Funnily just as I was joking about the website from the IT crowd where you enter all your details and it gives you your deathday (3pm Thursday if you don’t eat your greens, also the motherless ovens is driven by Scarper Lee’s impending deathday), I’ve found out that we can all access the system the GPs use to calculate our risks of cancers or heart disease (caveat caveat, don’t use it without your doctor…). The GP one sits inside their system and can safely use your data on your GPs database. The one we can use just sits on the website and you have to enter you data yourself, but it’s a really simple one page checklist. On the symptoms checklist, out of 19 options, blood comes up 7 times depending on where it is. It’s as if random bleeding is a bad thing. The postcode is used to estimate your deprivation. It could be an idea to compare the numbers you get with the numbers your GP gets. If your numbers are lower perhaps the GP is missing some family history. Also, who would have thought that difficulty in swallowing is more of a risk factor than a family member with type 2 diabetes and a family member with breast cancer put together? Who even knew difficulty in swallowing was a thing? These three risk factors give you a sprinkling of frowny faces on the overview panel if you are 68, but not if you are younger. The moral of the storey is that staying young is the best way to avoid cancer, it’s also good for Hollywood careers, so really, we should all follow Orlando’s example (not Bloom).
Though I was a bit dismissive of caveats above it’s interesting to note that though the system was built using (anonymised) data from 2.5 million people, it is still only sensitive enough to give meaningful risks for the ‘big 10’ cancers (the ones on the list when you click calculate) and it’s not good for rarer cancers (yet, adding more data will help). Also the tool is created with data from 25-80 year olds, so is only suitable for those age groups (I’m pretty sure bleeding randomly is bad whatever your age though! Get that checked out).
My impression of how useful a tool like this is, is influenced by a tool to predict genetic causes of diabetes (e.g. MODY). While the overwhelming majority of diabetics are now Type IIs and Type Is are next (their pancreas has been knocked out, most likely by the immune response to a mystery virus), there is a small category of people whose diabetes is caused by a mutation, and if you can identify these patients, it may change the way they are treated. For example some people just have a tiny stunted pancreas. Giving these people drugs to squeeze out more insulin from the pancreas probably won’t work and they should probably go straight onto insulin. Some people have a permanently higher blood glucose, but the body works perfectly well to maintain that higher level. A bit as if your body thought your body temperature should be 38 degrees. It is crazy hard to bring these levels down, and luckily it seems to be unnecessary, these people seem to be at no greater risk of long term conditions. so these ‘diabetics’ need to not be treated. And best of all is the condition setting in before six months (before you have a functioning immune system that could kill your pancreas) where you have the blood glucose sensing system, you have the insulin producing system but they are not connected. They can be joined back together by massive doses of sulphonylureas (basically a pill) which will eliminate your need for testing your blood sugar and injecting insulin. Almost a ‘cure’ for this tiny minority of a tiny minority. The point about algorithms here is that Prof. Andrew Hattersly who has led the research into these conditions, who can safely be considered an expert on genetic causes of diabetes has tested himself against the algorithm he built and he found the algorithm can guess better than he can. So swing on over if you’re diabetic and have family member with diabetes and want to review yourself.
So with this endorsment for one algorithm, I’m inclined to believe a different one can probably at least flag up useful pointers.
This can further be tied to the outguessing machine described by Poundstone in Predicting the Unpredictable. The machine was built in a pre digital age (the ’50s) and by simply remembering whatever you guessed the last two times a given option came up (with a 16 bit memory), it could consistently outguess anyone. Because (from Thinking Fast and Slow) multiplying lots of small numbers in our head is not a major human skill, so leave it to the algorithms.