Idea for heart attack / stroke prediction

neeb

I was just thinking yesterday that increasingly, a lot of people are recording their training figures (HR curves, power etc) on a regular basis and publishing them on Strava, month after month and increasingly year after year. With thousands of people doing this (for running as well as cycling) that represents a very large and accumulating body of data. Some of these people (hopefully a very small minority) will doubtless go on to have heart attacks or strokes within a short period after posting their last Strava upload. If that data was available, it should be possible to statistically evaluate any trends in training data that are correlated with a high chance of a sudden health issue. The more data available and the longer the periods (years, decades even), the more accurate any such associations would be likely to be.

Obviously the sort of data that Strava records is not the best type for evaluating risk of heart attacks or strokes, but most people at risk don't know they are at risk and so wouldn't have any reason to be tested "properly". The correlations between trends in Strava data and risk wouldn't need to be functionally explainable, there would just need to be associations with high statistical significance.

So what I am envisaging is that Strava might inform you if your pattern of data started to look statistically suspicious, and you could then go and get checked out. Of course it might turn out that no significant patterns existed, but it seems reasonable to assume that they might.

imposter2.0

Sounds great - then we can all sue Strava when they fail to spot an underlying health issue or pre-existing medical condition.

ai_1

Nice idea but I think it's very likely there would be far too much noise in the data for this to be useful.
Chances are you'd produce a load of false positive indicators and/or fail to predict actual problems. Despite any disclaimers, people who's data hasn't triggered a warning will be less likely to get checked properly if they have minor concerns as the data will help them justify just ignoring it.

I think to do this responsibly you'd need a level of demonstrable correlation that's impossible to produce. HR is effected by far too many factors that are not recorded in this data. You might have speed and gradient but you don't have a record of wind speed and direction, whether you were drafting, alcohol consumption (major impact), sleep patterns, etc. If I understand correctly the main way in which HR can indicate a heart problem is if it doesn't track smoothly against effort - I suppose if you had a power meter this could be done. However the main data used to predict heart risk is blood pressure and ECG neither of which you would get from bike computer data.

neeb

Just an idea for people who have more knowledge than me to think about.

Someone would need to analyse the data to see if any demonstrable correlation emerged. I get the point about HR and effort, perhaps it would only work if both power and HR data were available, but give it 5 or 10 years and power meters may be fitted as standard to high-end cranks/pedals whatever.

Other factors that affect data on a daily basis (sleep, alcohol) could average out, but then the task would be to find patterns that emerged over a period of weeks rather than days.

Maybe an idea for the future rather than right now, when more data of higher quality is available and we are all multiply socially and administratively networked... Perhaps the algorithm could even factor in an estimate of lifestyle stress based on Facebook posts.. ;-)

ai_1

neeb wrote:

....Maybe an idea for the future rather than right now, when more data of higher quality is available and we are all multiply socially and administratively networked... Perhaps the algorithm could even factor in an estimate of lifestyle stress based on Facebook posts.. ;-)

Oh dear. I have visions of insurance companies monitoring your online data to decide what premiums to charge! Actually that doesn't seem so far fetched, it's no worse than a lot of the "security monitoring" that already happens.

cougie

neeb wrote:

I was just thinking yesterday that increasingly, a lot of people are recording their training figures (HR curves, power etc) on a regular basis and publishing them on Strava, month after month and increasingly year after year. With thousands of people doing this (for running as well as cycling) that represents a very large and accumulating body of data. Some of these people (hopefully a very small minority) will doubtless go on to have heart attacks or strokes within a short period after posting their last Strava upload.

I don't really see that there would be much of a link you know. It's just isolated data. You know so little about family history or diet or medical history. I doubt very much that you can say the risk of a stroke increases after such and such.

neeb

cougie wrote:

I don't really see that there would be much of a link you know. It's just isolated data. You know so little about family history or diet or medical history. I doubt very much that you can say the risk of a stroke increases after such and such.

The point is though that it's not isolated data - it's data that exists within a much larger body of data (for that individual over time as well as for thousands of other individuals). Medical history / diet / genetics etc is not relevant here, all that matters is that if you are training (especially with power), you have a very detailed record of how your heart rate is responding over time to effort, including things such as variations in heart rate recovery, plots of heart rate against power etc. It may be that there are strong trends in that data at the population level, e.g. (just to take a completely hypothetical example) perhaps people with higher than average maximum HRs between the ages of 49 and 54 who experience a rapid change in heart rate recovery after anaerobic efforts over a period of three weeks are statistically very likely to have a stroke... or whatever. You wouldn't find any such correlations until you looked for them and were able to statistically verify them using a very large body of data.

imposter2.0

You're assuming that HR data has any meaning beyond the individual it relates to - which it doesn't.

neeb

Imposter wrote:

You're assuming that HR data has any meaning beyond the individual it relates to - which it doesn't.

Of course it has meaning beyond the individual given a large enough sample - we are all human and we don't vary so much that these variations won't be normally distributed if you have enough data. This would actually be the power of the method (if it worked), to be able to detect correlations that might not be apparent (because of individual variation) if you looked at a smaller sample.

imposter2.0

neeb wrote:

Imposter wrote:

You're assuming that HR data has any meaning beyond the individual it relates to - which it doesn't.

Of course it has meaning beyond the individual given a large enough sample - we are all human and we don't vary so much that these variations won't be normally distributed if you have enough data. This would actually be the power of the method (if it worked), to be able to detect correlations that might not be apparent (because of individual variation) if you looked at a smaller sample.

No, sorry, it means nothing. Without knowing everyone's MHR (which can vary significantly between individuals anyway), you will have no way of knowing how hard anyone is working at a given HR.

shockedsoshocked

Not strictly related to the OP, but Strava is being used to assess training methods of 'successful' cyclists. Check out - “Engine matters”: a first large scale data driven study on cyclists’ performance.

neeb

Imposter wrote:

No, sorry, it means nothing. Without knowing everyone's MHR (which can vary significantly between individuals anyway), you will have no way of knowing how hard anyone is working at a given HR.

You are missing the point - it doesn't matter how hard anyone is working at a given HR, if it happens to be the case, say, that people who have a HR within a certain range for a given power output for a given training profile (or whatever) and whose figures then change in a certain specific way then go on to have a heart attack with a certain high statistical probability. If there are correlations you can use them as predictors without knowing what is actually going on at a physiological level. Just like insurance companies might up your premium for having a red car without being able to explain the psychology behind car colour and a statistically higher accident risk.

imposter2.0

ok - so it's about as useful as using car colour as a predictor of accidents? That makes much more sense...

Bozman

I had a minor AF moment before Xmas. I could feel that my ticker wasn't quite right for about 5 minutes and then my heartrate spiked at 250bpm for a couple of seconds. My Garmin picked that up and it was quite an impressive spike but there wasn't a hint of the irregular heartbeat before the spike.

cougie

Are there any apps that record heartbeat that could spot irregular heartbeat ? You'd need to be able to change the scale to look at the detail - ideally at rest to see ?

Bozman

cougie wrote:

Are there any apps that record heartbeat that could spot irregular heartbeat ? You'd need to be able to change the scale to look at the detail - ideally at rest to see ?

I doubt it.
You can see it on an ECG if you're lucky enough to catch it, you can feel it, with me it feels like a car running on three cylinders but it didn't effect my heart rate.
That was the first time that I've had it while I've been out with a heartrate strap, I was interested to see just what was recorded. ....... Nothing a part from the final spike.

coulddobetter

Arrhythmias cannot be diagnosed via a basic cycling heart rate monitor . As a coronary care nurse who has been involved in multi national clinical trials forget strava and leave it to the professionals

Mad_Malx

ShockedSoShocked wrote:

Not strictly related to the OP, but Strava is being used to assess training methods of 'successful' cyclists. Check out - “Engine matters”: a first large scale data driven study on cyclists’ performance.

This is interesting, thanks Shocked.

neeb

coulddobetter wrote:

Arrhythmias cannot be diagnosed via a basic cycling heart rate monitor . As a coronary care nurse who has been involved in multi national clinical trials forget strava and leave it to the professionals

Just in case it's not clear, I certainly wasn't suggesting that.

ugo.santalucia

I'd say ask your GP for a health check... a blood pressure exceeding the norm combined with a high LDL cholesterol and/or glucose level are probably a better indicator than heart rate

canoas

ugo.santalucia wrote:

I'd say ask your GP for a health check... a blood pressure exceeding the norm combined with a high LDL cholesterol and/or glucose level are probably a better indicator than heart rate

I'd say this is a sensible way of approach to avoiding a HA or stroke.

I suffered a Heart Attack only a few weeks ago, I was one lucky person I can tell you, to survive!
Had blood tests one year ago, results were fine. My resting HR is about 45-47 bpm. I had a main artery that was blocked, all other arteries were fine/normal and clean. Now I have a stent in the bad one. I'm 6 foot and weigh 73kgs and eat healthy.

But I did suffer from hypertension, reading of 140/85 or 130/80...so maybe this was an indicator, but Doctor use to say you are too fit and nervous must be wrong reading! An ECG is a good test also.