Stat counter

View My Stats

Sunday, August 27, 2017

Understanding Donald Trump

After speaking to my oldest child last week (and he is not a kid anymore), I believe I have gained some insight into Donald Trump. What Andrew observed was that it seems that Trump is critical of anyone who is critical of him and furthermore, Trump is incapable of being critical of anyone who is supportive and uncritical of  him. It is a very simple principle of tit for tat and very consistent with the reasoning of a man who does not appear to be especially capable of any deep and reflective thought.

After thinking about this observation, it brought to mind the writings of Jonathan Haidt and his work "The Righteous Mind". This work is nicely summarized in 2011 article in Scientific American (Link)

"To understand what constitutes these moral matrices Haidt teamed with Craig Joseph from the University of Chicago. Building on ideas from the anthropologist Richard Shweder (with whom they both had studied), they developed the idea that humans possess six universal moral modules, or moral "foundations," that get built upon to varying degrees across culture and time. They are: Care/harm, Fairness/cheating, Loyalty/betrayal, Authority/subversion, Sanctity/degradation, and Liberty/oppression. Haidt describes these six modules like a "tongue with six taste receptors." "In this analogy," he explains in the book, "the moral matrix of a culture is something like its cuisine: it’s a cultural construction, influenced by accidents of environment and history, but it’s not so flexible that anything goes. You can’t have a cuisine based on grass and tree bark, or even one based primarily on bitter tastes. Cuisines vary, but they all must please tongues equipped with the same five taste receptors. Moral matrices vary, but they all must please righteous minds equipped with the same six social receptors."
It may appear that he is thoroughly unprincipled but that is not the case. Trump values loyalty beyond all other principles. His earlier dealings with Jim Comey provides support for this. Recall that in his initial dealings with Comey what he asked for ( LA Times link)

President Trump demanded "loyalty" from former FBI Director James B. Comey and asked him to drop at least part of the bureau's investigation of former National Security Adviser Mike Flynn, Comey plans to testify to Congress Thursday. In a prepared statement posted on the Senate Intelligence Committee website, Comey says that in a private dinner with Trump on Jan. 27, the president asked him if he wanted to remain as head of the FBI and told him “I need loyalty, I expect loyalty.”"I didn’t move, speak, or change my facial expression in any way during the awkward silence that followed. We simply looked at each other in silence," Comey says. He told Trump that he could promise "honesty," he says.
Comey did not get it. To Trump, loyalty is much more important than honesty. Trump's mind is likely equipped with all of the same six social receptors, but he applies a different emphasis on particular receptors. Haidt showed that this differential emphasis is a characteristic of people with certain patterns of emphasis associated with particular political leanings.  One can take a test to see where your individual leanings might be and which moral senses you might emphasize at (  Haidt has noted that liberals and conservatives tend to place emphasis on different priorities, where liberals placing great emphasis on care and fairness and conservatives placing more emphasis on sanctity, authority, and loyalty.  Surprisingly, there seems to be an inherited component to this differential weighting that goes beyond environment. Thus, our moral intuitions may be driven by genetics to some degree.

Although I realize this is rank speculation, I believe if Donald Trump took this test his results would be skewed toward a single minded emphasis on loyalty.  His focus on loyalty is so strong and so single minded, it is hard for the rest of us to comprehend.  Yet, it does seem to explain what otherwise defies explanation. Donald Trump becomes quite understandable once you realize that no other principles are in play.

Monday, August 21, 2017

The Promises and Perils of AI - Artificial Ignorance v. Artificial Ignorance

AI - New Yorker

There is a buzz out there in the health care delivery world about the promises of artificial intelligence (AI). There are fears among physicians that they might be replaced by computers. There is excitement and there is fear and there is hype. In my opinion, at this point there is mostly hype. The reason I believe that what we have most is hype is that for most of the important tasks we can delegate to AI, we are missing one key element. AI is not something programmed. It is something learned and in order to learn a computer needs validated data sets which contain unambiguous right and wrong answers. There lies the rub.

The recent article in the New Yorker by  (AI - New Yorker AI v. MD) describes studies done by Stanford where they trained computers using images taken from patients diagnosed with melanoma.
Thrun, who had maintained an adjunct position at Stanford, enlisted two students he worked with there, Andre Esteva and Brett Kuprel. Their first task was to create a so-called “teaching set”: a vast trove of images that would be used to teach the machine to recognize a malignancy. Searching online, Esteva and Kuprel found eighteen repositories of skin-lesion images that had been classified by dermatologists. This rogues’ gallery contained nearly a hundred and thirty thousand images—of acne, rashes, insect bites, allergic reactions, and cancers—that dermatologists had categorized into nearly two thousand diseases. Notably, there was a set of two thousand lesions that had also been biopsied and examined by pathologists, and thereby diagnosed with near-certainty.......
...Thrun, Esteva, and Kuprel then widened the study to include twenty-five dermatologists, and this time they used a gold-standard “test set” of roughly two thousand biopsy-proven images. In almost every test, the machine was more sensitive than doctors: it was less likely to miss a melanoma. It was also more specific: it was less likely to call something a melanoma when it wasn’t. “In every test, the network outperformed expert dermatologists,” the team concluded, in a report published in Nature. 
So should our dermatology brethren be afraid that Watson and its prodigy will supplant the mole spotting workforce in dermatology? Perhaps, but there is a flaw in this work.  What does it mean to use "biopsy proven" images? What exactly does a biopsy prove?  It may not prove anything and there lies the problem. The teaching sets upon which machine learning is based may be validated (or not) by a not so shiny gold standard.

In a recent paper published in the British Medical Journal by Elmore et al (BMJ) the reproducibility of histology in melanoma diagnosis was examined.  The results are a bit concerning and call into question the gold standard status of anatomic pathology and its ability to "prove" anything. The best concordance found was about 80% for lesions believed by experts to be frankly malignant. That means any training set the computer viewed likely had at least a 20% error rate built in. For the more subtle lesions, the concordance rates hovered around 50% (and some lower).  How about comparing this to coin flips?

Training machines to learn to make diagnoses by using flawed teaching sets will generate AI; perhaps more likely to generate artificial ignorance than it is to generate artificial intelligence.

Sunday, August 20, 2017

Civil War Reconstruction continued

I have been a bit quiet for a while. I, like much of the country, have been appalled by the antics of President Trump for a number of typical and also contrary reasons. I do not disagree with everything that he is attempting to accomplish, just because embraces awful positions on other issues. In some sense, I am even more deeply resentful of his appalling white supremacist and neo-Nazi apologist statements. They are appalling in and of themselves, but also because these statements also can prejudice others against anything else Donald Trump may embrace. They poison his entire agenda, good and bad.
When one looks into the history of Confederate memorials, it presents a not so virtuous justification of erecting them in the first place. The great surge in erecting these monuments coincides with the promulgation of Jim Crow laws and of Lost Cause mythology. The Old South prior to the Civil War was built upon a foundation of racism at its worst. It was an economic system where people owned other people. The Lost Cause mythology attempted to soften that ugly fact and revise history to make the actions of those who led the rebellion,  to preserve the system which allowed certain people (whites) to own other people (blacks), to appear more virtuous than they really were. It also attempted to have the oppressors and former slave owners assume the roles of victims.

The monument building occurred at the same time that Washington DC was resegregated by Woodrow Wilson, the KKK membership spiked, and many gains African Americans garnered in the South after Reconstruction were rolled back through enforcement of Jim Crow laws. The nation put in place immigration laws at that time that were blatantly racially exclusive. The monuments to the Confederacy were erected for a purpose, and a not so benevolent one. We need to recognize that was part of the plan.

Do the statues need to come down? I like the idea that they need to come down only after a period of reflection and education. Here they are in plain sight, where some have been for more than 100 years. Many of us simply ignored them as empty and boring markers of distant history no longer relevant to our lives. However, the history involves events which happened not so long ago and we need to be more aware of what happened then because it is relevant to what is happening now. I believe they were erected with malice intent with an agenda focused on fear and intimidation. Some people in current times share this same agenda. There is nothing benign about this. I believe that once the truth about these statues becomes widely understood, they will come down because they will be a source of embarrassment.

In my opinion, it is not justified to honor people whose behavior and actions were neither heroic nor morally sound. They likely were people with many virtues (or at least some virtues) but their decisions to fight for a way of life predicated on human enslavement put them on the wrong side of the moral divide. It might be said that our founding fathers (Washington and Jefferson) should also be viewed in this light but I would like to make a basic distinction. Washington and Jefferson were hypocrites in that they were slave holders,  but they did not mount a rebellion based upon preservation of a morally unjustifiable tenant of slavery. Their monuments were erected to celebrate their accomplishments in creating a Republic which, although imperfect, has been a remarkable accomplishment.

Confederate leaders led a rebellion primarily motivated by the desire to preserve this abhorrent institution. We do need to recognize that their views were not unique for the times. Slavery was the norm for thousands of years and these men and women who embraced it in the early and middle part of the 19th century represented the tail end of slavery acceptability, at least in the developed world. Their lack of insight may be explainable given history, but it is not morally justifiable and clearly does not warrant any monuments celebrating their lives and actions. In the end, they accomplished essentially nothing. What is there to celebrate about their lives and accomplishments?

This series of events also highlights the perils of focusing on moral equivalencies. At the most basic level, everyone is flawed and everyone makes mistakes. This creates the opportunity to level all moral transgressions, elevating minor infractions to major status and lowering major one to minor status. We need to recognize that some transgressions are simply worse than others. Yes, I understand that this can lead one into another slippery set of slopes but we can for our work on the ends of the spectrum and not in the middle. Advocacy of Nazism and White supremacy has no middle way aspects. The philosophy is poison and has recent history marked by brutal violence and mass murder. We would be absolutely mortified if current day Germany started erecting statues honoring Adolf Hitler and Herman Goering.

Sunday, July 16, 2017

Repeal and Replace Agonies

The Republican attempts to change the ACA into something else are destined to fail. They are destined to fail not necessarily because their ideas are devoid of any merit, but they are destined to fail because they are charged with a series of impossible tasks. The only acceptable outcomes which are politically sellable are ones that must include the following characteristics:

1. All pre-existing conditions must be covered and insurance companies cannot discriminate on the basis of age of illness. To be actuarial sound you need broad participation but you can't force people to buy insurance they do not want to buy.
2. People must be insulated from the costs of care - this means minimal to non-existent copays and deductibles
3. Since health care is a right, all reasonable services need to be covered, including new and innovative treatments, drugs, and procedures
4. Cost of care must be kept in control and increases in costs must not break state or Federal budgets. However, you need to cut costs without cutting expenditures. Cutting expenditures means you will be killing people. This is where the game of political Twister really becomes interesting.

This is an impossible task. They simply cannot succeed. The insurance markets are a complete mess with the prospect of complete breakdown without intervention. However, no intervention is likely to happen without total collapse. The collapse will be used to justify movement to a Federally administered health system.  Everyone will gain coverage by fiat but that is when the next set of challenges will become apparent.

With a single payer which presumably will coopt present insurance markets, the question will be whether it will represent a floor for care or a ceiling. If a Federal "universal" program has ambitions to provide a comprehensive package of services to everyone covered, it is no trivial task to decide what is covered and what is not covered. Who is going to do this? Are we simply going to contract with the current insurance carriers to do what they have been doing? What are the gains that we will see from this approach. We as physicians and patients will end up bargaining with the same people we have been bargaining with all along. How will these parties be incentivized to administer the system. I suspect they will be rewarded for stinting on care. Sound familiar?

I have worked within a Federal single payer system call the VA Health System. The VA system is amazingly comprehensive. In fact, there are days where it appears that there is no service which might be delivered to a veteran which cannot fall broadly into the scope of health care service delivery. Furthermore, there is really no one who is charged with the job of defining what the scope of health care services might be. The end result is a perpetually expanding scope of services all defined as within the scope of their right to health care. If the job of defining scope does not fall upon former insurance companies, it will end up in the hands of Federal employees who will not be empowered to anything other than allow for scope creep.

The point is that a Federally financed universal health insurance program will not be administered by the Federal government. The Feds to no have any experience in dealing with the systems required with the exception of the VA Health System and the only thing less politically sellable than the Republican alternatives to the ACA is to put everyone into the VA health system.

Despite the explosive growth of Medicare costs, the care of patients on Medicare patients has been cross subsidized by patients whose care is covered by private insurance. A movement to Medicare for all will represent a price shock for suppliers of care They will push for acceptance of Medicare for all only if the system allows for patients to purchase supplemental policies that do more than help pay co-pays and deductibles. Care can be delivered for Medicare prices only by paring costs dramatically. With fewer financial resources coming in, health systems will need to figure out how to operate under these conditions. They will figure out how to do less and spend less and justify these actions.

We are already seeing hints of this under the current system. Rural health delivery is disappearing. It is simply too expensive to maintain a comprehensive set of services where the costs to deliver these services are higher. The first services which go away are ones with small or negative margins. There are also massive movement away from using physicians, who are expensive. The move to a single payer would in essence make everything look like rural health. The drive to reduce costs and to do less, especially less of anything low margin, would translate into whole swaths of care services disappearing. If you have a hard time finding something now, it will only get more challenging.

This is actually happening already. A shift to single payer would only accelerate this shift. However, universal state sponsored health plans also exist within the context of private insurance. This happens in France, Germany, and Great Britain. The Universal Plans provide more of a floor than a ceiling. From my understanding (and I may be wrong) the Canadian system historically had few non-state outlets, other than travelling across the southern border and paying cash to the doctors at the Mayo Clinic. A more basic coverage model is clearly possible with non-covered or poorly covered services being available via supplemental insurance. Whether this would look like our current alternatives for Medicare covered patients is uncertain. If more options are made available to non-Medicare patients, these additional choices will ultimately diffuse into the Medicare population. The program may end up to be a Medicare for all program, but I predict that the end result would be a disruption of Medicare as we know it. Giving options to one group will obligate us to provide those options to all.

What do I predict will happen within the next five years?

1. Single payer in the US. It will not likely happen in the current administration unless there is a complete meltdown of the insurance markets.
2. Single payer will result in dramatic changes to the Medicare program
3. The big debate will be how much of a parallel private insurance market will pop up  - the other tier

Sunday, June 4, 2017

Drug Recognition Experts and other snake oil salesman

I saw a story on the local news yesterday where they showed a video clip of three different people all stopped for minor traffic issues and subsequently arrested for being "under the influence" (News Link). The assessment of their compromised state was made by a single policemen using an algorithm of dubious utility.

The news story focuses on a particular Officer Carroll, a decorated Cobb County Officer who is one of the approximately 250 officers in the state of Georgia who has been trained as a Drug Recognition Expert. In the three cases identified in this news story, Officer Carroll's assessment was at odds with the final laboratory evaluation, which did not demonstrate the presence of drugs. The response of the Cobb County police was nothing short of remarkable. They commented -

"Commanders would not let Officer Carroll talk with us, but they stand behind the arrests. The department doubled-down on their assertion that the drug recognition expert is better at detecting marijuana in a driver than scientific tests."
Say what? The gold standard is an poorly validated and subjective test which can be trusted over the actually measurement of a certified lab whose machines and assays use actual positive and negative controls? What drugs are these people on? 

RAPID decision making

I learned about a new tool for decision making. It is called RAPID and it has been credited to Bain and Company Inc.

While I was introduced to this tool's use in committees structure within ate large health organization, it seems potentially even more impactful within clinical environments. In clinical environments teams of care workers participate,  yet roles and responsibilities are generally very ambiguously defined. A patient who presents with a set of complaints such as shortness of breath, decreased visual acuity, and a new onset rash on the background of hypertension, diabetes, anxiety, and history of opioid abuse will undoubtedly require a large care team to address their issues. Within any team charged with addressing these problems there will be a host of overlapping roles which currently are defined on an ad hoc basis at best, and more often than not never defined at all.

One could imagine creating a modified SOAP note or problem list which would include team members charged with making actual decisions and execution of specific plans. Accountability never happens without actual ownership of problems and definition of roles and responsibilities. 

Assessing outcomes in healthcare: Do we need the equivalent of double entry bookkeeping?

We tend not to understand how what is now viewed as mundane was once revolutionary. The simple act of balancing our checkbook is a legacy of a revolutionary and transformational process which was first propagated not much more than 500 years ago. That process is double entry bookkeeping (DEBK). There is some contention as to where DEBK was first used or at least widely used, but there is little contention that is the big picture, it is a relatively newly adopted human practice. Humans have been accumulating and trading for thousands of years. We have been formally accounting using DEBK for only a few hundred years.

What is the big deal about DEBK? It allowed individuals and more importantly larger organizations to organize large amounts of information into relatively compact journals with visual displays which allowed people to accurately assess whether their activities where leaving themselves and their organizations better off after transactions. DEBK is one of the foundations of the modern trading economy and served as a foundation for the growth of wealth and the unprecedented  improvement of the human condition which has happened in the past few hundred years.

Let's move from commerce to healthcare. In the healthcare economy, at the most fundamental level the ultimate goal is to leave patients with more health assets after encounters than they started with. While appearing very simple when boiled down to this principle, setting up the ledgers is not so straight forward. What exactly do the entries look like? What are health assets and what units can they be measured in? What specifically do people value in terms of health?

There is the absence of undesirable symptoms whether pain, anxiety, or fatigue. There is also the presence of particular functionality. Can you walk, run, climb stairs, think clearly, solve problems, or function sufficiently to work and earn a living. This sounds complicated to measure but there are already various patient reported outcomes tools which measure many of these elements.

There are also financial tools which can be used. For any given intervention, financial assets are needed to deploy. There is always a financial and time costs to devoting resources to address health issues. If we are able to measure health assets over time, theoretically we can begin to assess whether a financial commitment to a person results in a good investment.

The barriers to deploying such a project are not insurmountable. One issue we will need to anticipate is how we will execute the conversion of financial resources into health assets. There will be huge variation in terms of the preferences of specific individuals. Some people will want to invest large amounts of money to improve their health asset picture while others will want to invest those resources elsewhere.

Furthermore, improvement of health assets may be best done via investment directed toward activities and services not commonly viewed as health care services. There are currently biases skewed to certain directions of investment driven by health insurance which probably drive inefficient allocation of resources that leave people and populations worse off than if the resources were invested via some alternative approach. However, without the equivalent of DEBK for health assets, the pernicious effects and the asset losses created by these suboptimal investments are hidden and opaque.

For those who repeat the mantra that health care is different, DEBK application into the world of health assets will be a wake up call. Improvement of the human condition requires human activity driven by incentives and systems that allow us to measure whether investments of time, effort, and money actually provide on return on that investment. Do we leave people better than we found them and are the investments we make best applied to where they can best accomplish this goal?