A Decade Of Tracking Headaches

So, this wasn’t my first choice in science, exploring the wonderment of life. I stumbled into tracking, it took 14 years to get into, and the London Migraine Clinic was a big influence. They started me on it and it went against the prevailing culture about not making a fuss about health issues. So, it took me until 2015 to properly join the smart phone revolution and what a relief that was. So after 10 years I have lots of data.

The thing that’s important about this is the interventions; the actual definition of a headache is difficult and the significance of headache is difficult. The interventions were all single doses of triptans, a powerful anti-migraine drug and this was a proxy of the severity and the duration of a headache, so it’s great for analysis. I also had lots of data on possible triggers over about seven years, and these did help. They gave me a feel for the problems caused by dodgy behaviours. But as a part-time statistician I struggled with not knowing when these things happen, and I didn’t have a headache, so the data was very much lacking.

Last year, at the conference in Amsterdam, one thing that struck me was how usefully the time was being used, and this got me thinking and made a lot of sense. Everyone has lots of information between their ears, and they have this structure of a model of our world and this is based on causality, which has got a primary key of time. So, I wanted to exploit these things. How do we look up using time?

My Apple database has the number of seconds since 1970, it was just not useful. But our brains need associations, so I’ve thrown a table together just to illustrate it. I wanted to tackle my headaches like this. If you look at the biggest variant there, the decade, the graph shows how it went over the decade. In the early years, any medical professional would say you don’t really want to be doing this. You are flirting with habituating yourself to a drug. Luckily, thankfully it got better after that.

So, if I map my life circumstances onto this period, in 2008 I started quite a demanding IT contract in Brussels, and eventually, in 2011 I became more useful and negotiated myself onto local hours, a 10% reduction. And then the work fell off in 2013 in the Department and I became self-employed in 2014. So the self-employment I should say is 50% IT work and the rest renovating a flat and doing a few manual jobs. It’s more of a variety. So I went down, I took the first and the last periods of these two life circumstances and I dug deeper.

I went down to the breakdown by month and there was no clear pattern in here. And the sample size was very small. But when I went down to week days, I started seeing some patterns. So, the top one, the contract hours shows a big stress relief nature to the headaches, whereas in the bottom I just spread my work over six days much more evenly. I will do things in the middle of the week like go shopping or do some manual work, so it’s a much more even pattern of work which seems to be corresponding to that.

And then if you break it down by hours into a day, it had two distinct patterns again. So in the contract hours, I think this is to do with being forced to medicate by external circumstances. Even at the weekend I would go to French lessons or on a Sunday I would really value a day off, so I would make sure it’s a nice day. While self employed, the constraints were maybe social events in the evening, or are trying to get to sleep in the evening, so I had more choice then.

Finally, I looked into the correlation of delaying the medication with the duration of headache and its success. To summarise briefly for less severe headaches, there was a correlation. The shorter delay to medicate, the shorter the duration and the greater success. It was less so for the more severe headaches. So, the basic problem here is delaying a medication might mean that I don’t need to medicate at all, so it has limited practical use.

So, what I learned from this was until I can find a physical cause for my headaches, I seem healthier when the edge is taken off the stresses of modern life. The second one is to medicate as soon as I know that it’s going to be one of those headaches when I need to. And a time span can turn a bi-variate series into a huge multi-variate one. However, we are a little too good at pattern matching, so need to watch out for confirmation bias. Probably lastly, a more general one; tracking brings ideas out into the open, so it’s harder to consider for example, illness a personal failing while addressing it scientifically.