Stat counter

View My Stats

Thursday, January 30, 2014

Subsidize something and drive others to consume more of it...duh

Oregon's deployment of a lottery to allocate scarce health care resources has provided us additional insights into human behavior. For those of you who might not be aware of this interesting pre-ACA interlude in health care policy, in 2008 Oregon expanded Medicaid on a limited basis, granting Medicaid coverage to 30K of a 90K pool of those interested and they did so by lottery. How they got this past their human subjects board is beyond me. I get push back when I want to administer surveys. I guess one can experiment in the policy world more readily than the medical world. One not need to dwell on the possibility of harm as long as your intentions are noble.

One of the major rationales for Medicaid expansion has been to relieve Emergency Departments of the crush of the uninsured and presumably if a major portion of the uninsured were covered by Medicaid, they would not frequent the ED, or at least that was hope, assumption, and selling point. No expected that this hypothesis to actually be tested.  It is always most convenient to be able to declare such items as fact by consensus, particularly when it has a useful political purpose. How could you ever test this?

Enter the Oregon random allocation scheme and viola, a randomized control trial with huge numbers and the tracking systems in place. The results of this "experiment" were published in a not too shabby journal Science in January of 2014 and the results are pretty unambiguous. ED visits in the covered population were increased by 40%. The authors looked at all sorts of subgroup analysis and what they found was no significant decline in ANY subgroup and increased in all subgroups examined except one.

They further subdivided the types of visits looking at visits which ranged from ones that only could be handled via the ED to various gradations to ones that absolutely were not emergencies and should have been handled in a routine ambulatory setting. There were increases in all types of visits except true emergencies.

The most obvious lesson from these findings is that our predictions as to how our policy interventions will end up "working" should be viewed with great skepticism. Heath care delivery is accomplished by a complex system with many different actors (none of them angels), unique circumstances, context specific motivations and incentives, all interlinked. Push here and watch where things bulge elsewhere. So much for relieving the crush on ED's. I am sure I will still hear bold claims that states not expanding Medicaid will create undue stresses on ED's. Never let actual data get in the way of passionate claims.

The accompanying commentary to the article also had some interesting insights. The first was a statement which said:
 " From the patients perspective, insurance is essentially an across-the-board price cut that makes all health services cheaper. The direct effect is, naturally, more consumption of health of all sorts - primary care, ER, and everything else".

Talk about stating the obvious, but perhaps it is just the obvious that needs to be stressed. The holy grail of free market based economic systems is the drive down costs and make better and unique services available to ever larger segments of the population. One can use subsidies to make things more affordable to specific segments of the population but this does really drive down costs in the long term. Subsidies simply hide costs from people and may actually make subsidized products and services more expensive. Look at health care and higher education. Huge subsidies, poor productivity gains, and rising costs characterize these industries.

The subsidized ED visit for that newly minted Medicaid recipient is just as expensive to deliver as the visit delivered to to the previous customer. That it has become more affordable to one end user provides little incentive to figure out how to deliver the needed service in a more cost effective way to a larger audience.

The second comment in the accompanying editorial provided an potential explanation for the unexpected  increased visits. Perhaps they should not have been so unexpected.  One of the reasons for the increased ED visits can possibly be linked to primary care physician behavior. As it turns out the primary docs are strapped and despite the newly minted coverage, there is no room on the PMD schedules, particularly to handle unscheduled and complicated patients. When faced with this, their responses were to tell their patients to go to the ED. One can view this a a rational decision given their resources. Their offices are functioning at capacity and have no surge capacity built in.

The lesson here is the problem is not the insurance. All insurance is in these cases is a price cut for the consumer without sending any sort of concordant price signal to suppliers. If fact, setting prices too low sends a signal to producers of particular services that what they produce is of little value. The rational response is to make less of what has little value. The problem is the value setters are not those who consume the services. The results of price mismatches are shortages. There is no reason to assume that what we may desire should or will be available when we desire it. However, many of us  have lived through a remarkable era of human history where we have become unaccustomed to any sort of scarcity. Everything we want is available virtually when we want it.

I have news for readers. That situation has existed in this brief moment in human history in a narrow portion of the world because the remarkable growth of wealth has occurred in areas using a price coordinated economy,  which in other words is a market based economy. Through the miracles of market set prices, signals travel at the speed of light to goods and products providers throughout the world who can see where there are opportunities to "make a buck". We are awash in goods and services because parties who best meet are needs and wants are rewarded handsomely. Everywhere this system was trashed you end up with some command and control totalitarian  nightmare like North Korea, Zimbabwe, and Venezuela.

We are experiencing more and more pain, more and more dysfunction, more and more scarcity in health care because we have all but removed the price coordination mechanism, attempting to create a better health care delivery system by trying to use the lessons of Eastern European command and control economies. When we see price signals that tell us that services are too expensive, we use subsidies to mask the signals and then are surprised by the unintended consequences of the subsidies. Predictably we never get not what was expected but some other less desirable outcome. Don't we just love surprises!

The way to get past scarcity is to incentive productivity which will drive down the actual costs, not the artificially set prices and subsidies using the insurance vehicle.

Lessons from being frozen in, Atlanta style

We are thawing out from a winter storm debacle here in Atlanta. There is no question it was a truly awful experience for a number of people who were trapped on the highways for hours on end. On the bright side, I am not aware of any weather related deaths.

The finger pointing as to blame for this has been going on. The Mayor of Atlanta, the Governor of the State, and the heads of various agencies were all targets of the angry press. How we could learn so little from previous similar episodes and allow this to happen yet again. It was only three short years ago when Atlanta was buried for an entire week. I can also recall multiple other episodes in the past when ice or snow laid waste to our ability to function. Northern cities can deal with this issue. Why can't we?

I always like to tell my trainees that the key to management of patients and their diseases is expectation management. Remember, happiness = results - expectations. I think the public expectations for dealing with snow and ice for a southern city which gets a significant snow and ice event on average every five years should not be that it should be able to handle this like Buffalo New York. Northern cities invest huge sums of money in equipment, training, and people to deal with what is a common occurrence. For Atlanta to make a similar investment would be one  that would have a minimal ROI.

I listened to various leaders discussing their efforts after the fact. The city of Atlanta actually did a reasonable job of dealing with surface streets which may have actually ultimately worsened the gridlock (see below). The state DOT which was responsible for clearing the interstates made a strategic but understandable error. They pre-stationed equipment where they thought the storm was headed, south of the city. The computer models were unreliable since the conditions prior to the storm were very unique, not being seen since the advent of computer modeling of weather and the predictions from the National Weather Service were off. Yes some forecasters got it right but how was one to know they were the correct ones before the fact.

The storm hit very quickly and when EVERYONE left work at the same time, the clear city streets allowed them to rapidly get on to the interstate highways, creating gridlock so rapidly that the DOT equipment, which was not optimally placed could not be effective. The well salted surface streets likely facilitated rapid access to the interstates and perhaps worsened gridlock on these unsalted streets.

What can be learned from this? Complex systems are complex and unpredictable. I suspect that for the next 10 potential events, we will be closing schools and offices early, cancelling events, and have a low threshold for action. That will continue until there is a hue and cry that we are over reacting and being silly at which point we will revert to a less cautious behavior. The institutional memories will dissipate and we will have yet another snow and ice disaster followed by renewed finger pointing.

Good systems are compatible with bad outcomes. As we become more and more dependent upon technology and the ability to travel and have less and less reserve built into our systems, we become more prone to huge disruptions from unpredictable events like weather events.  It is simply bad policy for the public to demand their public money be spent on bad investments.  Atlanta could handle these events if they invested in fleets of snow plows which could sit idle for years on end. Because it makes no financial sense to invest heavily in the infrastructure to deploy low frequency contingency events (which are almost infinite in variety), we should expect that we are going to have some bad days. Not even big government (and perhaps especially big government) can stop us from having bad days.

Sunday, January 5, 2014

What the public (and many doctors) don't understand about cancer

Gilbert Welsh has published another commentary in the New York Times (Welch NYT Dec 29, 2013). What he writes in the piece is interesting as usual, although not especially novel. His comments were based upon the results of a study published in JAMA Internal Medicine in the previous week. He wrote a similar commentary piece published as a companion piece in the JAMA Internal Medicine journal. He has presented his arguments skeptical of cancer screening before but his message is very well worth restating. Because his reasoning is counter-intuitive, it takes a while for the message to sink in and, for many people, it simply is met with blank stares. Some view this with great hostility. Check out the comments. They are perhaps as interesting as the article.

Why is that? I think the reason is that there is emotional baggage associated with the word "Cancer" (well dah!). In addition and perhaps more importantly there is a gross misunderstanding of how cancer is diagnosed and the limits of pathology as a tool to predict biological behavior. This issue needs to be pushed front and center into the debate because  the response to divergent responses to his message come down to reasonable people hearing different things when they hear the word "CANCER".

The origins of pathological (histological) diagnosis of disease date back to the mid-to late 19th century. The German chemical industry developed out of the demand for new dyes for cloth. It created vast libraries of chemical compounds which resourceful entrepreneurs sought new applications in other industries. One set of applications was found in the developing pharmaceutical business. Many of even our current drugs are based upon dye backbones.

A second set of applications was found in laboratory medicine, initially at the autopsy table. Tissue was removed from corpses, fixed in formaldehyde, embedded in wax, sectioned and placed on a slide, and then stained with a variety of dyes borrowed from the textile business. An empiric approach led to the foundations of contemporary histology. With the advent of anesthesia, samples of tissue could more readily be taken from live human beings, processed as cadaveric tissue had been processed, and examined under the microscope.

The power of the pathologic exam of tissue became apparent because it could be validated at the autopsy table. The microscopic appearance of a large mass (for example a breast mass) biopsied from a live person may bear a stark resemblance to the microscopic appearance of tissue obtained from another at autopsy. It is not a huge leap to predict that one living with such a mass will suffer the fate of those already on the autopsy table. In contrast, if the pathology of the large mass showed something distinctly different, such as an abscess, one could predict that they would experience a different fate. At this point pathological diagnosis was pretty black and white.

Fast forward a century and the tool that is used to predict patient's fate is essentially unchanged. We are still placing pieces of tissue into fixative, embedding in wax, staining with textile dyes, and looking under the microscope. However, what we look at has changed dramatically. We no longer look at large pieces of tissue taken from people with advanced disease but instead look at tiny slivers taken from patients who are essentially well. Because this transition tool place over many decades, we have lost sight of the gradual changes of the inputs and the contextual change. We no longer have the autopsy to validate whether the increasingly subtle changes observed by pathologists have any biological significance. Most patients diagnosed with cancer currently do not die of their cancer.The 5 year survival rates of breast and prostate cancer are 89% and 99% respectively according to the SEER numbers.

I pose the question: why should we be able to predict the future by taking human tissue, fixing it in formaldehyde, embedding it in wax, placing them on slides, staining with clothing dye, and looking at them under the microscope? I have no doubt that within certain clinical contexts this technique has value but as one moves away from advanced tumors, when does this technique lose value? The value of the diagnosis is that it can predict who will be done in by their disease and identify who will benefit from some sort of intervention to prolong and/or improve their lives.

This is not some trivial question of interest only to the high priests of academic medicine. Read the comments to Gilbert Welch's article and our understanding of cancer and  the ability of pathologists to make these diagnoses and make reasonably accurate predictions rests at the center of this problem. No one appears to want to state the obvious. The tools used to make these predictions is simply not up to the task. In particular the gold standard for looking at those screened and identifying real disease with a high likelihood of killing those affected is using histologic tools - anatomic pathology. Just how good or bad are these tools at giving us the information we desire?

It appears they are not so robust and perhaps when applied in certain circumstances border on simply terrible. Try to use them in neonatal screening contexts and you end up with epidemics of neuroblastomas of no clinical significance. I could go on and on. The public has no clue and much of the medical community is equally as blind to the limits of this tool. Anatomic pathology operates like the Oz in the Wizard of Oz.
"I am Oz, the Great and Terrible. Why do you seek me?"
They looked again in every part of the room, and then, seeing no one, Dorothy asked, "Where are you?"  
"I am everywhere," answered the Voice, "but to the eyes of common mortals I am invisible. I will now seat myself upon my throne, that you may converse with me." Indeed, the Voice seemed just then to come straight from the throne itself; so they walked toward it and stood in a row while Dorothy said: "We have come to claim our promise, O Oz."
"What promise?" asked Oz.
What is the promise of the pathologist? If is the promise to be able to look at a specimen, apply a test,  and tell  their clients whether they face a terrible fate. It has been said that extraordinary claims should require more stringent levels of proof and what claim is more extraordinary than a claim that a healthy and well person is going to die. Remember it is simply a test and like all tests it has inherent problems with specificity and sensitivity, all of which is context specific. Anatomic pathology is completely subjective and requires the operator to define thresholds. The test is only as good as its ability to provide useful information.

What useful information is provided when a specimen is sent in to rule out a malignancy? When the term cancer is invoked by a pathologist, the general public sees this as a prediction of death. However to the pathologist the term cancer can be invoked when there is little or no expectation of malignant behavior. How useful is that to use a term implying aggressive biological behavior for classes of growths than are essentially benign? We end up using tortured terms such as non-progressive cancers or non-malignant malignancies. This is like describing a cold hot summer day, a rock hard soft pillow, a short tall man. or a skinny obese person.

The term cancer needs to be redefined or at least its meaning needs to be clarified. I propose that if this used as a diagnostic term by a physician, it must be attached to some sort a prediction of the likelihood of death if left untreated. If a 60 year old woman is found to have DCIS after screening mammography, what are the odds of her dying of breast cancer if left untreated? I suspect that physicians, especially pathologists, would be much less willing to use the term cancer if they were required to attach a relatively precise prediction to it. As long as they can remain behind the curtain like the great Oz they can label people as having cancer without having to explain to them what the meaning of this label is. As long as they can use the term in a clinical vacuum, confusion will reign. It is reasonable for patients to insist that all cancer be treated if they assume that all cancers are malignant. One perhaps make the case that aggressive treatment is warranted if 90% are aggressive would go on to kill the host.What about 50%? 10%? 1%?

What does this confusion look like? Read the comments on Dr. Welch's article. The numbers are stark that he presents and provide little evidence that current cancer surveillance practices do what they intend to accomplish for the overwhelming majority of those involved. However, the cancer label invokes such a visceral response that these numbers are meaningless. Once the term cancer is used all nuance tends to disappear. What doctor in their right mind would recommend leaving cancer untreated?

I suspect that if patients knew how crude anatomic pathological tests are and how meaningless the label of cancer might actually be, they would initially be relieved followed by simply being appalled.  The JAMA Internal Medicine article which the Welch commentary was based surveyed over 300 randomly selected people between 50 and 70 and found that less than 10%  had been informed of the numbers behind the screening push for cancer and that most would not agree to common screening tests (PSA or mammograms) once they were informed of just how good (or bad) these test function.

I would be remiss for not mentioning another set of incentives and pressures which have driven this process, that being the threats of litigation for a missed diagnosis. Pathologists are terrified at the prospects of missing a diagnosis of cancer with the prospect of over diagnosis rarely being hyped. However, if the limits of the test were recognized, this is not such an issue. All tests are struck between the rock of sensitivity and the hard place of specificity. Call everything cancer and  you will never miss a case.

These concepts are not new. I became aware of them through the work of Dr. Elliott Foucar more than 20 years ago. They received scant attention from the broader medical community but now that the concept of over diagnosis is rising to the attention of the general public, the source of this systemic error is going to shine a brighter light on the limits of anatomic pathology as being the gold standard of cancer diagnosis. Using legacy approaches in anatomic pathology to make extraordinary claims is reaching the end of its useful lifetime. The curtain should be pulled back on the pathology Oz.