Gary has now had his Fitbit HR for about three months and, as promised, here is an update.
As I reported before, the device really does show how much variation there is in my activity levels. It captures steps taken, miles walked, calories burned, heart rate, minutes of activity and sleep pattern. The software that comes with it gives the ability to download data on a monthly basis, so here are some statistics from my first three months in a few key areas:
- Steps taken – ranged from a daily low of 5,500 to a highest of 31,000 in a day; average daily steps around 13,000
- Miles walked – lowest day 5.6 miles (9 km) with a highest of 24.5 miles (39.4 km); average of around seven miles (11.3 km) per day
- Activity levels, which are measured in minutes per day are: sedentary – up to 650 minutes per day; lightly active – between 50 and 430 minutes per day; fairly active – ranged between 50 and 260 minutes per day; very active – ranged between 0 and 143 minutes per day
- Calories burned – a low of 2,900 in a day and a high of 5,200; average 4,000
- Flights of stairs climbed – ranged from a minimum of 12 in a day to a high of 68, with an average of 24.
So, all very interesting. And through the software package you have the option to link up with friends and make all this nice and competitive.
To give my results some context the American Heart Association (AHA) recommends that individuals take either 30 minutes of moderate aerobic exercise (walking briskly) on at least five days a week, or 25 minutes of vigorous aerobic exercise (running or jogging) on at least three days a week. The AHA also suggests a minimum of 10,000 steps per day.
Maximum heart rate is usually calculated as 220 minus age (in my case 166). And my target for cardio zone activity is 80% or 130. Fat burn zone activity would be a heart rate of more than 110.
Fitbit ‘active minutes’ are measured in METs (metabolic equivalents), where a MET of 1 indicates a body at rest and 3 is the threshold for being active. Fitbit splits activity into sedentary, ‘lightly active’, ‘fairly active’ and very ‘active’. It shows heart rate in three categories: ‘fat burn’, ‘cardio zone’ and ‘peak’. These minutes don’t make it through to the data that can be downloaded (at least in the software that comes as standard) but you can view it on the Fitbit dashboard.
Has ownership changed my behaviour? Well I think it has. I tend to walk up escalators on the London Underground more often than I did and I’ve lost count of the number of times I’ve remarked ‘That will be good for my step count.’ But I’ve stopped short of taking a walk round the block after dinner if my step count is way below my 10,000 target.
Interestingly, gut feel is also supported by my first three months’ worth of data. Minimum step count per day has increased, as have another couple of key metrics. On the basis of my experience, maybe people using one of these tracking devices take their fitness more seriously than those who don’t.
So it has been an interesting three months in many ways. I will continue to provide updates on this subject from a personal perspective. Interest in wearable devices continues and I have had a number of conversations about the previous article with people in the UK market. At least two reinsurers are providing Fitbits to a number of their staff as a way of exploring the potential of the data. Expect articles in the near future on some of the science behind wearable and the use of such devices in wellness programmes. Also, we’ll make some comments on the usefulness of sleep patterns as a predictor for early mortality.
Based on my three months, do I think that the outputs are usable for risk assessment or basic pricing? The answer is ‘probably’. Does it replace other information sources? Maybe not, but in conjunction with the other information it does help the underwriter build a picture of the applicant. Maybe this is stating the obvious, but I think that the more data you have the more interesting it becomes. Data gathered over a three-month period must give an underwriter a better picture than a snapshot of one day or one week.
But which of these pieces of data is likely to be of most use in an underwriting or pricing context? Step count, which is what most people talk about, is a pretty blunt instrument and, as I describe above, some folks will clock up additional ‘mileage’ in their own homes to inflate their steps count to meet daily targets. And this might not be particularly energetic. Calories burned and the various activity levels might have more potential through being more representative of activity and the intensity of that activity. However, I’m not sure that calories burned is that useful without also knowing calories taken in. So active minutes, and how many of those minutes were ‘highly active’, have the most potential for underwriters.
Using this data in conjunction with application answers, physical measurements and other information could be very powerful and lead to better risk categorisation. Combining it with four or five ‘killer’ questions and data from external sources could make an effective risk selection process. If the customer continues to provide data then one could think about moving towards a model of continuously variable risk pricing or further premium discounts.
But what about anti-selection and fraud? Will applicants fabricate their data or give the wearable to a friend who is training for a marathon? Possibly, but are the premium discounts likely to be worth that much? Are the people who might go to these lengths the sort of folk who are going to non-disclose or anti-select anyway?
A certain amount of protection can be obtained by obtaining upwards of three or four months’ data. If insurers are able to obtain the data direct from the technology provider, this might also add some protection against anti-selection. In fact it leads one to think about a tie-up between Fitbit and a risk carrier – constituting maybe a natural extension of the Fitbit brand. Target customers could be pre-selected via three or six months of data and offered a suitably discounted price or fewer health questions dependent on duration of usage and the results obtained. The Fitbit data would be historic and so more likely to be genuine. Overlay that with socio-economic data, credit score, prescription history or other such externally sourced information, and maybe there’s a credible selection process supporting a powerful proposition.
Is this part of what the future looks like?