Individualized Evidence Based Medicine

Today’s Managing Health Care Costs Indicator is  $340,000

The early July Annals of Internal Medicine  has a fascinating simulation which suggests that there isn’t a single “rule” about who should get mammograms.  This isn’t surprising; we’ve been thinking about individualized medicine for some time. But this throws into doubt much of the quality reporting we’ve been putting in place over the last 20 years – we’re going to have to go way beyond the percentage of women between 50 and 65 who got a mammogram each year to determine if a provider group is delivering  the best possible care!

For starters, this simulation shows that annual mammography is so costly that it costs more than $340,000 per quality adjusted life year to perform annual mammography instead of biennial for all women at any level of risk.  Annual mammograms would not be recommended for any women by value-based purchasing guidelines.

Click to enlarge.  Source 

But this is not all about costs –the rate of false positives is strikingly high.  And false positives don’t just cost money – they take an enormous emotional toll on women and their families.  This simulation considered the potential negative quality of life for the brief period of time between a false positive screening mammogram and a negative biopsy – and even if that is very small, it still has a big impact on the “value” of mammographic screening.

Click to enlarge. Density by BI-RAD methodology (see article for details)

The value of mammography is highly dependent upon two well-accepted factors:  family history and previous biopsy. Breast density is such an important risk factor for breast cancer that this alone could be a reason to recommend different screening intervals.

This is a simulation study – performed with robust sensitivity analysis – and the accompanying editorial warns that we should not change our current mammography practice based on this paper alone. The simulation assumed we would do a screening mammogram to determine breast density at the beginning of each 10 year period, and required a host of assumptions any of which could be disputed.

This paper elegantly demonstrates that we need to start thinking about evidence based therapy based on highly individualized considerations. These considerations will involve genetics, previous medical history, and individual patient preference. It’s going to be harder to effectively develop evidence-based insurance plans – since what is medically necessary for me is different than what is medically necessary for you based on criteria not likely to be found in claims  It will also be harder to effectively practice medicine without decision algorithms built into electronic medical records. Neither a simple rule nor a physician’s intuition will be adequate to give us the best medical care.

Our quest for certainty makes health care expensive AND exposes us to excess risk.

We intuitively believe that medical diagnostic tests are more likely to banish uncertainty than they really are.

Here’s an example. Many executives have an “executive physical” that includes an exercise stress test to be sure they don’t have coronary artery disease, and many patients pay to have a CT scan to assess coronary artery calcification.

Take a 45 year old executive with a normal blood pressure and a normal cholesterol – his “pretest” probability of having a heart attack or cardiac death in the next 10 years is about 1%.  Here is a link that lets you put in age, gender, blood pressure and cholesterol and calculates risk based on data from the Framingham data.

Imagine 1000 such executives, each with a 1% chance of coronary disease. That means that ten of them will have a heart attack or cardiac death in the next ten years.   If they all had exercise stress tests, with a 70% sensitivity (chance a person with disease will have a positive) and a 90% specificity (chance a positive is a “true” positive” we would have 107 positives – but only 7 of them would be correct.  The overwhelming majority of positive tests would be false positives.   There would be almost 900 negatives, and of these only 3 would be “false” negatives.  

      

This calculation of “posterior probability” is called Bayes Theorem

So – even with a positive test a patient has a low chance of serious cardiac event.  A negative test is much more reliable. However, the executive had a 1% chance of heart disease before the test, while after the test the probability remains about 1/3 as high. 

So – we have more data – but not much more information.  Here is a graphic way to look at this:

Worst of all, there is not evidence that people with no symptoms benefit from invasive therapy to correct cardiac disease.  Even the “lucky” executive whose hidden cardiac disease is discovered through this testing might not be so lucky either!

We desire more information – and insist on more diagnostic tests in a vain effort to banish uncertainty.  Alas, tests done under the wrong circumstances don’t do much to diminish uncertainty.  We want exercise stress tests, mammograms and prostate specific antigen tests, even when our risk of disease is low.  The “cascade” of follow-up tests costs substantial sums (the point of this blog).   This cascade also causes discomfort and anxiety –and sometimes real damage to patients.  

.

Cheap But "Good Enough." Can We Accept Decrementally Cost Effective Interventions?

We can all easily agree that if something increases cost and offers worse outcomes (red in diagram), we shouldn't do it. That's why prescriptions for Vioxx and similar antiinflammatories cratered after reports emerged of higher cardiac mortality for these high-priced medicines that were only as effective as Advil.   We can also all agree that if something decreases cost and increases quality - we should do it (Green in chart)

The challenge is with the blue slice of these pie --medical services that increase quality (a little) and increase price (a lot).  That's true of a majority of novel interventions tested in the literature - they displace less expensive standbys, but offer some new advantage.  It's tough to say no for higher quality.

Another challenge, that is ill-studied, is an intervention that is less expensive, and not QUITE as good -but really "good enough," at least for most people (yellow in this diagram).  These decisions are clinically (and politically) difficult.

The early November Annals of Internal Medicine has a literature review on “decremental cost effectiveness” entitled “Much Cheaper, Almost as Good.”  Authors Nelson et al from Tufts Medical Center point out that in other industries we value products and services that are “good enough,” but cost much less. Examples include personal computers (not as good as mainframes), IKEA furniture (painful to assemble, but much less expensive), and the Nano automobile from Tata Motors (sacrificing comfort and safety but costing only about $2,000).  However, in health care, we only value the highest quality  - regardless of the price.  A good example of this was an NPR report this evening about MRIs.  They often cost $2000 in the US - and we have the best machines in the world. In Japan, the government-enforced price of an MRI is $120.  Is the best image worth this price differential?

The authors in the Annals article found 2128 cost effectiveness ratios in the medical literature between 2002-2007 – and only 33 involved sacrificing any quality for price. In general, these studies involved the sacrifice of between 8 hours and 1 week of quality adjusted life (0.001 – 0.021 QALYs), and the savings were beween $122 and almost $12,000 per patient.  (Of the studies they reviewed, only 9 were of high enough quality to fully evaluate).

Of course, in the majority of instances where there was an innovation that represented increased costs and increased quality, the “usual care” before the innovation might have been “decrementally cost effective.”

This is relevant to the firestorm over the US Preventive Services Task Force finding that evidence doesn't support recommending annual mammography for women under 50 (and only supports biannual mammography from 50-69).  This recommendation was NOT made to save money - but much of the vehement disagreement is focused on whether lives would be sacrificed to save dollars.  What if the number of "quality adjusted life minutes" saved by mammograms is less than the number of minutes women would spend getting additional mammography and having followup biopsies? Listening to some of the vitriol about this finding on the radio reminded me that we don't like to make tradeoffs, especially in decisions around our health.

We are unwilling to "settle" for a decrementally cost effective treatment for ourselves and our families and friends.  Further, the highest margins are often associated with newer, discretionary technology - so the medical system often does not discourage overuse.  This is the conundrum we face that makes technology continue to ratchet up the cost of health care.