Risk, Reality, and Healthcare Disruption: Separating Hype from Impact

Physician colleagues and epidemiologists reading this NY Times article understand something that engineers have a hard time with; uncertainty and probabilities in medical reasoning.

Physicians also understand that early detection does not equal prevention, as this article suggests. There is an inherent tension between screening tests that have the potential to find a treatable disease early, or when screening finds incidental findings that leads to unnecessary testing and iatrogenic harm, or when the screening test finds a “cancer” that will never harm you.

For non-physician readers, iatrogenic harm would be when your medical professional orders a test or procedure that harms or kills you. For example:

1. The medical startup screening MR scan reveals a concerning spot on your kidney.
2. The doctors affiliated with the company order a biopsy because they are concerned about cancer.
3. An artery is punctured during the biopsy.
4. The bleeding leads to surgery.
5. Then you die from the surgery.

All for a benign spot that you did not need a biopsy in the first place. It happens more than you think.

I understand that basic healthcare epidemiology and topics such as pre-test likelihoods, sensitivity/specificity, and lead time biases are not part of engineering or founder thinking. It should be, because repeatedly we see engineers saying they are going to fix healthcare through early detection. Learn here about the problem of overdiagnosis.

To be clear, I’m dermatologist with a career focus on information technology and a believer that AI/ML and large language models  (LLMs) will greatly improve the recognition of skin cancer by non-dermatologists. The idea of full body photography as a tool to detect changes in nevi evolving towards melanoma, or recognizing new melanomas, remains incredibly promising. These evolving technologies should be used, but only in those with a risk factor for skin cancer. Full body photography for everyone makes sense when the technology is 100% accurate and extremely low cost. When you screen people at low risk, then false positives soar! Bad things can happen to people and you can drive up costs—despite your good intentions.

What is the likelihood of someone that says they are going to “disrupt healthcare” actually will improve the lives of many?