Healthcare as a Digital Good
To solve healthcare's biggest problems we need an aggressive push towards digiization and automation
In The Second Machine Age, McAfee and Brynjolffson write:
“When things are digitized—when they’re converted into bits that can be stored on a computer and sent over a network—they acquire some weird and wonderful properties, they’re subject to different economics, where abundance is the norm rather than scarcity.”
They were thinking of VHS tapes turning into YouTube and books turning into eBooks.
The Magic of Digital Goods
The economics of digital goods are pretty magical.
Consider the humble book.
Before a hardcopy book can be read it must be printed, bound, shipped, stacked on a shelf, bought, and carried home. While that copy is being read no one else can read it. At any moment it could be water damaged, torn, faded, misplaced, and eventually trashed. Each of these steps themselves require thousands of ingredients: printers, ink, trucks, truck drivers, bookstores, cash registers, landfills and so on,
Contrast that with an eBook.
An essentially infinite number of copies can be made at a virtual marginal cost of zero. It can be sent to Lagos just as quickly and freely as to Guangzhou, Tehran, or Buenos Aires. The same eBook can be read by an uncapped number of bookworms without diminishing the number of copies. An eBook can be deconstructed and recombined in creative new ways, The text can be incorporated into blogs, fanfiction, tumblers, memes, and Twitter threads. Self-publishing platforms take off. Barriers to entry are slashed, everyone’s a reader, everyone’s a writer.
These features of digital economics can be summarized as follows:
Non-rival: consumption does not diminish supply
Ubiquity (aspatial): goods transcend space and time
Recombinant: goods can be deconstructed and recombined
Immaterial: fewer material ingredients, asymptotically low marginal costs
If you’d told an economist 50 years ago that you had a good subject to these dynamics they’d think you were a wizard. But here we are.
Next, let’s look at what these “weird" and wonderful” dynamics mean for healthcare.
A Match Made in Heaven
The strengths of digital economics align beautifully with the weaknesses of modern healthcare, namely poor access, ballooning costs, and slow innovation.
Firstly, healthcare can be divided broadly into Tangible and Intangible elements:
Tangible: things we can touch. Things like scalpels and incisions, casts and slings, pills and injections. Think of this roughly as Hardware.
Intangible: things we cannot touch. Things like the clinical expertise of a doctor or nurse, exercise and diet plans, and explanations and reassurance. Think of this roughly as software.
In the short-term, Tangible elements are difficult to digitize. If you have a cut that needs sutures you have a decidedly analog problem.
In the short-term, Intangible elements are most amenable to digitization. Clinical expertise and decision making can be codified and automated. These aren’t trivial tasks but they’re within reach.
Let’s look at an example.
Illustration: Medical Imaging
Medical imaging is an integral part of modern medicine. Like any piece of healthcare it has Tangible and Intangible elements.
The Tangible elements of medical imaging are things like MRI machines, ultrasound probes, and sonic gel.
The Intangible elements of medical imaging include the digital images, and the clinical expertise, and visual reasoning skills of a human radiologist.
How does the current, setup look like?
Analog Version
Radiologists are doctors who interpret medical images like X-rays, CT scans, and MRIs. They spend four years in college, four years in medical school, and five years in residency to become general diagnostic radiologists. They then spend an additional one to two years to become radiologic subspecialists like neuroradiologists or cardiothoracic radiologists. All the while they incur debt and opportunity cost. To justify this herculean effort, radiologists are hired at a cost of around $500,000 per annum.
Having spent that time, energy, and money we can finally start interpreting radiographs.
Radiographs have to be interpreted individually. Depending on the complexity of the study this can take some time. Radiologists are human so they require food, drink, and rest. They may be able to work, eight, ten, or twelve hours a day but that still leaves more than half the day unserved. To err is human and invariably even the best radiologist will miss a critical finding. Of course not all patients have access to the best radiologist. And some parts of the world, say my home of Khartoum, there may be no access to a radiologist at all.
Now the digital alternative.
Digital Version
An automated medical image interpreter would obviate the entire radiologist training process: no medical school, no residency, and no fellowship. Accordingly: no tuition, no decades in school, and no labor costs.
Once built, a new instance of a robot radiologist could be created instantly at zero marginal cost. We would go from 100 radiologists per million in the U.S. to 1 per image. It would interpret images in real-time, simultaneously, or in bulk. Indeed, we could have X-ray machines that output findings rather than images.
The cost of medical imaging would plummet and access to imaging would soar. We’d generate enormous radiographic data which could be cross-referenced against patient records, fed back to the algorithm, and the robot continuously and iteratively improved. All patients would have access to equally skilled interpretations, any time of day or night, virtually for free.
Using the framework above:
Non-rival: simultaneous interpretation of tens of thousands of images
Ubiquity: access to the best interpreter instantly, anywhere, anytime
Immaterial: no reading rooms, no reading monitors, marginal costs of zero
Recombinant: self-enhancing interpretations, new tools and use cases
Fantasy?
The digital version may seem farfetched but it isn’t.
First, computer vision and visual reasoning has already surpassed humans, including radiologists, in many tasks. Next, we needn’t replace all medical interpretation wholesale. We can start with a single modality (e.g. X-ray), a single study (e.g. Chest X-ray), a single view (Posteroanterior Chest X-ray), or even a single finding (e.g. Pleural Effusion).
We’ve already seen the fruits of simpler digitization. Picture Archiving and Communication Systems (PACS) mean radiologists can be hundreds of miles away from the scanner. The beginnings of Ubiquity. Speech recognition and dictation technology has meant reports are digital, less error-prone, and faster to turnaround. The beginnings of Immaterial.
More digitization, gives us exponentially more of these fruits.
Summary
Digital goods have a weird and wonderful economics. They are non-rival, ubiquitous, recombinant, and immaterial. These features correspond nicely to some of the biggest hurdles in healthcare: poor access, ballooning costs, and slow innovation. Healthcare can be broken into Tangible and Intangible elements. In the short-term, Intangible elements are most amenable to digitization. We should make aggressive efforts to digitize more and more of healthcare’s Intangible elements.