Is AI useful for health information?

In her new bestseller Good Energy, Dr. Casey Means makes a novel assertion about conditions ranging from depression to stroke. She says the same metabolic problem (the failure of cells to make sufficient energy) is the root cause of every single one. She goes on to explain that unfortunately for patients in the United States, the health care system is designed to treat the organ-specific consequences of bad energy, not the source itself. As a result, “the more we spend on health care, and the more we work as physicians, and the more access to health care and medications we provide to patients, the worse outcomes get.”

In fairness to the current health care regime, a system designed around 42 medical specialties could work … if bulletproof excellence existed in each area. Where do we stand? New technology now makes it possible for a simple heath hobbyist like me to evaluate the medical profession’s overall competence.

Artificial Intelligence platforms — I mostly use ChatGPT — work by analyzing billions of pieces of human-written content (books, news articles, scientific journals, podcast transcripts, web pages) posted on the Internet. To answer questions, ChatGPT doesn’t access a database of facts, but instead responds based on patterns it sees in the data. This computation means that if you ask ChatGPT a medical question, the chatbot’s response will be a summary of: the materials used to teach medical professionals, what these practitioners say publicly about the subject matter, and the conclusions they have published.

I recently came across a pee-in-your-pants-good Substack essay about ultraviolet blood irradiation (UVBI). UVBI is the process where doctors expose a small quantity of a patient’s blood to ultraviolet light and then infuse the blood back into the body. Only a small amount of blood needs to be irradiated for treatment because the body conducts light inside. (Next on my reading list: The Body Electric). UVBI is so miraculously effective against everything from cancer to polio that the American Medical Association was compelled in the 1950s to commission bogus studies to discredit the procedure — the profitability of the entire medical establishment was at risk. Nevertheless, research continued in Russia because that country’s financial condition required that health care spending return actual results. Breakthroughs have been reported related to bacterial infections, autoimmune disorders, and psychiatric conditions.

For my purposes, I asked ChatGPT if UVBI had ever been considered as a treatment for plantar fasciitis. Tendon and fascia injuries take so long to heal in aging athletes because of the poor blood flow and lack of healing factors delivered to these tissues. UVBI is thought to be an effective intervention against many illnesses in part because of the way it improves circulation.

ChatGPT started off reasonably enough — it responded that UVBI is not a recognized or proven treatment for plantar fasciitis. Then, everything went off the rails. “Plantar fasciitis is a common condition characterized by inflammation of the plantar fascia,” the chatbot wrote, “a thick band of tissue that runs along the bottom of the foot.”

In reality, plantar fasciitis (more accurately: “plantar heel pain”) is a result of tissue degeneration, not inflammation. When you look at unhealthy fascia under a microscope, you will see no signs of inflammation.

ChatGPT went on to describe eight “standard treatments” for plantar fasciitis. They were:

1) Rest and ice
2) Nonsteroidal anti-inflammatory drugs (NSAIDs)
3) Corticosteroid injections to reduce inflammation

Ice is contraindicated in cases of inflammation and irrelevant in cases like this of tissue degeneration. In addition, while a change of exercise habits is useful to allow for healing, complete rest will delay fascia remodeling. Anti-inflammatory drugs will likewise be pointless. In addition, steroid injections into the plantar fascia will do nothing to address tissue degeneration, but they will increase the chance of heel fat pad problems later in life.

4) Stretching: “Specific exercises to stretch the plantar fascia.”

Stretching already overstretched fascia is a terrible idea.

5) Orthotics: “Using arch supports, orthotic devices, and wearing shoes with good support.”

Orthotics and supportive footwear prevent your foot from gaining strength in the vulnerable areas and will prolong healing. More appropriate would be a recommendation for wide toe box, zero drop shoes, and barefoot activity.

6) Extracorporeal Shock Wave Therapy
7) Surgery

These are expensive in-office treatments (ESWT is not covered by insurance) that have a high rate of failure.

8) Physical therapy: “Professional guidance to perform exercises and techniques that help alleviate the condition.”

YAY!

In conclusion, according to AI, you have a 1-in-8 chance of finding a doctor whose direction will do more good than harm. Put another way, only 12.5% of what a doctor tells you is worth doing.