completely misunderstanding the "Experiment", that sentence you quote is comparing the outcomes of those in the "free" plan to those in the plan with co-pays, that is to other insured people. I don't think there was any comparison at all of those with some insurance and those with none.
So what? The only difference between having no insurance and having each of the three cost-sharing plans is in the amount you have to pay out of pocket. The study found that cost-sharing reduces utililization (as expected---since out-of-pocket costs are a disincentive to utilization), but that this makes little difference or no difference to health. This is consistent with a large body of evidence from other sources indicating that health care has little effect on health. Yes, for some individuals in some circumstances, health care makes a huge difference. But in general, across large groups of people, it does not. And the relationship between health insurance and health is even weaker, as the RAND study shows.
Posted by: JasonR | Oct 24, 2007 11:05:18 PM
So what? The only difference between having no insurance and having each of the three cost-sharing plans is in the amount you have to pay out of pocket. The study found that cost-sharing reduces utililization (as expected---since out-of-pocket costs are a disincentive to utilization), but that this makes little difference or no difference to health. This is just silly. There's a thing called a demand curve. If the costs are so high as to reduce utilisation to dangerously low levels (and you could make an argument that there is no such thing, ie no one ever ignores chronic symptoms because they can't afford a doctor and ends up dead or in the ER, but there's nothing in this study to back you up) then there will be negative health outcomes. No one in the study had costs that high.
Yes, for some individuals in some circumstances, health care makes a huge difference. But in general, across large groups of people, it does not.
Assuming health care doesn't ever have negative health outcomes (from Munchausen's Syndrome maybe? A lot of conservative health care arguments seem to point there) these statements are mutually contradictory.
Posted by: kali | Oct 25, 2007 7:57:15 AM
Gah! Your imagination will have to supply the italics tags. Sorry.
Posted by: kali | Oct 25, 2007 7:58:09 AM
This is why wonks are dumb. RAND, as a corporation, just doesn't have the correct expertise to be running any study that oversees health care. They have massive ties to the Republican party, and they have a very clear right wing agenda.
Why we are treating this like it's a peer reviewed study of healthcare run by a medical university I will never know. It clearly isn't, and it clearly doesn't have the credibility to be discussed as it is currently being discussed.
Posted by: soullite | Oct 25, 2007 8:25:30 AM
Jason: "This is consistent with a large body of evidence from other sources indicating that health care has little effect on health."
I await with bated breath your citations to these "sources." And I presume you never go to a doctor, eh? Because that would have "little effect on health." And people over 50 shouldn't get flu shots (and just risk dying), and people with cancer shouldn't bother getting radiation treatments (and just die quietly), and people who've suffered traumatic accidents shouldn't get rehab care (and just atrophy at home), etc., etc., because these things "have little effect on health." Christ on a cracker, man, grow up!
Posted by: David in NY | Oct 25, 2007 9:35:03 AM
And the relationship between health insurance and health is even weaker, as the RAND study shows.
JasonR, you are an idiot. The RAND study did not look at insurance vs. no insurance, it looked at insurance with copays vs. insurance with no copays.
The only difference between having no insurance and having each of the three cost-sharing plans is in the amount you have to pay out of pocket.
Yeah, and the only difference between being rich and being poor is the amount of money in your checking account. And the only difference between hot and cold is, you know, the temperature.
Better trolls, please.
Posted by: DMonteith | Oct 25, 2007 9:57:15 AM
Hey JasonR, you have canceled your health insurance now, haven't you? In light of your epiphanies regarding the RAND study, I'd be very disappointed to hear that you have health coverage.
After all, it makes no difference to your health whatsoever. And if anything untoward should happen to you the only difference will be the amount you pay out of pocket, right?
In fact, if people who have read what you've written here discover that you do have insurance, it might lead them to take you less seriously than they would otherwise. You need to cancel that crap now, dude!
Posted by: DMonteith | Oct 25, 2007 10:38:58 AM
David,
I await with bated breath your citations to these "sources."
We've been over all this before. See, for example, this piece.
Posted by: JasonR | Oct 25, 2007 12:01:23 PM
JasonR:
Incomprehesible.
For you, Jason, that's plainly obvious. Your seem unable to comprehend how monumental the difference is between modest co-pays in the health-care environment of 23 years ago and costs to the uninsured today. There just isn't a reasonable basis for comparison, here.
Posted by: idlemind | Oct 25, 2007 3:14:43 PM
JasonR
I don't know about RAND but you are definitely asserting that 85% = 100%. To few to affect the conclusions would be zero. I think you mean too few to affect the conclusions significantly (using the word with its conventional meaning and not as an abbreviation for statistically significantly).
However, you present no argument for your conclusion. How do you know how much bias could be created by the omission of data on drop outs. For example death rates in the study were 1%, if the untracked drop outs all died, the death rate in the copay samples would be twice that of the free care sample. That would be significant.
In the unpublished article to which the RAND reply links, the RAND affiiliated authors note that, at the beginning of the study those who later dropped out were on average sicker than those who didn't. Thus given the fact that only 4 of over one thousand participants in the the free care program dropped out, the loss of data on drop outs should have biased results against free care.
Your non calculation of the magnitude of this bias is based on nothing except your apparent belief that 85% = 100%.
I suggest you look into a remedial arithmetic class.
ps
I used the raw 77% number not the 85% number, because
I am not sure whether people who dropped out and were known to have later died were ignored completely (I can't tell from the RAND reply or the article to which it links)
Posted by: Robert Waldmann | Oct 25, 2007 3:50:36 PM
The comments to this entry are closed.
