“When the apocalypse comes, everyone will be on their own to buy healthcare.” Dave “ePatient” DeBronkart dropped this casually into a phone call. I had to ask about the apocalypse. “It’s when the system collapses under its own cost and complexity”. Our conversation happened a month before the COVID epidemic wiped out revenue for hospitals and pushed them toward bankruptcy.

Maybe this will be good! Maybe we will find that we can buy great healthcare with the money that we save because we were too sick to go out and buy a $4 coffee. Maybe machines will concoct a personalized cancer treatment while we wait at a local Minute Clinic. Maybe we will get a bundle of advice for free with our fitness tracker. Maybe healthcare will be too cheap to meter, or at least too cheap to insure.

Image for post

Cheap healthcare sounds unbelievable now. Healthcare is one of the most expensive pieces of our personal life. It is also the most expensive and complicated piece of the US economy. Costs increased relentlessly over the last 60 years from 5% of GDP to 18%. The supply of healthcare is tightly rationed. It is metered in ten-minute increments.

The phrase “too cheap to meter” is a famous fail from 1954, when Lewis Strauss of the Atomic Energy Commission gave us a vision of “transmutation of the elements,–unlimited power, ability to investigate the working of living cells by tracer atoms, the secret of photosynthesis about to be uncovered,–these and a host of other results all in 15 short years. It is not too much to expect that our children will enjoy in their homes electrical energy too cheap to meter.” What actually happened was exactly the opposite. We got the energy crisis, a generation obsessed with the scarcity of energy, and a baking planet with a contaminated atmosphere.

Image for post
Lewis Strauss — Great speech. Famous fail on cheap fusion energy

But we eventually turned the energy market around. In 2009, computer networking pioneer Robert Metcalf shocked audiences with his thoughts about “energy in squanderable abundance”. Technical progress has made the bits that flow through our networks too cheap to meter. Maybe that will happen to energy. Only eleven years have passed since Metcalf’s musings. Renewable electricity is cheaper than the old forms of energy, and still declining in price. The price of oil traded below zero. Fusion power is back on the agenda.

55 years later, Metcalf saw a squanderable abundance of energy, with self-driving bonus

Where health care is getting cheap

The same type of cost declines can happen in healthcare. I’m working on automated advice, driven by broad-spectrum diagnostics. This will give each human an agent that looks out for our health. Let’s look at how this might play out.

A machine can read through a health record and generate advice and reminders. It can do this cheaply because it is just software. This automated advisor cannot pat you reassuringly, but it does some things better than human doctors.

  • The software and rules are quality controlled and continuously improved.
  • They are comprehensive and frequently updated.
  • They can interpret big, real-time, and complex data, including images, DNA sequences, and sensor and device streams
  • They are always available. They can pay attention to their patients overnight, or all day.

We will eventually have AI advisors that understand us at the molecular level. They will be personal agents that truly look out for our individual health, and they will not be swayed by the desire to sell services.

We can start with the current technology. We have an expanding number of rules that can look at a standard FHIR health record and offer reminders, followup, focus attention on risks, and check for problems in drug-drug and drug-gene interactions. They can also help us select care that is high quality, low cost, and reimbursed.

To provide good advice, we need information from tests and other diagnostics. There are thousands of different ways to look at blood, or biopsies, or DNA, or images. It is complicated to decide which tests to run and expensive to run them.

Gathering information will become simpler and cheaper because we will use a smaller number of general-purpose diagnostics. For example, we will use a “whole genome sequence” that lasts a lifetime and replaces many different “panels” that test for various genetic characteristics. Instead of many blood tests for specific biomarkers, we will have one array of sensors that can see many different biomarkers and conditions.

Image for post
Hanna Rennert — Validation and Implementation of Whole-Exome Sequencing

A fascinating example of this genre is “digital pathology”. The current practice for diagnosing breast cancer is to take a biopsy and send it to a pathologist, who looks at it under a microscope. If the sample looks dangerous, they send it for some chemical tests (IHC). In the digital version, the pathologist takes a picture of the sample and sends it to a computer. The computer, after training on millions of images, can help the pathologist how dangerous the tumor might be. The computer can ALSO detect characteristics of the tumor that humans can’t see, that would normally require additional IHC tests. One picture and some software can replace a series of more expensive tests.

Image for post
From “Epithelium Segmentation”, Bulton et al

Computers can use similar techniques to detect conditions using only phone sensors. They can look at a picture of a face, listen to a voice (“how do you feel today?” asks the bot), and feel motion and gait. That service is free, all day every day.

Machines can manufacture custom treatments at low cost and high volume. Making glasses is a simple example. In the US, a routine eye exam will often cost $100, followed by $200 to buy new glasses. This process can be automated so that eyes are measured by an inexpensive machine. The measuring machine can then send instructions to a milling machine that grinds the lenses for glasses. In this case, automation and custom manufacturing can save a few hundred dollars.

The payoff can be a lot bigger for complicated problems like cancer. A “personalized neoantigen vaccine” is a new and expensive cancer treatment. The machines for creating this treatment can theoretically fit into one room. One machine will find the DNA sequence of a cancer, and understand the cancer and the patient’s immune system. Using this information, a different machine can synthesize molecules that stimulate the immune system to attack the cancer.

Over time, more and more drugs will fall into the bargain bin at your local pharmacy.

Image for post

US drug expenses have risen from $2.7B in 1960 to $360B in 2019, and drug spending is near $1T globally. Recently, cost increases are driven by higher prices for each drug. Vendors justify these higher prices as a way to pay for R&D spending that brings us wonderful new treatments. However, most of the price increases are on older drugs that paid off their R&D decades ago. The high prices on these older drugs are a consequence of payment complexity and regulatory capture (a fancy term for allowing the vendors to write the rules).

The solution to price inflation in these older drugs is to free the market. Countries with less powerful pharma lobbies have much lower drug costs. For example, a popular modern insulin called “Lantus” has a list price of about $320 in the US, and $33 in India.

Walmart is now opening clinics where you can pull out a few bills and see a doctor without insurance. They post fixed prices like $30 for an annual checkup, or $1 per minute to see a therapist. If you have a credit card, you can buy telehealth consultations for similar prices.

Self-pay systems are inherently cheaper than the bizarre system we have in the US, and probably cheaper than single-payer or capitation. They drive down costs by allowing customers to shop for value, and by cutting out the middlemen that have an incentive to raise costs.

This is a virtuous loop. When costs fall to the point where a lot of people choose to pay directly, that will drive further cost declines.

We can make a lot of progress with an effort to exploit the cheapest and most automated forms of advice, diagnostics, and treatment. This will set up a move to self-pay.

Care That Reaches You at Home

Cheap and automated advice will flow into the vacuum where doctors aren’t available.

We still need humans to care for sick, injured, and emotional patients. But, our established healthcare providers rarely go into the home. Human clinicians do a great job working with patients that come to a healthcare building. However, 99% of the time, patients are not in a healthcare building. That’s when automation can take over and apply continuous attention, too cheap to meter.

We are seeing a rapid development of remote care. This year doctors switched many consultations to “telehealth” — phone and video. Payers like the US government are agreeing to pay for at-home monitoring. New cheap and automated services fit into this remote delivery channel.

Humans can’t monitor all of the information that flows from at-home monitoring. They need automated agents to look at what is going on in real time, and alert humans about problems.

Personal Agents That Care

Cheaper, more available agents can work for patients as well as doctors. Patients are often inattentive to their own health, and poorly informed about care options. We will program software agents to be attentive, motivated, and able to tap a full global network of advice and services. Automated advisors can set up diagnostics. Then, they can indicate where and when their client should go for hands-on care.

When humans and their agents get smart about where and when to buy care, this will force changes in the healthcare industry.


Smart shopping and declining cost is likely to force the industry to become more specialized and concentrated.

Currently, community healthcare providers need to be ready to handle almost any problem that a patient can have. Diagnosing and providing every required service is expensive. It is also equalizing. Each provider out of hundreds or thousands does approximately the same thing. And, it reduces quality, because many treatments are handled by providers that don’t have enough volume and experience to drive them down the cost and quality curve.

When customers have their own smart diagnostics and agents, they will know where to go for treatment.

  • They will be comparison shopping, which will finally exert downward pressure on prices.
  • They will concentrate business. They will identify the providers with the best price and quality, which will tend to be the highest volume providers for that particular treatment.
  • They will fragment the industry into specialized providers.

Insurance companies will also be affected. If daily care becomes cheap enough, we won’t need the complicated insurance and payment schemes that we have now. We will use “catastrophic” health insurance to pay for expensive, hands-on treatment.

Healthcare is a huge industry in the US and many other countries. If health care starts becoming cheaper, it will drive changes in the industry that are so big that they will change the overall economy.

Ideally, we will need all of these services less frequently because we will be healthier.

Previous articleCan Startups Disrupt Medical Errors?
Next articleArtificial Intelligence, Leadership, and Talent Management
Andy Singleton
Andy is a software architect, CEO/CTO, and student of innovation. He is currently working on HumanDB, a new architecture for delivering automated medical advice, and Unbundled Fund, a better investment structure for digital private markets. He is the founder of Assembla, a SaaS collaboration company, PowerSteering Software, an enterprise SaaS company, Creation Mechanics, an evolutionary algorithms company, and employee #2 at SNL Financial. He has an undergraduate degree from Harvard.


Please enter your comment!
Please enter your name here