Illness (word 'illness') prediction is not hyped enough

How and if we can tell that you will be sick tomorrow, and how you can do it already today yourself? PS. fonts with the same look of capital 'i' and small 'L' characters should be banned.

Illness (word 'illness') prediction is not hyped enough
Photo by Volodymyr Hryshchenko / Unsplash

We are getting better and better with predictions and forecasts, I would even argue in many areas the best is still ahead of us with the advent of AI and ML models. We measure and monitor our sleep, steps, weight, or ovulation among others.

Let's try to step up the game and predict an "illness" 👨‍🔬 #WCGW

Disclaimer - This is not medical advice, and if you are ill you should seek professional help. I do not have a medical degree and the information below is based only on my case - which might be very different from person to person.

Lately, I noticed that when I'm ill (COVID or other virus infection) my resting heart rate (RHR) increases by ~20% ... (to over 90 BMP from avg. of 74). This is an observation I made when I was in bed due to illness and Apple Watch nudged me to "stand up" to achieve my daily goal of standing hours. Great "health device" which is unable to notice I would like to, but not so much such or can.

This is the chart that initially cought my attention (exported from Apple Health as this app has a horrible UI, but it is a story for another time).

My average daily resting heart rate (from last 10 months), red marked are times of visible symptoms during infections (high temp, weakness, sore throat...).

This probably is not that surprising, but if we look a bit closer, zooming in at 3 months period:

Last 90 days. Red = visible signs of infection/illness 

In at least one case we can see average resting HR a lot higher than usual almost 2 days before the first symptoms, in the other case a day before. It looks like anything above ~80 average RHR should trigger a warning (in my case).

🪄<Still frame from Minority Report here>🪄

What if we can get a notification BEFORE we are ill?

Well, it is possible, even today... kind of. My solution without writing a new app and dealing with demanding API, lots of data, and securing it - was to do an MVP just for me.

It turns out Apple Shortcuts can utilize Apple Health data and in it, and we can create basic automatization! It is not super readable or compact - due to the generic interface of Apple Shortcuts but idea is to take an avg. RHR from the last 24h and comparing it to the same measurement from the last 7 days. If it is higher - notify me as it might be brewing infection/illness.

Apple Shortcut Link (you can use it to import automation).

// Warning long screenshot

Does it work?

Yes. Just don't forget to add a trigger in the Shortcut app to run it at least daily.

Is it correct?

I have no idea.
Time will tell. Once I will get a warning about the possibility of incoming illness I can get a bit more sleep, vitamin C, and be careful with my plans for the next couple of days.
In the worst case (false positive) I would get extra sleep and rest.
In the best case (correct positive) I would be a bit more ready for incoming illness.

Isn't HRV (Heart Rate Variability) considered ideal to detect "under the weather"?

Seems good as well. Due to limited testing and time on analysis, I wouldn't say it is worst or better than resting HR, but it seems resting HR (in my case) is much more stable.

FYI HRV decreases in case of being ill, and higher HRV is a positive thing (in opposition to a high resting heart rate which is considered bad).

See it yourself, same charts as before but with HRV in green:

Last 10 months
and the last 90 days
I hope that I don't have to remind anyone about "correlation doesn't equal causation" and it is always worth approaching any data and conclusions with a grain of salt.

What next?

🕵️ I encourage you to take a look at your data (if you have a smartwatch and monitor it) and see if you notice any trends in areas where you were ill. Maybe in your case, it will be something different as a better warning sign.

💪 There is A LOT of more work that can be put into this and yield very good results (data analysis, ML models etc.), especially in terms of tracing "feeling" vs. all quantitative data we have about our bodies and behavior.

📱 On a related note there is a Bearable app - it focuses on manual daily data input to find any correlation between our behavior and for example migraines or sleep quality.

👨‍🔬 There is Fitbit and a separate one from Oura research about COVID detection based on health data, which is great input on this topic. There are many scientific papers out there on HRV that concludes that it might be used as a marker to detect issues or illness - for example, this, or this. I haven't found any comparison between HRV and RHR as biomarkers for being ill - if you have please leave a comment.

📚 A professional term for a metric that can give a bit of insight into the overall state or condition is "biomarker".

💍 Oura is tracking with labels your illness, probably to train their models, and might release this feature in the future.

✍ if you have an opinion, advice, or some experience in that field - please consider leaving a comment.

💡"Foreknowledge" is such a great word. To know something beforehand, it can be calculated, and then we call it a prediction.

🤔 This is all great that there is research but I would like to have those tools in my hand right now, even if not perfect. I understand that this is a complex topic in a highly regulated area, and often results can be not conclusive enough to move forward with the global availability of those features.
But... I think we are ready to take the first steps, especially as the cost of false positives is not high (extra rest), and for now, a simple Apple Shortcut has to do.

I would love to see better automatization allowed in health-smart wearables, so defining such an alarm for RHR or HRV is not such a pain to do.