• How much could we expect the Oregon Medicaid study to reduce blood pressure? – ctd.

    Superstar blogger Adrianna McIntyre of Project Millennial got me those Lurie et al papers. They are two reports of patients who were dropped from the Medi-Cal program in 1982. It’s a case-control study, one report 6 months following loss of coverage and the other at one year.

    The first paper examined 186 patients who lost coverage and compared them to 109 patients who did not. Of these, 51 had hypertension in the group that lost coverage, and 38 had hypertension in the group that did not. They found that those patients in the “study” group had their diastolic (not systolic) pressures go up in the six months after losing coverage (+ 10.0 mm Hg) and those who maintained coverage saw them drop (- 5.0 mm Hg).

    Where to start? This is a very small study, not an RCT. So it’s nowhere near as powerful as the RAND HIE or the Oregon Medicaid study. But, yes, losing Medi-Cal coverage was associated with an increase in systolic BP in those who had hypertension. But it’s more complicated than that. The analysis didn’t control for any covariates at all. A number of the people who had their coverage dropped got health insurance from other sources, or got their Medi-Cal reinstated. So… are we saying that Medi-Cal was better than those other insurance products? Moreover, this was a plugged in group of patients. Check this out:

    More than 95 per cent of the patients in both groups had, and could identify, a regular source of medical care. Nearly 85 per cent of the patients in both groups said they thought they could get medical care whenever they needed it, and approximately 90 per cent were “extremely” or “very” satisfied with the care they received.

    In fact, this study came about because the researchers cared for these patients and were helping them to regain coverage or get care elsewhere. I applaud the efforts. But this is not a random design conducted with a blinded and disinterested group. And there’s no way these patients are representative of those in the Oregon study.

    It’s also a very small study. It was statistically significant because the difference seen was huge, and they didn’t control for any other factors. It was also after a 6 month period. So it’s worth looking at the follow-up study conducted at one year which managed to include between 81% and 86% of those in the first study. At this time point, the systolic BP had decreased 4 mm Hg from the 6-month point in the group that lost coverage. The systolic BP in the control group (which kept Medi-cal) had increased 3 mm Hg from the 6-month point. In other words, they saw a regression towards the mean, which is what you might see from a small study like this with surprisingly large initial results. So at one year, the difference between the two groups in systolic BP was 4 mm Hg (Table 2). Would it have continued to lessen by two years? We don’t know.

    Nonetheless, this is a very small study of a select and non-random group of patients. They saw that losing coverage was associated with (not caused) a large difference in systolic BP at six months that was much reduced (to about 4 mm Hg diastolic) by one year in a simple, uncontrolled analysis. These were patients really plugged into the system, who had physicians who were actively involved in their care and coverage. I’d expect a bigger difference from a sudden disruption in phenomenal care than from the other end, where you take a patient with no insurance and potentially less access and give them some. In other words, this study could be more appropriately used to describe the dangers of dropping people from the Medicaid rolls, than to describe the benefits of giving them new coverage. And I still wouldn’t say it’s powerful enough to be slam-dunk representative.

    I’m not the only one who thinks this way. These ideas were also included in a letter to the editor here.

    Moreover, all my previous arguments hold. We’re talking about patients with hypertension here. Not a much larger group of people, some of whom had hypertension.

    But still, this study is nothing like the RAND HIE or the Oregon Medicaid Study. I believe some of the authors were involved in both these papers and the RAND HIE. I bet they’d tell you that the RAND HIE is much, much more significant. And for my feelings on that, go back here.

    This study does not convince me that we should see large results from giving people with relatively well controlled blood pressure Medicaid in Oregon. Tell me why I’m wrong.

    @aaronecarroll

    Share
    Comments closed
     
    • I find it incredible that the authors of the Oregon study reported BP data for the population as a whole, rather than for those with hypertension. It leads me to wonder whether they had a political axe to grind, rather than a detached scientific perspective.

    • I think the practical relevance of this comparison is limited. In the 40 years since HIE we’ve had:

      1. Improved ability to control BP from many, many new medications.

      2. Very different financial implications of use due to dramatic shifts in pricing. They’re currently very cheap, likely improving the ability to control BP.

      3. Much greater treatment intensity (# of meds for the same disease burden), due to cultural shifts, better data, performance assessment and more. (Kaiser DOR has done some great work on this, but the effect is really massive, especially in the last 5-10 years.)

      4. We’ve learned a lot more about the general intractability of nonadherence. There are many factors in people not getting medications, so the implications of saving $4/month was never going to be 100% effective.

      I think what you’re driving at here – that a practical power calculation for Oregon is very difficult and involves more than just sampsi equations – is an important complement to the sampsi equations. I’m not sure that 1982 and 1971 are very telling. I wonder if some of the recent work on implementation of value-based purchasing might be more relevant?