Thoughts on macro goals for health care policies

In an earlier post (also here), I wrote some thoughts about setting goals for healthcare policies.

Part of the difficulty of making progress in the healthcare debate is that the popular debate never covers a workable goal for setting healthcare policy.   As in this video, the goals are often overly simplified in terms of universal access (insurance cards) or platitudes such as better quality for more people at lower cost.   Usually no one objects to these goals: most people agree that health-insurance facilitated access to healthcare is a good thing, and no one objects to striving for better quality for more people at lower costs.    The arguments concern how to achieve all these goals simultaneously, not whether these are the correct goals or the complete list of goals.

Now that government is intricately engaged in funding and regulating the healthcare industry, the questions of healthcare policy are politically important.   While the tendency is to hide the details of policy making behind opaque government agencies, we should be demanding more democratic participation in these questions.  This is especially urgent when the current policies are resulting in deficit spending to deliver healthcare services.  At the government level, healthcare spending should be revenue neutral where all of the annual expenses are balanced with revenue sources specifically set aside for healthcare expenses.

I don’t think healthcare expenses should add to the debt.  Eventually, we will paying debt for ancient healthcare expenses that benefited individuals who are no long living.  In the extreme, several decades from now we’ll have elder people whose healthcare expenses will be restrained by obligations to service debts for healthcare expenses of their long-dead grandparents.   We need to approach healthcare policies with a constraint that the its expenses must be strictly revenue neutral.  There is no long-term return on healthcare expenses, especially after the treated patient eventually dies.

The trouble with the debated healthcare policy goals of universal access, decreasing costs, or improving outcomes is that these have no limiting conditions.    Even if we have tremendous advances in healthcare technologies and practices, we’ll still be overwhelmed by providing even better care for even older people who have even more complex needs.  Currently, the deficit spending for healthcare is a secondary concern.   We currently accept deficit spending for current healthcare.   We observe deficit spending as a scoring mechanism for alternative policies that will either increase or decrease the growth of healthcare spending.   Savings is expressed as the difference between alternative future deficits, not whether the deficits will ever be zero.

Revenue neutrality of healthcare policies should be a primary goal with the secondary goals of universal access, improving quality, and lowering cost burdens on the population.   I think this debate is hard because none of these secondary goals appear to be negotiable.   No one is going to argue for trade-offs between these goals.   No one will argue that lowering expectations for quality can result in a more egalitarian and more affordable healthcare.  Nor are there advocates for the other trade offs.   Yet, given the priority of balanced budgeting for healthcare, these are the trade-offs we should be debating.  To his credit, the above video does broach the subject though it is hidden in the context of the frontier where one accepts more risks from less regulation.

Part of the reason why the debate over trade-offs is so difficult is that the objectives (universal access, lower costs, better quality) are incomplete.   There are other objectives we should be considering as peers to these beneficial goals.

We should take a clue from the machine learning advances that tells us that working with a larger set of data with a larger number of considerations (features) can lead to actionable and beneficial conclusions.  The lessons of machine learning is the opportunities provided by the large number of features available for consideration.  The technology for big-data analytics behind machine learning are readily available for everyone.  We now have the capability to debate policy choices based on the full range of considerations instead of just the few (universal access, high quality, and low cost) we normally address in sound-bites or tweets.

Although we debate healthcare at the national level, healthcare is inherently a local problem.   Even with the national policy, we subdivide the market into approximately 3100 counties.  The questions of access, quality, and affordability vary for each county.   Adding counties into the set of features provides more granularity to discussion and modern tools makes this detail available for public debate.   For example, in some rural and economically depressed areas insurance does not make as much sense because there is no locally available providers or that there are higher priorities for spending money than healthcare.   With today’s technologies, we can easily incorporate these data into public debate even though it multiplies the considerations by 3100.

As we look into details for different counties, we become aware of additional features that should fit into the debate.  Each of these features can apply to all counties but with varying relevance to local problems.   Adding these features will further multiply the number of considerations.  The features are independent dimensions that each offer something to the debate.

My point is that we have the technology to debate over this volume of data and this debate is likely to be more productive than refining the issues into two or three ideals convenient for the ancient model of sophistic debate.   With modern data mining and visualization tools, the public can discuss the details of all of the relevant features of the healthcare debate.   We should learn from how learning arises in machine learning from feature selection among a large set of features with diverse measurements.   In analogy to neural networks, for example, the same volume of data can be presented to networks of humans each with access to tools empowering them to observe the vast richness of the data and then use debate to mimic forward propagation of accumulating conclusions and backward propagation of derivatives of errors.

At the county-level considerations of healthcare, we can identify many additional features that we can now effectively debate about:

  • Attraction of qualified healthcare providers to where there are needs.
  • Healthcare as an employment opportunity for local populations who prefer careers in helping others.
  • Local access to alternative medical practices from cultural traditions and experimental practices.
  • Local variations of expectations of quality of life improvements from healthcare, balancing disease management or cure against tolerable disruption to life.
  • Accommodation to the differences in risk tolerances among the non-consumers of healthcare.

Healthcare needs a workforce

The most valued healthcare providers are motivated with the desire to help others at an individual level with an emphasis of helping another person in achieving their goals.   The following quote captures the dilemma that providers face in their jobs:

Transactionally, Emily’s medical team did everything they should have for a consumer patronizing the Cleveland Clinic’s renowned facilities. They informed her of her diagnosis, gave her medication and made her feel better physically. Yet, at the moment of medical crisis, those services were not top-of-mind as she thought of her future. Was Emily a “consumer” of healthcare, or was she a patient?

The fortress vs frontier story makes a valid claim that freeing technology to automate medical practices can provide benefits of lower costs and broader access.

In general, these technologies automate the specialists who often have limited if any direct contact with the patient.  Healthcare practices that are ripe for automation includes surgery, diagnostic interpretations of tests, monitoring and testing.   These happen to be professions that currently command high compensations so automation offers promise for significant cost savings.   Also, these professions are more appropriate to machines, once the technology is sufficiently mature, because the results will be more consistent and not prone to human fatigue or stress.   These specialties also involve very little direct conscious interaction with patient so they have very little to offer the human element of health care delivery.

In contrast, the general practice primary provider offers a patient benefit of a close personal relationship that contributes to better health outcomes.   The patient benefits from a relationship to discuss their health concerns with someone who knows medicine but also can understand the human stress of dealing with the conditions.  The primary care physician (or equivalent) needs to practice geographically close to the client to permit frequent and routine visits that maximizes the patient’s benefits.

Even if we are successful in automating all of the specialties, there will remain an acute shortage of primary care providers that are an essential element of healthcare delivery and outcomes.  Our policy debates must include consideration of appropriate incentives to attract qualified people into becoming primary healthcare providers at the county-level.

Because the primary motivation for primary care providers is the personal connection with the patient, advanced technology introduced into primary care sessions can become a disincentive for attracting the best candidates for primary care providers.   An example of the disincentive to primary care is the electronic health record (EHR) technology that distracts the doctor’s attention from the patient in order to spend more time in front of a computer screen.

In our debates on healthcare, we can distinguish the need to maximize the human interaction between client and primary care providers from the need to increase automation of specialties.   In particular, we need minimize technology’s distraction of the primary care provider.  It produces a frustration that discourages people from entering the field, or from remaining active in it.   We need primary care providers in each county.  Demanding a consistent level of technology for all counties may adversely affect many counties with rural and less mobile populations.

A workforce needs healthcare careers

While healthcare needs workers, there is also a substantial number of workers who prefer to work in healthcare.  The delivery of primary and direct patient care needs workers to provide these services.   Optimally, the workers would be drawn from local residents in order to minimize commute time and stress.  Our simplistic debate over healthcare often overlooks this critical issue resulting in rules that prevent willing people to deliver healthcare to their communities.

In addition, the number of people desiring careers in helping others (such as in healthcare) is a consistent fraction of population everywhere.  The current healthcare economy employs roughly 10% of the workforce.  I imagine the majority of those workers prefer this work over other jobs, even as they may complain about stress or compensation.   Many people in the remaining population know of relatives or friends who they can tell would be happiest in a job in healthcare compared to other options.   There is a constant portion of the population who are happiest when they are helping others.

As I watch the healthcare debates, I notice that this employment opportunity problem is rarely discussed.   A politically sustainable healthcare policy should generate sufficient jobs to employ people who want to be involved in helping others.   I do not think this objective necessarily is at odds with the benefits of automation for cost savings and quality improvements.   For overall healthcare policy, the primary benefits for automation is in the specialties such as diagnostics and surgery where there is minimal direct relationship with the client.

Automation has a very different value proposition for the long-term relationships of the various primary care providers.  In some cases, technology may interfere with the objectives of providing better primary care.

There are other helping-career options than officially approved medical care.  Helping careers closely related to healthcare may include physical therapy coaching, holistic medicine, or other culturally-specific traditional medicine.   These opportunities vary dramatically across the different counties.    Our healthcare policy debates can include the county-level considerations of accommodating these alternatives as career opportunities for those who wish to pursue them.  Such economic considerations are relevant to the debate.

Consumers do not want to be patients

Great quote byJan Oldenburg, senior manager at EY and vice chair of the HIMSS Connected Patient Committee:

Coming from the patient empowerment movement, one of the problems with ‘patient’ is in its very nature it implies a subordinate role, that of a person who needs to be taken care of as opposed to having autonomy. It’s probably true that at the time you are hospitalized, you perhaps have the least autonomy of any time in your life, but at the same time, people still have agency, preferences and needs in that setting that are not always the same as those assumed by the hospital or caregiver.

As I follow the healthcare debate, I get the impression that there is an assumption that everyone considers themselves to be healthcare patients, or aspires to become patients when they encounter health problems.   The unchallenged assumption is that people with health problems want to fix those problems in the same way that a car owner will want to employ a mechanic to fix his car.

Although modern policy encourages discussions around consumers instead of patients, there remains a conviction that most clients of healthcare prefer to be patients to get fixed like new.  Similar to the expectation of the car sent into the mechanic shop, patients want to be fixed.    This is a part of our culture so that even when we attempt to poll people, their responses will favor the fix-me-up patient status rather than the more ambiguous consumer model.

In some ways, the very name healthcare is misleading.  Health is a part of life and living doesn’t get suspended when a person turns into a patient.   In the scenario of the car being sent into the mechanic, the owner is free to go about his normal life while someone else works on the car.   Patient status demands that the person go along with the body to be fixed.  Patient status interrupts life.

Undoubtedly, most patients want to feel better to avoid degeneration caused by their ailments.  They also want to continue living their lives outside of being a patient.   They have other obligations where they need to be elsewhere than in the healthcare system where most of the time is spent in waiting rooms, life-interrupting procedure preparations or recoveries.

People want a balance of how much of their life they want to devote to their healthcare.   Some people are willing to sacrifice long periods of their life to better healthcare.  An analogy are athletes who spend all of their time for years developing their capabilities.  Similarly, someone with a very active social life may see an intensive patient experience as an investment for maintaining that active life.  But not everyone is so eager to make this kind of investment of their life as healthcare patients.

In general, everyone has different assessments of the value of the investment of their life into being a healthcare patient.  Although when asked in surveys they may respond that they want to be patients, they may respond differently when addressing their specific situation balancing their healthcare needs with other life priorities.  An illustration of this is the frequent complaint that patients do not follow doctor’s advice to change behaviors even if that involves diligent schedule of taking medications.  One way to explain this lack of cooperation is that the individuals are not committing themselves to the notion of being a patient to be fixed.

For individuals, managing their healthcare is inseparable from managing their life.   Some people would prefer to live with as little interruption by healthcare industry as possible.  When they need care, they may prefer an approach that manages the symptoms rather than cure the problem even if that may involve a worsening condition.

The healthcare policy has a good reason to strive to take this choice away from people.  Allowing conditions to worsen (benefiting the consumer by postponing care) will eventually lead to higher costs with lower measurable outcomes.  I still think that we should respect each individual preferences for how he wishes to spend his time.   Time spent in a medical situation is time not spent elsewhere.   The individual feel there are better things to do with his time than be a captive of a schedule imposed by his patient status.   There are many cases of unique opportunities where we will not object to diverting attention attention away from healthcare, especially if that diversion results in some accomplishment.

The distribution of population with different expectations or tolerances for life-interrupting healthcare will vary greatly in different counties.   More urban localities will generally be more accepting of life-interrupting healthcare than more rural counties.   Similarly, poorer localities may be more likely to avoid the interruption from their ability to continue working or to preserve what little they own to pass to their survivors.

In our debates over healthcare policy, we should consider local variations of tolerance for life-interruptions of healthcare.   The goal should be to improve happiness of the individuals.   That happiness may not be consistent with resigning to a period of patient-status.

Balance of risks in pursuit of happiness

The healthcare debate should include features that captures its role in the broader economy.   The outcomes of healthcare policy will have impacts far outside of the healthcare market.   Similarly, the economy will have impacts on health care policy.  Prosperity brings benefits of reducing demand for healthcare through healthier lives.  Lower prosperity will make healthcare worse by overloading with new ailments of lower quality of life and with less access to revenue.

Healthcare debate creates a new form of class-based political conflict: one class needing immediate healthcare and another not needing any immediate or near terms health care.   The former class is further divided into those needing financial assistance to cover true cost of their care, and those who do not either because their conditions are cheaper or they are more wealthy.   The conflict occurs in trying to find a way to subsidize healthcare for those who need it but can not afford true cost.   This conflict is at the root of debates about choosing between rationing and cost sharing.  We can make progress on this debate by making these choices as explicit features of healthcare policy.  The features (data dimensions) for rationing and cost sharing are valid topics for considerations for assigning weights or for feature-pruning compared to all the other features.

An example of the conflict occurs with prices for pharmaceuticals.  Different consumers of same drug pay different prices depending on their insurance who in turn pay different prices depending on independently negotiated terms based on risk pools.   After subtracting what the patient can not afford, the remainder of the price must be shared somehow.  Currently the sharing is through inflated premiums among all members of the same insurance plan with government backing such as reinsurance and risk corridors.  The design makes the assumption that rationing can be avoided by clever financing.   From experience, we should know that technology will continue to introduce new potentially effective but very expensive treatments.   Because we exclude rationing from consideration, these innovations will place more pressure on the financial gaming to solve the problem.  More money has to be found elsewhere to pay for universal access to anyone needing the technology.

This example reminds me of the earlier arguments justifying long-lasting patents of medicines and more restrictions on generics.   The earlier arguments claimed that the inflated prices for widely used but cheap-to-produce drugs subsidized higher prices for more expensive and less commonly required drugs.   Assuming that argument to be correct, it shared the cost of delivery of care among the sub-population of all people needing some form of pharmaceutical consumption.   If it did work that way, then it had the benefit of relieving the remaining population (who consume no prescription drugs) of the cost burden.   Today of course, we’ve removed this option, forcing companies to give up rights to generic producers and for everyone to sell drugs at or near the cost of producing them, at least after a patent expires.   This redirects the cost subsidization for expensive drugs to come from insurance premiums covered by both the healthy and the sick.   The insurance approach balances the cost over a larger pool, but it does raise conflicts by burdening people who are not receiving care with the costs for those who are.  This conflict would be easier to resolve if we allowed this specific choice of rationing and cost sharing to be explicitly part of the debate.

The healthcare debate should explicitly consider the interactions of policy on the broader economy.   In current policy, we have imposed mandates on employers for the types of insurance to offer their employees, and on individuals to have insurance meeting the requirements.   In general, these standards for minimum coverage for mandated insurance significantly interfere with the non-healthcare markets by increasing costs and limiting options for balancing risks.   For example, the community rating for setting premiums prevent employers of generally younger and less risky employees from offering cheaper coverage for more catastrophic care.   Alternatively, the self-employed must obtain insurance even if it consumes up to 9.5% of his income.

Prior to the current policy, there was freedom to accept lower standards for health insurance (with annual or lifetime limits, or with no guarantee of reissue) or to accept risk of no insurance at all.   Current policy sets the expectation of obtaining insurance or paying a fine or tax for not obtaining it.   This has a debatable impact on the economy especially in the more dynamic small-business economy.

Inherent in the starting or running of a small business is directly balancing financial risks.   Accommodating the risks of needing healthcare through premium payments or setting aside savings is just one of the many risks facing a business.   In many cases, there is just one opportunity to make a business work.  The owner may choose to take the risk of under-insuring for healthcare for the opportunity for future success that can eventually pay for healthcare insurance later.  This is clearly a risk as anyone at any time can encounter unexpected but major healthcare expenses.   However, it is just another risk to manage among all the other financial risks facing a business.

It would have been more helpful if the policy had left this mandate open for continued evaluation as another feature to consider for optimizing the healthcare policies.  The slowing the introduction of new businesses or slowing or restraining the growth of small businesses eventually will have detrimental impacts on healthcare outcomes.  The loss opportunity of better prosperity will increase healthcare expenses correlated with low incomes.  It will also provide less tax revenue to finance the healthcare policy.


This list of additional features to consider for healthcare debate could continue endlessly.  For example, I could identify the various tolerances for total out of pocket expenses (premium, deductible, and uncovered expenses), or the relative availability of support networks of friends and families.   A healthy healthcare debate can consider hundreds or thousands of features for each county.   Modern data science technologies permits the public to engage in this level of debate.   The only reason why are not considering more features is because the thought leaders insist on sticking to easy to defend platitudes such as wider access to better health for less cost.   Their arguments are not getting anywhere.

Machine learning solves problems by considering the full range of features and selecting the most important features supported by evidence.  We can learn from machine learning that we can solve our problems by including more features in our debates.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s