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

    So some people have asked why I have focused on diabetes so much, and not blood pressure or Framingham composite results, when looking at the Oregon Medicaid study. There are a few reasons. First, I’m a pedatrician. I know much less about hypertension, as I don’t see much of it nor do I do any research on it. Second, I do research in diabetes, and know much more about that. Third, I think diabetes is a looming problem with more short term concerns. Finally, I can’t remember the last time I used Framingham stuff (other than to advise friends), and I try not to wax philosophic on things unless I feel like I know enough about them.

    But it’s a fair point, and so I decided to look into the blood pressure results further. They don’t make me feel much better.

    Let’s start with what the Oregon Medicaid study reported recently. It’s in their Table 2. They found that in the control group, the systolic BP was 119.3 mm Hg. Medicaid changed it by -0.52 mm Hg. That wasn’t statistically significant. But let’s acknowledge that this result is for all people in the study, with and without high blood pressure (a systolic BP of 140 mm Hg or higher). Obviously, most people didn’t have it. The average systolic BP in the study was under 120 mm Hg, which is normal. so I’m not sure how much you could expect the (normal) average BP of this group to go down.

    There seem to be two main sources of “expectations”. The first are two non-RCT papers from Lurie et al, published in the mid-1980s following people who were kicked off Medi-Cal. I’m told they saw a big jump in blood pressure, but even my institutional access won’t grant me permission to see those papers online. I’ll keep trying. If someone wants to send them to me, I’d be thrilled.

    But I did find the RAND HIE paper on BP. That study found that for people with hypertension, giving people free care resulted in a change of systolic BP of -1.9 mm Hg. It’s easy to find this result. It’s right there in their Table 2.  Then you can go to Table 3. Among low-income people with hypertension (which I’ll concede is arguably more comparable to Medicaid populations), the change was -3.5 mm Hg.

    But those results were for people with hypertension. There were 856 of them in the study. The average systolic BP in the free care hypertensive group was 137.1 mm Hg. So there was lots of room forimprovement.

    Table 2 in the Oregon study, on the other hand, is all participants. Not hypertensives. You can tell that from the average systolic BP (119.3 mm Hg). I don’t see how you can apply the results from the RAND HIE (-3.5 mm Hg) to that group, because that group is not all hypertensives.

    I’ve been told to look at sub-analyses. So be it. I’ve scoured the Appendix. Table S14a is a subset of those in the Oregon study who were age 50-64. They were less healthy, in that their average control group systolic BP was 127.3 instead of 119.3. And their systolic BP changed -2.5 mm Hg. Still not significant, of course, but much closer to the RAND HIE effect. And again, their average systolic BP was 127.3, which is still much less hypertensive than those in the RAND HIE.

    The most on-point subanalysis I can find is Table S14c. That’s the pre-randomization diagnosis grouping, and it includes 2225 people who had a diagnosis of hypertension before the study began. But there are a few things to consider here. First, these are people who may have been more likely to be plugged into the system already because they had diagnoses. So they might see less of an effect by getting Medicaid. Second, their control BP was 129.8 mm Hg, still much lower than the hypertensive subjects in the RAND HIE. In fact, only 38% of them had an elevated blood pressure. But there’s more. There were only 2225 people with a diagnosis of hypertension pre-randomization. That’s 2225 of the 74,922 who were randomized. That’s 3% of the sample. If the percentages hold (and they should if it’s random), then they accounted for only 3% of the 1903 OHP-enrolled survey responders in the intervention group.

    How many is that? 56 or 57 people. Of whom, only 38% had an elevated blood pressure. So how much effect could Medicaid have?

    The more I look into this, the more baffled I get. Here’s how the RAND HIE defined people with hypertension:

    Thus, participants were called “hypertensive” at enrollment if they (1) reported taking antihypertensive drugs, (2) had a repeated systolic  blood pressure greater than or equal to 160 mm Hg or diastolic blood pressure greater than or equal to 95 mm Hg at the examination, (3) had a repeated systolic blood pressure greater than or equal to 140 mm Hg or diastolic blood pressure greater than or equal to 90 mm Hg and reported that their physician had previously told them they had hypertension, or (4) reported that a physician had told them more than once they had hypertension and either were assigned to miss the examination or had systolic blood pressure greater than or equal to 130 mm Hg or diastolic blood pressure greater than or equal to 80 mm Hg. Others were called “hypertensive” at exit from the HIE if they met criteria 1, 2, or 3 at exit or if (5) they had both repeated enrollment and exit systolic blood pressure greater than or equal to 140 mm Hg or diastolic blood pressure greater than or equal to 90 mm Hg or (6) a physician had reported hypertension on an insurance claim form and the participants reported they had been diagnosed as hypertensive, or the physician had reported hypertension on two or more insurance claim forms.

    Those people had hypertension. There were more than 850 of them in the study. They had average systolic BPs near 140 mm Hg. And they found that in the low income sub-group, giving free care resulted in a change of -3.5 mm Hg.

    No matter how I slice it, I can’t generate the same enthusiasm for the Oregon study. The study had fewer people, and less power. The study doesn’t seem able to define or quantify the hypertensive group in a similar way. The study seems to feel that it should be able to see a drop in systolic blood pressure that is higher than what was seen in the RAND HIE. And the study seems to think that we should be able to see those drops in groups with starting systolic blood pressures that are much closer to normal than those seen in the RAND HIE.

    It’s totally possible I’m wrong. If I am, I will post it right here. Tell me what I’m missing. Tell me why when we start with a population of people with much better control of their blood pressure, we should expect a larger absolute drop in systolic BP than the RAND HIE did. And tell me why this study should be powered to detect that difference with what seems like far fewer participants with hypertension.

    I’m still going to look for the Lurie studies. But it seems that the RAND HIE, which is the only other real RCT here, should be most on point.


    • Aaron,
      I’m 66 and the Doc said I have hypertension. He recommended meds. I said I would try a natural way. I got a cuff blood pressure device. There are lots of ways to change your blood pressure and it changes by a lot within minutes of readings. When I see statistics from people getting one or two readings I feel it’s almost valueless.
      It would be difficult to monitor B P readings without taking more readings and taking all the variables into consideration.
      For me the biggest changes come from
      – fast walking
      – deep breathing
      – eliminate caffein
      The difference can be anywhere from 1 to 30 mg BP. Systolic.

    • I’m sympathetic to the arguments about statistical power, but one point worth addressing is how appropriate comparisons between RAND and Oregon point estimates should be given that RAND began in 1971 and we’ve had 40 years (40 years!) of improved medical care and new drugs since then.

      • That would be a more compelling argument if it turned out that ALLHAT hadn’t proved that some of the oldest and cheapest drugs are still the best.

      • The goals of blood pressure control are much more strict today (lower) and despite the fact that many older drugs are excellent the newer drugs have permitted better control of blood pressure permitting people to reach optimal blood pressure levels. Thus it is quite likely that Aaron’s argument is very compelling.

        Of course when Hillary Clinton was pointing out how expensive medications were your argument was correctly used against her. She claimed outrageous costs for medications for relatively common diseases, but the vast majority were actually able to be managed at a small fraction of the costs she used as examples because she was using the list prices of the most expensive new medications.

    • Something I wondered about when I first read this Oregon medicaid study was what kind of education and follow up did the patients receive? As a person who lost her brother because he had high blood pressure when he was relatively young, I am very interested in this aspect of the study.
      My brother, an intelligent person, did not understand the long-term complications and things that could happen to him if he didn’t get regular treatment for his high blood pressure.
      The medication isn’t expensive, but getting a prescription can be, for an unemployed person who didn’t qualify for medical coverage back in the 1980’s.

    • You hit the nail on the head. The 12-thousand-patient sample size is misleading when talking about improvements A1c or BP — the true sample size for DM or HTN control is the group who already has either, which is a tiny fraction of each. We expect the HgbA1c to stay the same for the 95% of people who don’t have diabetes. Plus, 2 years is a pretty short time frame for big improvements in either of these, particularly given that the intervention isn’t a specific treatment but Medicaid, which is potential access to a physician who then will give treatment…