Where's My Health Tech?

Imagining What Could Be

For the first 24 years of my life, I almost never got sick or injured. Perhaps I’d be out of commission for a little bit once every few years. It’s easy to discount just how much of an advantage this can be in both school and work: to truly retain all your hours to dedicate towards whatever you want or need. I attribute this to remarkable good fortune, a sleep schedule verging on religious devotion, and a childhood aversion to getting in other kids’ germy business.

In the 25th year, however, this changed. I started falling ill perhaps every other week. I found myself losing enormous time to lying in bed hoping for recovery, and to avoid vanishing from the world completely, undertaking a lot of activities and responsibilities at only 50% capacity. It became obvious that the most important determinant of my near-term happiness would be whether I could resolve this issue.

Searching for Answers

First, I figured it was time to schedule my annual doctor’s appointment and see if I could get any advice on possible causes or things to investigate by describing my approximate symptoms and life circumstances. Unfortunately, this was a total dud. Although the hospital was ingenious at describing my single visit to insurance as 4 different elective visits with thousands in out-of-pocket expenses, I was given precisely no useful information for my health, instead left to puzzle through it on my own.

Still, even alone this shouldn’t have been a particularly difficult search. I’d had a lot of life changes which lined up well with these changes in health status, and these were probably somewhat causal. Here are a few:

• My environment had changed. I had moved from the East Coast (New York) to the West Coast (California), with its warmer weather, reduced rainfall, and similar.
• My residence had changed. I went from an NYC high-rise apartment to living in my parents’ house for a few months, followed by renting a townhouse.
• My work had changed. I quit my quant job to work at a startup, which changed offices once as it grew, although neither had great ventilation.
• My eating had changed. My old cafeteria gave way to less healthy delivered work lunches, and my dinners adapted to shopping mostly at Costco rather than Whole Foods based on proximity.
• My exercise had changed. I initially swapped Pilates studios to one more focused on strength, but then cancelled my subscription when I started getting sick more often and would need to bail on classes at the last minute.
• My lifestyle had changed. Among other things, I walked much less, losing my daily commute over the Brooklyn Bridge, but had tried to bike more rather than converting it all to driving.
• My socializing had changed. I had left behind my weekly D&D group among other NYC activities, but met new folk here, including getting into a relationship, and spent more time with my family.

Any of these things and more could’ve produced the sort of physical or even psychological change that might harm my health. I just had to figure out which one of them so that I could fix it.

Technological Solutions

The fact that so many things had changed around the same time gave me a lot of possible candidates to sort through, but they were not inseparable. My move to California, the start of my new lease, the cancellation of my Pilates subscription, the swap to a new office, and more had all happened at discrete times and even varied in how much they should impact me throughout the week. I was coming from a trading firm where we spent vast energy searching for subtle signals in data that might be predictive of price moves. Surely I could handle this sort of simple analysis.

Now, my memory is not that great. I couldn’t recall exactly which of these milestones had precededmy first noticing that I was falling sick more often. Even if I did have perfect memory, the low resolution of the data meant that single date might not be conclusive. Still, I had the upper hand here, or so I thought.

You see, when I’d quit my job, I needed to quickly spend the remainder of my health-dedicated FSA account money, and so had bought myself an Oura Ring, self-advertised as a “Smart Ring for Fitness, Stress, Sleep & Health.” For multiple years now I’d also been wearing an Apple Watch, describing itself as “the ultimate device for a healthy life.” This was the moment they’d been waiting for, their chance to shine and use all the data they’d been collecting in order to provide real value. Right?

Well, not so fast. I quickly came to find out that although their respective apps had been well optimized for gamification, making sure you’d check back each day on the number of minutes you slept or exercised for, they were not nearly as suited for this sort of retrospective analysis. I could barely find any correlation between their summary stats and my perceived well-being, let alone try and trace back to causes.

A Detour Through Data

Although I was disappointed to be so thoroughly failed by some of the biggest health tech out there, I was perhaps not entirely surprised. Supporting real health would be riskier legally, and not as obviously profitable. Nonetheless, I had an out. Oura and Apple Health both provide a way to export their raw data, likely due to the GDPR “Right to Data Portability”.

I should note here that Oura tries its best to make this painful and unusable: phrasing the export as a “request” which they might honor in around 10 business days. Therefore, my initial efforts had to be exclusively from Apple, whose data I fed into Claude to help me with some basic analysis.

The results here were honestly a little surprising and disappointing. For the most part, the basic health markers recorded did not seem to show an obvious spike in the time period I experienced it, or otherwise to correlate nicely with my perceived well-being. After some finagling, the best individual predictor turned out to be sleep duration: I tended to stay in bed longer (generally celebrated on health metrics) on days I felt sick.

Of course, this was the wrong direction of causation. I was staying in bed to try and recover because I felt sick. Staying in bed was not truly “predictive” of when I would get sick. Perhaps in addition to being hard to get and analyze, the data these gadgets provided was still insufficient to draw good conclusions?

Going Low Tech

And so, ultimately, I went low tech. I created a basic spreadsheet, where each day I logged whether I had felt sick, how sick I felt, and what kinds of symptoms I had. I supplemented this information with various columns representing signals that might help me tease apart the different causes.

• Wondering whether I had insufficient time outdoors, I started taking vitamin D pills, and logged whether I had done so.
• To determine if my office might be harmful, I distinguished weekdays and weekends, and marked days when I worked from home.
• Suspicious of my new apartment, I slept at my childhood home some days, recording when I had done so.

Manually collecting data can be time-consuming, so I did not pursue all hypotheses in detail, limiting it to just what seemed easiest or most likely. What makes this all much worse is that the data is inevitably low resolution: I cannot make clones of myself to run multiple experiments, or speed up the pace at which I catch and then recover from sickness. I can’t even be trusted to self-report with any degree of precision!

Nonetheless, after now months of monitoring, my data seems to suggest my apartment as the most likely culprit. For the time being, I am concluding the experiment and moving myself out of it. I am hopeful this means the whole episode is behind me and that I shall be back to my usual low rate of sickness. Still, I can’t help but wonder: could this all have been better? In an alternate world, could I have been more confident in my conclusion, or even have to it months earlier to save myself the pain?

What Could Be

Perhaps I’m an idealist, but it feels to me like we could do much better. I can imagine health tech which allowed for and even helped with this kind of analysis, to understand what actions and environments might make me feel better or worse. It feels like that could be joined with all sorts of other data not just from a health tech device: if I stay at home all day besides visiting a pharmacy, could we make inferences from that, or if I wrote about discomfort in my journal? It’s even possible that such investigation could be going on constantly in the background, tipping me off to things which might be harmful before rather than after I notice them.

I know that some cursory analysis is already available, and that perhaps it is hubris to believe what these corporations already provide is not pretty close to the best it can be. Still, I guess I just feel that if we put in a fraction of the effort we spent on tracking what causes someone to click on an online ad instead into tracking what causes people to feel well or badly, we could do so much better.