On Getting Burned by Population Health Thinking: (Why has a Decade of Aggressive Intervention not Prevented More Deaths From Brain Aneurysms?)

In our house we take a lot of fizzy vitamin C to prevent colds in the winter. As doctors, my wife and I both know that science doesn’t support our decision… We take it anyway.


We shouldn’t. After all, the well respected Cochrane group (which grades and distills multiple studies into a single “meta-analysis”) clearly finds that vitamin C supplementation has no effect on preventing colds. We also know about the very modest side effects of too much vitamin C: how you can get kidney stones and diarrhea and how one woman lost her (albeit transplanted) kidneys from taking too much of the stuff. So in the absence of an upside, and in the presence of information to suggest potential harm, we take the tablets anyway. Daily. In bulk. From Whole Foods.

Why? I think that, honestly, we’d feel stupid if we didn’t take it and then came down with a runny nose and cough. As much as the science discounts benefit to the population, we have a personal belief that the stuff works. We don’t want to be one of those outliers who might have benefitted from the drug but didn’t take it.

We’re not the only ones:

A neurosurgeon friend recently sent me a great study recently published in the respected journal, “Stroke”. The paper, written by authors from Columbia, looked at nationwide Medicare data to describe trends in the use of aneurysm surgery in the elderly. What they revealed was fascinating, both from a population health perspective, but also because it tells us a lot about how we make medical decisions, both as consumers and as physicians.

Here’s the background:

A certain fraction of the population (about 3%) has a cerebral aneurysm, or a “bulging” of the blood vessels in the brain. These are usually asymptomatic and are picked up on CT or MRI scans of the brain done for other reasons. A certain fraction of these aneurysms go on to burst, leading to a dangerous and sometimes fatal hemorrhagic stroke (subarachnoid hemorrhage). The risk of rupture varies, but is estimated to be around 0.5% to 1% per year depending on the size of the aneurysm.

Over the years, neurosurgeons have intervened on these asymptomatic lesions by performing one of two procedures: either they surgically insert a clip at the neck of the aneurysm, or else they increasingly use minimally invasive techniques to route a coil of metal from an artery in the groin into the aneurysmal pouch in the brain, occluding it. The surgeries are both performed with the goal of preventing downstream rupture of the aneurysm.

The decision to intervene on the lesion like this one is hard to do without considering population level numbers and outcomes. The older that a patient is when an aneurysm is discovered, the lower the likelihood that the patient will die from this aneurysm (given the risk of rupture is cumulative and < 1% annually).

Surgeons also have to balance the real risk of causing significant harm to the patient: the most current data show that 30-day mortality in the clipping population was 1.6% (25% of patients had complications and 44% were sent to long-term care facilities after discharge). In the coiling population, mortality was also 1.5% (13.5% of patients had in-hospital complications but unlike coiling 81% of patients were discharged home after the coil).

Clearly, neither clipping or coiling are benign procedures. But coiling is easier to sell: it doesn’t require open-skull surgery and seems less dangerous then clipping. Coiling can also be done by a variety of medical specialists (interventional radiologists, interventional neurologists and the like) which means that the number of doctors willing and able to intervene increased through the early 2000’s.

Screen Shot 2015-08-23 at 8.55.37 AM

What the country saw over the year 2000’s was a dramatic increase in the number of aneurysm interventions. Coiling led the growth in interventions from 1.4/1000 patients in 2000 to above 6.0/1000 patients by 2010.

It also turns out that the >75 years age cohort was responsible for the largest increase in coiling volume. Older patients increasingly got aggressive interventions designed to prevent aneurysmal hemorrhage even though their liklihood of rupturing before they died of other causes was low.

Part of this increase was due to the growth in the number of doctors able to treat these aneurysms. Part was the fact that coiling proved to be safer then clipping, and easier to sell to both patients and doctors. Perhaps most important, intervention seemed common-sense and as the procedures grew safer, the 1% rate of annual catastrophe began to look comparatively more sinister.

Now here’s the kicker: Over a decade when interventions increased four-fold, the rate of subarachnoid hemorrhage didn’t actually go down.

Screen Shot 2015-08-23 at 9.29.32 AM

The Stroke data suggests something worse than the obvious issue of whether the morbidity/cost of intervening on an aneurysm outweighs a <1% annual benefit (particularly in the elderly).

In a chart that accompanies the Stroke article, authors showed that while the amount of clipping and coiling going on in America increased dramatically from 2000 to 2010, the number of hemorrhages didn’t actually fall at all. In fact, they inexplicably rose by the end of the decade.

The authors of this brilliant study concluded:


[I]n spite of the dramatic increase in the rates of procedures, particularly coiling, performed on Medicare beneficiaries with UIAs, [unruptured inter cranial aneurysms] the overall rate of SAH [subarachnoid hemorrhage] among Medicare beneficiaries did not decrease. This is the opposite of what would be expected if procedural treatment of UIAs was preventing aneurysm ruptures that would have occurred without treatment.

Their takeaway:

“…it is reasonable to suspect that for some patients in this age group, the value of clipping or coiling their [unruptured aneurysms] is either modest or nonexistent… particularly in light of procedural complications and the effect of procedure-associated morbidity and hospitalizations on quality of life.

It’s easy to challenge a study like the one in Stroke, and undoubedly the issue has yet to be definitively resolved. Observational studies are interesting, but not definitive and a proper head-to-head trial of intervention versus watchful waiting seems in order.

Medicine, at the end of the day, is all about balancing potential good and potential harm. The problem is often that it’s impossible to get a sense for either extreme until you can study a large population of people. The nuance to this otherwise simple calculus is that every doctor will tell you that it seems far easier to rationalize a complication resulting from action intended to prevent a greater harm, than it is to than it is to justify passivity and watch as someone dies of something preventable. 

I know this first-hand: a few years ago I diagnosed metastatic breast cancer in “Maria”, a 29-year-old immigrant woman. She had arrived at the emergency room of the university hospital where I worked short of breath and we soon found a large mass in her swollen, red breast and spots in her lungs on x-ray.

As we were making plans to admit her, Maria told me that her primary care physician– quite appropriately– had previously told her that she was too young to benefit from screening mammograms according to national guidelines. The guidelines, published by the US Preventive Service Task Force, and others, generally recommend that young women not receive a mammogram because the harms from working up incidental findings outweigh the low liklihood of finding real cancer.

For these kinds of population-level tradeoffs to work — for thousands of people to benefit from the avoidance of unnecessary care — there needs to be some small number of unlucky souls who would have benefitted from the treatment or workup but didn’t get it. Statistically, it has to work that way.

Nobody wants to be that unlucky soul, that Maria. And, for our part, physicians are hardwired to over-value the impact of missing a rare case of cancer while under-valuing the cummulative benefits of avoided testing. It’s primal.

This is why recent campaigns to deemphasize prostate screening antigen testing, and routine mammography in young women have been such hard sells. In both of these cases, population health studies are conclusive: on a aggregate basis, the side effects of testing outweigh the benefits of catching the rare cancer. But, tell that to the patient whose cancer you missed. It feels better to do something, particularly when the “doing” gets safer. 

Even though I trained as an epidemiologist and have spent my career trying to rationalize healthcare, I can tell you that if push came to shove and I had an aneurysm, coiling would probably seem pretty attractive. Nobody wants to find themselves in Maria’s shoes. The population health specialist in me would consider it a wasteful and illogical decision. The vitamin taker in me would certainly understand.


Images: Apotek Hjartat via Flikr cc search.  Charts from Dartmouth Atlas and Stroke