The Tyranny of Health Care Research

What is must important for health care leaders, designers, and researchers to understand is their limitations. Leaders have chosen the designs and research that they like without understanding the limitations and defects of the research. My vote for what has most set the US behind in health care goes to the tyranny of the aggregate. 

Once again, Dr. Saurabh Jha comes to the rescue by nailing this tyranny the best.

"What is the central tendency of a distribution but a lazy generalization? The aggregate, the mean, is wrong about everyone but the few closest to the mean, yet is so revered because we mistake the aggregate for the truth. The tyranny of the aggregate is the most extraordinary tyranny of our times. The aggregate is built by people who vary, yet it imposes itself on the individuals, the very variation which creates it. It literally bites the hands that feed it."  SAURABH JHA, MD (associate editor with The Health Care Blog)

This is likely the most valid critique of quality studies using regressions - the ones that have resulted in hundreds of billions spent on medical error, insurance expansions, and whatever variables are loaded by the researchers for little actual gain in health care.

Drug researchers use the same types of studies to gain approval even if there is marginal benefit at major cost of the drug. When this was pointed out to me regarding widened approvals for statin drugs, I did not yet understand just how this worked. But clearly those who understand and manipulate aggregates their way can save research costs and gain approvals for drug distribution - even with the lack of longer term studies or any real indication of relevant benefit.

It is important to understand that there are many researchers named Jha. This is something that confused me. Saurabh Jha MD has done very fine critiques of ACOs, hospital studies, and the tyranny of the aggregate noted here. Another Jha was the senior author of the male to female physician comparison study - enjoying its short time in the sun while taking advantage of lazy generalizations via the tyranny of the aggregate.  This is pointed out by Dr. Saurabh Jha.

The comparison of male to female physicians has other issues such as differences in distribution, volume, types of locations, types of patients seen. If would take 200 hospital admissions to only female internists to save your life one time. Also the study never considered excluding physician types known to be quite different in key areas such as communication skills. Previous problems with apples to oranges studies have already been reviewed. Different subjects require different studies and do not deserve being placed  together for comparison.

I have performed tens of thousands of regressions if not hundreds of thousands in various studies of health care workforce, county level studies, and other areas. Playing with regressions can be come an addiction. It is quite wonderful, until you begin to understand the limitations of the variables and your findings. Even worse, you can see how the studies can be manipulated intentionally or not. You may even forget why you did the studies or even the question that you asked. Now we know not only why some studies are difficult to replicate, but how.

If you thought physicians were hard to change, think how hard it is to change the beliefs of innovators, health researchers, digitalization worshippers, and others who have seen the works without the limitations.

The Tyranny of Health Care Research 1995 to 2020

Inconsistent is the best description of what is presented as research in areas such as quality, cost, and access. It is important to understand the assumptions of regression studies and other approaches. Perhaps it is important to highlight the impact of such works under categories such as
  • Health Insurance Expansion and Intervention - expansions of insurance did not impact important areas such as primary care, mental health, and basic services which have remained stagnant or are in decline. The lowest physician concentration counties with concentrations of lowest paying insurance plans still remain behind.
  • Medical Error Studies and Interventions - studies have claimed dramatic numbers such as 100,000 to 200,000 deaths a year in US health care that supposedly could be avoided. But 20 years of interventions have not demonstrated significant progress at great cost.
  • Pay for Performance Studies and Interventions - Pay for Performance has been another assumption that has failed to significantly change outcomes in small to large scale studies. Even worse, the pay for performance, MIPS, MACRA, value-based, and Readmission Penalty designs discriminate against providers that care for more challenging populations with lesser resources, greater complexity, and more difficult situations/environments.
There are just a few that fight against the Tyranny:
  • Competent Researchers
  • Those who deliver care and understand the compromises
Competent Researchers are outnumbered and outgunned. Popular research gains the attention of journals and the media. The research demonstrates that the limitations are far too few and studies are taken out of context and beyond limitations. Even worse, health policy interventions in areas such as payment have been implemented nationwide before reasonable study and critique.

Those who deliver care have long sensed the problems involving assumptions, rapid chaotic change, and compromise of patient care. But often their concerns are minimized or worse. Often they have been called incompetent or resistant to change. It is the duty of all who provide care to be diligent regarding patient care. Their compromise is a compromise to all of health care.

Research Dividing the Nation

Research that fails to understand most Americans left behind and what impacts their health, education, economic, and societal outcomes is research that helps to facilitate national divisions.

At one time research was focused upon disparities between populations. This led to important changes decades ago.

Recent research has been more dominated by what is popular. Research with value for advertising revenues and social media appears to be preferred and has a fast track to publication - even if pages of limitations and necessary explanations are missing.

Forgive me for going with the dramatic and the popular, but these various blogs and articles are easy targets.

Everyone with large databases is able to demonstrate what looks like large differences in populations voting for Trump. The differences have been linked to numerous social determinants, health outcomes, and other demographic differences.
  • The first problem with these studies is that actual voting is inferred. 
  • Also most people did not vote. Voting is one of the least representative characteristics as not everyone can vote. 
  • Those that do vote are a select group and can be different from the people in the places where they vote. 
  • Americans can also vote differently than predicted as is well known now. Researchers, policists, and the media were off in their estimates, possibly because they understand most people left behind the least.
  • Research often looks at averages and fails to examine distributions of voting. Inclusion as a red county or blue county may involve only a small margin of difference. This can lead to the need to manipulate the changes to magnify the difference - such as changes in voting last election to this one as seen in some of the research. 
  • The fact that people did change from last to this election is lost in the stereotyping that goes on across the choice of research, the capture of data, the analysis, the selections that lead to acceptance and publication and promotion by the media. 
  • Red county or blue has varied over elections for many reasons. 
The difference for most counties was small. Some studies apparently needed to use their own take on the variable to use such as magnitude of change in voting from past to the current election to show differences. This is an obvious demonstration that the candidate makes a difference, or what is perceived about that candidate that makes people vote for or against them. This, of course, is usually not presented in the research and is too complex for media or social media reduced to a few words or a small politically charged poster or posting.

More Tyranny of the Aggregate

Rural people as a whole do have lesser social determinants, but there are wide variations. Picturing Trump voters as less educated, or more likely to be dependent is not helpful.

Few consider that painting rural people or minorities or others in a bad light helps to divide the nation and may result in voting behaviors guided by emotion rather than other considerations.

Understanding the Oppression of Agendas

Does it matter now to understand where Obamacare came from? Does it matter than the managed care and Dartmouth researcher assumptions became the law of the land - assumptions based on a small portion of the population? Does it matter than most Americans and their needs were not considered along with small practices, small hospitals, and care where needed? Does it matter than most Americans are not desired by ACOs, insurance plans, practices, and hospitals because they are associated with adverse outcomes because of who they are and their behaviors, situations, environments, and deficits of local resources and workforce?

Guidelines for Researchers Desiring to Get Published

  • If you want to get published, choose an area of study involving quality measurement, male vs female, or divisive politics - the more the merrier.
  • Be sure to develop databases with hundreds if not thousands of variables. Choose the variables most favorable to your cause.
  • If you want to show no difference, choose the same or similar populations for care (NP vs MD, Resident work hours limitations studies before and after). You can also focus on a distracting area such as handoffs implying problems. Since you are the researcher you can find your way to a study that gets published even though it may be a distraction from the population and local resource factors that are much more likely to shape outcomes.
  • If you want to show a difference, compare populations who are different or providers who are different with different populations served (Primary Care Medical Home, Urban vs Rural Hospitals, Pay for Performance). Including states that perform well and are advanced in information prowess such as Washington State makes independent nurse practitioners look good in Medicare, since half of the combined populations of the 7 states for comparison were from Washington State.  
  • Avoid areas such as health access or real solutions to health care problems. The solutions are a poor fit with current beliefs and assumptions and are usually too complex to grasp. In addition, the evidence against is mostly common sense and global and macro as compared to micro biomedical or micro policy research studies that are simplistic and work for comparisons involving one independent variable which is often not really independent.

Male vs Female Internist Studies Compare Apples to Oranges

Few would contest the statement that males and females are different. Somehow it is hard to see that a comparison of males to females is problematic. When there are differences between the two types of subjects being compared, it is difficult to compare them or there will be compromise or there will be flaws in the comparison
  • We should question the differences when there are differences between males and females such as origins, ages, years in practice, training, practice location characteristics, practices of different type, and practices in different regions of the nation. Without complete inclusion of controls for all of these areas, then females will look different but not necessarily because of gender. 
  • Males are found in places that have less healthy hospitals with lesser payments and poorer outcomes. The study did not account for lower ratios of internist to patient which was implied as indicated by higher volumes assigned to male internists. Hospitalists with high patient ratios lose the ability to discharge faster and other adverse outcomes are likely. When there are differences in the populations receiving care, it is very difficult to provide controls as these are also apples to oranges different. 
  • Pay is different for many reasons not involving gender. Such studies are popular because of advocacy groups, beliefs, and assumptions. The differences make if difficult or impossible to tell what is gender and what is not.
  • We should question the differences based on females - female physicians more likely to be in settings with greater support from more lines of revenue and higher payments.
  • We should question based on the higher volume of male physicians compared to females.
  • We should question based on the consistent differences. When a wide range of outcomes all have consistent differences between males and female rather than some better for one outcome and at least a few better for the other this implies consistent differences such as is seen in distribution.
  • We should question based on the communication skill differences of certain internal medicine males such as those born in other nations - known to have higher discipline rates.
Males are more likely to be found in places with lower concentrations of workforce, health resources, social determinants, and local support along with patients who are older, less healthy, and smoke, binge drink, eat more, exercise less, and have poor to fair health at higher levels not to mention higher levels of preventable deaths.

Readmission Improvements - Shaped by DRGs or ACA?

Then of course there is the problem of DRGs and their cost cutting focus. This results in a change in the priorities to lesser cost. Lower costs are best addressed by sending patients home quicker and by cuts in personnel. Nurses suffer the most under such cuts. DRGs was a knife at their throat and a major motivation to much activity and advocacy. More importantly the combinations of dumping out patients faster and marginalization of the safety and education contributions of nursing should have some impact - and there is such evidence as provided by  The readmission rate improvement claims are another problem as readmissions were likely increased by DRGs which left some room for the recent decrease. Whatever happened to the usual caution when interpreting changes in outcomes - known to go slowly when millions of people are involved. The readmission rate is seriously flawed as a measure of hospital "quality." One size of 30 days does not fit all diseases, diagnoses, DRGs, patients, patient populations, places, and care settings. This is seen with highest penalties where expected at 14% for lowest physician concentration counties (red counties) with 9% for rural and three times less at 3% for urban.

Managed Care to Dartmouth to ACA/MACRA Distortions

It is bad enough to have research that lacks evidence basis. It is worse when this is applied to important areas such as payment policy. The lack of ability of CMS to understand people and local factors is why MedPAC and RAND have had differences with CMS. The fact that these differences surfaced at all is unusual. But there are major issues such as the CMS estimates of expanded health care coverage twice as high as CBO.

That Readmission and MACRA make matters worse where care is most needed should not be a surprise as this has been going on for decades of designs.

Rural vs Urban Hospitals

Assumptions that bigger is better and higher volume is higher quality have also ignored apples to oranges differences. The male to female study fits in the waste bin with rural vs urban hospital quality under the category of apples to oranges flaws since there are differences in locations, populations, local workforce, hospital personnel, local resources, local situations, local behaviors, and lesser payments.

The Future of Health Care Improvement Is About People Improvement Before, During, and After the Limited Impacts of Clinical Interventions

Some day we might just see through such studies when we do finally understand that 60 - 70% of outcomes in areas such as health and education are about the people and their behaviors, environments, situations and how they interact with the community over decades before the event, admission, test or other outcome is measured. And in the case of readmissions, almost as bad as MACRA in inability to discern "quality," the various people and local interactions after discharge come into play.

Then we can understand that To Err is Human has been a great diversion from the real determinants of health outcomes just as spending 1 and then 2 more trillion on health care has also been higher cost for little change in outcome - the opposite of value based.

Even worse we can look at data from Blue States and Red States and begin to see how the massive increases in health care costs are worsening investments in people and their communities - and worsening health, education, economic, and social outcomes.  The most blue counties receive health spending at 29000 per person while the red counties only receive about 3500 per person.

But the journals and the media are too busy polarizing and paralyzing the country and making it easier for the misguided to make matters worse by slashing and burning spending that matters such as Social Security, disability, Veteran benefits, Medicare age 65 and 66, and insurance for poor children - all slightly more concentrated where outcomes are least and where most Americans are found - left behind by numerous designs.

Then there might be some understanding regarding those who would vote against the candidate most closely associated with the establishment that has done little for them for decades as disparities widen for most Americans.

And those perceiving correctly that they are left behind by prominent areas such as health care and have had little attention in other areas might just vote differently in the next election as they did previously - except for the venom being released by social media and other media outlets and fueled by numerous journal articles that are as much about stereotyping as the material found in the media. Isn't it great to write articles that gain you praise from academics and the media and various advocates as the material adds to the advertising and marketing revenues while leading the nation to greater divisions and impaired healing. 

Recent Blogs 

Why Prevent Doctors and Nurses from Teaching and Nourishing?

Office Visits Do Not Break the Bank But Insurers Can

Necessary Rather than Disruptive Transformations

Ending the Disruption of Pay for Performance and Payment Plans that Lack Evidence Basis and Discriminate

Best of Basic Health Access Blogs

Do Family Medicine Leaders Deserve the Trust of the Students Choosing FM?

Family Medicine Leaders Must Move Access Forward Not Backward

Readmissions Better from ACA or Preexistingly Worse from DRG?

Does Academia Compromise Health Care for Most Americans?

Demographics Distributions and Discriminations in Health Care

Demographics Against the Democrats

Not Easy Being Swiss Cheesy in Health Info Tech

The 25th Anniversary of the COGME Third Report and No Change By Design

Why Is Value So Hard to Recognize in Health Care and why does family medicine not value family physicians and the high value places where they practice

The Four Horsemen of the Primary Care Apocalypse - Medicaid, High Deductible, Veteran, and Medicare Plans shape failure by payment design

Plea to Academic Leaders - Please No More So Called Primary Care Solutions - No Training Intervention or Practice Rearrangement Can Work without Payment Reform

What Is Stunning in Primary Care Is No Change By Design - Numerous failed attempts to recover primary care all point to insufficient payment made worse by accelerating cost of care. 


Popular posts from this blog

Training Cannot Overcome Deficits By Financial Design

Information Technology Cannot Heal

Critique of Commonwealth Fund Report on Ensuring Equity