Sunday, May 31, 2009

McAllen Texas and Healthcare Utilization: A Function of Statistical Variation, Not Poor Policy

How about McAllen Texas? Readers may recall that locale* was the topic of an essay in The New Yorker magazine by the impressive surgeon-essayist and Democratic advisor Atul Gawande. The McAllen 'hospital referral region' has the dubious distinction of having been identified by the Dartmouth Atlas as being close to the top nationwide (pretty cool graphic) for the amount of total fee-for-service Medicare dollars spent per beneficiary in 2006.

The DMCB’s first reaction was “so what?”

It took a second look when that hapless locale was seized upon by Peter Orszag of the Office of Management and Budget (OMB), Consumer Reports and some notable blogs as the symbol for all that ails American healthcare. Commentators are accusing the ‘McAllens of this country’ of consciously and unconsciously economically ripping off the system with precious little quality to show for it.

Yet, the dubious DMCB remains unexcited about Dr. Gawande’s faux discovery and disappointed that others haven’t considered the most likely cause of McAllen’s outlier status. Is there something really special about McAllen or is something else going on?

To illustrate the point, conduct a thought experiment by imaging many bags of pennies (to pay the primary care providers), quarters (for the rest of the physicians) and dollars (yes, they are available as metal coin) for the hospitals suspended above a huge map of the United States. Mentally open the bags, releasing coins that fall, clatter and roll across the map. Since the distribution of the coins is random, some areas of the U.S map will have no coins, others might have just pennies, others may have dollars and pennies, some will have all three and a few will have lots of all three. Somewhere on that map, however, there will be a pile of coins that is bigger than the rest. Perform the same experiment using Monopoly style plastic clinics and hospitals across a board-map and the same thing will happen. In these thought experiments, the area of the map with the extreme outlier status happens to be McAllen*.

It’s hard for non-statisticians/non-economists to think of human behavior in markets as being ‘randomly’ distributed around an average, but it’s true. The best (painfully so) examples of this are the performance of mutual funds, as well as what happens to losing vs. winning teams and their professional sports coaches . The same is true in healthcare: given the overall upward growth in the number of hospitals, specialists and clinics with an inevitable distribution (both high and low) around that trajectory, it is statistically inevitable that there will be a McAllen somewhere in the United States.

It is the nature of our minds to believe there must be something “causing” outliers. In other words, there must be something about McAllen that attracted all those coins, right? The DMCB, in reading Dr. Gawande’s article, thinks that may be true in Miami (which is number 1 in the U.S), but it doesn’t think that's the case for McAllen as described in the New Yorker magazine article. The gumshoe M.D. reporting clearly shows the McAllen providers are mystified by their status. It’s not as though they planned to take advantage of the system. In fact, they didn’t. That’s because it’s all random.

This is important because most healthcare providers involved in quality improvement learned long ago that ‘identifying’ and then ‘managing’ outliers with targeted interventions is a poor way to promote overall system improvement. Outliers naturally regress to the mean over time and they're not the problem anyway. Rather, the trick is to reduce overall variation around the mean (reducing the standard deviation) and to move all providers toward a better average level of behavior. That’s a lot of complicated work that, frankly, isn’t as enthralling to editors or the readers of The New Yorker. It's too much work.

While popular media can be forgiven for using simplistic descriptions of extreme outlier anedotes to pander to a political agenda, the DMCB isn’t too sure about Dr. Gawande. However, the DMCB is most frightened by potential reaction of the OMB. Short of complete central planning for the entire health care system, random distributions of performance, expense, quality, claims, satisfaction and countless other measures around a mean will be unavoidable. Of all persons, Dr. Orszag should understand that outliers are an ironic certainty, not evidence of malfeasance. Most are anomalies, not proof of anything. They are, in short, interesting, but not lessons and certainly not the stuff of policy making.

*correction: McAllen is not a county in Texas.


Brady Augustine said...

Yes, what you describe is the monumental shift from chopping off or explaining away outliers (i.e. QA) to focusing on "squeezing" and "shifting" the curve (i.e. QI).

I think it might be helpful to think of things less in the old experimental statistics way and more in the process control statistics way. For example, over a period of time, McAllen was just like El Paso in costs, demographics, and public health but somewhere special causes got introduced into their health system(e.g. doctor-owned hospital). [see wikipedia article at]. Thus, their high costs are a function of their own system's natural variation (common cause) and these changes to the system (special cause).

If we look at individual systems this way, negative outliers are informative in informing national QI efforts just like benchmarks (positive outliers). ~BAA

Anonymous said...

You blow it at "McAllen County." There is NO McAllen County.

Anonymous said...

One criticism of 'health care' charities in general: That industry has become one of the most profitable of all time. It already accounts for about 15% of our GDP. The industry as a whole can easily afford to cover its own research and development. Still, it lobbies for billions in government funding, tax breaks, and 'charitable' contributions. It affiliates with hundreds of public figures who 'raise funds' from ordinary people specifically for that industry in the name of 'humanity'. In other words, we are paying for a portion of their research and development. In return, they sell any 'breakthrough' made right back to us for MAXIMUM PROFIT. Their charges remain absolutely OBSCENE. They have been for years. So incredibly high, that thousands of families have already gone bankrupt as a direct result of health care expenses. Thousands of retirees have already had to 'reverse mortgage' their homes to pay for it. The average American is now losing sleep over health care expenses. Medicare and Medicaid are both projected to go bankrupt. Of course, the industry tries to cover for this injustice with one liners like "Today's drugs pay for tomorrow's miracles.". They also 'give back' a little just like every other industry and seek maximum publicity for it. Its a sham in my book. We don't need anymore 'good will' for or on the part of that industry. We need affordable health care in general. THAT MEANS LOWER PROFIT MARGINS. Along with fewer unnecessary tests, procedures, and pharmaceuticals. Of course, some of the work done is legitimate. But that holds true even for the government. Here is the problem. ITS GONE TOO FAR. Something must be done about this out of control 'drug and doctor' mentality. Otherwise, there will never, ever, EVER be affordable health care for the majority.

Jaan Sidorov said...

Oh my gosh. Anonymous is quite correct and the DMCB regrets the error. McAllen is a city in Texas and appears in the Dartmouth report as a 'hospital referral region.' The post has been corrected. Thanks so much for the heads up.

Jaan Sidorov said...

Brady makes several good points. It used to be that 'quality' meant finding the bad outliers and summarily executing them. In more modern quality management, outliers are an opportunity that involves first a) finding out if this is a statistical fluke and then b)assessing if there is really something driving the special status. A more nuanced interpretation of the Dartmouth data is that there is random fluctuation but and underlying driver MAY be present - like an investor-owned hospital.

Excellent points and worthy of their own post sometime. However, in reading the New Yorker article, Dr. Gawande did NOT find anything 'suspicious' going on that isn't also happening in locales everywhere. This led me to believe that McAllen is more fluke than fact.

The DMCB confesses to being weak in the science and the logic of quality improvement. That's another important point, here. At least it knows that it exists. It appears the Gawandes and Orszags of this world either don't know about it at all or are chosing to ignore it.

Jaan Sidorov said...

The DMCB mostly agrees with Angry Anonymous above. It pays over $1100 a month for health insurance and painful co-pays and deductables everytime it bumps into that money sucking machine called healthcare.

Everyone agrees that this has to end. Honorable people, however, are struggling with how to get from here to there. Most persons seem to think the answer lies in the right mix of 'market forces' and 'public interventions' that force consumers and producers to act in a more restrained way. What do you suggest?

Two other points:

The DMCB isn't too sure about the role of pharma. THAT is the topic of a whole other blog. However, while the cost of drugs is often breathtaking, a strict analysis of cost vs. quality of life years gained typically shows far greater value in pills than the routine stuff doctors 'do' to patients.

When it comes to government largesse, if you build the trough, they will come. Chances are your own Congressman is very proud of the money he or she has brought in to build that new research wing in your district. The DMCB is registered to vote and has told its elected representative that the approach is all wrong. You should too - in addition to continuing to post comments on this blog!