Preventative Diagnostics as Paul describes in #7 will really be the future - it's barbaric that we can only make a diagnosis when the disease has already manifested (in most cases). There are a few players in this space (Scanadu comes to mind), but it's seems like nano biosensors and the like are still very new technologies. Correct me if I'm wrong.
I'm not optimistic. I think pg's discussion of automatic diagnosis is a bit ill-informed.
For example, the recent trial that showed screening CT scans reduce mortality in lung cancer cost 250 million dollars to run. Even then, nobody is sure if it is even a cost effective measure.
It is difficult and costly to produce a screening test. It also takes many years to validate. Then there is the problem of what to do with the results - for example, if you are diagnosed with possible pancreatic cancer, the treatment is a massive operation to replumb your upper abdomen. 5% of people die because of the surgery alone, and the surgery costs a fortune.
Unfortunately a simple relationship like "find cancer early = good outcome" does not exist. There are incredibly high barriers for a startup developing diagnostic tests for screening. There is a good reason why the only people doing cancer screening studies are large government funded research consortia that can afford to wait 10 years or more to prove a result.
The example of Bill Clinton is misapplied - cardiovascular disease is really common, maybe 30% or more of people will get heart disease in western countries. We don't need to have a cool machine to screen for it, we need to risk stratify people with a few simple tests (ie ask them if they have a family history, check their cholesterol and blood pressure) and improve their risk factors (eat better, quite smoking, exercise, lower cholesterol etc). But then you are talking about modifying human behavior...
I live in Australia, and we are indoctrinated to check your skin for moles that maybe cancerous. There are claims that the high rate of early detection leads to higher survival rates[1].
My understanding is that early detection of bowel, breast and prostate cancer is relatively easy and produces good outcomes too.
There are radical ways to do early detection (sub dermal computers continually monitoring, etc etc) but there are ugly hacked solutions that just might work, too.
How much would it cost to build a toilet with a bowel cancer test kit built in?
Not sure what you are arguing here... if you are arguing that screening for cancer can be useful and saves lives, then I agree with you!
If you are arguing that a start up could have come up with a screening program for bowel cancer for example, then I don't agree with you for the stated reasons.
Breast cancer screening is not as useful as you would hope either. (http://dx.doi.org/10.1002%2F14651858.CD001877.pub4)
2000 women need to be screened for 10 years to save 1 life, with 200 initial false positives requiring biopsy. Also, I see lots of people diagnosed with breast cancer despite having mammograms.
Radical ways to do early detection are fine, but you have to prove that it works and that requires a lot of people for a lot of years and a lot of money.
Building a toilet with a bowel cancer screening kit built in is a form of behavior modification to improve uptake, and that is a great area for start ups to get involved in. pg was talking about something different however.
See, I don't think he was. "Ongoing diagnosis* doesn't have to mean new tests if you can make the existing tests radically cheaper and easier. Given that existing behaviour is always hard to modify it would seem sensible to try and piggyback on existing behaviour.
Toilets with cancer sensors that would check for bowel cancer everytime you go would be as about as "ongoing" as diagnosis can get.
Maybe toothbrushes could be modified to check for viruses in saliva.
I'm sure there are other easy tests that could be done if you have blood. There are obvious ways that could be integrated into everyday life (for women, anyway).
I've read some studies that showed dogs could be trained to smell cancer. Maybe people would pay to have their clothes sniffed (!) when they have them sent to the laundry.
I've previously suggested (on HN) the idea of payment companies partnering with food outlets and exercise software vendors to log the calories you are buying. That's a good input into diagnosis software too...
I'm sure there are a lot of other ideas - look for low hanging fruit and you can do radically better than the status quo.
System on chip PCR machine? The lowest hanging fruit would be miniaturization and better engineering of existing diagnostic machines. At the moment the medical diagnostic market is filled with overly expensive devices that could be easily made cheaper and more efficient (somewhere with a favourable patent/legal regime so you don't get sued to oblivion).
For insurance companies early detection of chronic illness can lead to a longer lifespan when old, which costs them more. Better you die fast after a certain age?
I look at medical diagnostics as though I'm looking at my production servers (a poor analogy in many ways, but lets go with it)
Watching charts on a production server, you see patterns over time - e.g you can see a weekend, or holiday quite easily. Keeping an eye on these, you can see when things start hitting bottlenecks and you need to do something to improve the situation.
When I get my bloodwork checked, its once a year - can you imagine checking in on your production servers just once a year?
I live in the Seattle area - Vitamin D deficiency is a huge (but not well known) problem in this area. I would pay real money to see a chart of my Vitamin D levels on a daily basis (without needing to draw blood!). I can then adjust my supplement dosage as needed.
I have a food allergy that is slow to flare up and slower to go away again - being able to check various levels of things in my system to track against food intake will help me find out exactly what foods cause what issues. The US food industry would pay a fortune to have this ability not be available!
