Predictive analytics
for prevention
What is 'Predictive Analytics'?
Predictive analytics uses historical data trends to indicate that something may happen in the future. In terms of wellbeing, this can indicate a possible health problem. Once a potential problem has been detected it may be possible to investigate and, if needed, for early intervention to stop it from getting worse.
For instance, a disturbed night’s sleep could be an early warning of a Urinary Tract Infection (UTI) and be a prompt for early testing and treatment. Unaddressed UTIs not only lead to many hospitalisations in their own right, sometimes leading to sepsis and death, but also falls in the home, often with broken bones and hospitalisation.
The Acticheck approach
Our approach of creating a ‘wearable tech for people who don’t do wearables’ means the life-saving wristband is worn all the time and gives us persistent movement readings every 15 minutes. These readings enable us to quickly build a picture of the wearer’s normal activity levels throughout the day. We have found that the general pattern of activity levels during the day are reasonably consistent and gives us a baseline to compare against the last 24 hours of activity.
The typical activity levels can be seen by looking at MY BAND on the dashboard. The yellow line – we call it ‘the golden line’ – is the baseline activity. The dark blue line shows the readings we receive every 15 minutes when the wristband is in contact with our server.
It can be easier to see the relationship between the two it you look at the ‘This week’ view.
If our analysis finds a discrepancy between the recent activity and the golden line we will report this. This is done both by showing it in the recent event log (this is updated at 10:22 each day) and by an email sent at 10:30 to every responder who receives ‘update emails’.
You can manage who receives update emails by clicking on Managing update emails and following the instructions.
What we tell you about
Currently, there are two email notifications you might get from us:
- Disturbed night’s sleep
When we have seen unusual activity during the normal wearer’s sleep period, maybe caused by the wearer getting up many times overnight - Inactive day
When we have not seen the usual level of activity during the day, maybe because the wearer is still in bed.
Each email notification will come with a link to a graph like this:
On the graph it is apparent that the level of activity between midnight and 03:00 was far higher than normal, hence the notification.
In addition to the activity lines we also show you the temperature being reported by the wristband. You may notice that this is inconsistent as it is not body temperature, so someone rolling their sleeves up in a cool environment is likely to show a consequential drop in the band temperature reading.
We should stress, that receiving one of these notifications doesn’t mean there is necessarily anything wrong. Some people do not have a typical daily routine or may go to bed late or get up early for entirely normal reasons, such as watching an international sports event.
However, these notifications allow for preemptive/proactive intervention if needed.
Future developments
We are working on refining the analytics with machine learning to bring more value into what we achieve with the data.
We hope to include:
- Cohort Analytics
It will enable us to eliminate behaviour that has been exhibited by a significant proportion of people in an area, for instance, a disturbed night’s sleep when there has been an extreme thunderstorm. - Identification of long-term decline
Comparing baseline changes over the duration of the wearer’s use to identify slow declines that might otherwise go unnoticed. This could be used to trigger timely assessments to offer support to help maintain independence for as long as possible. - Recovery planning
Gain a pattern for recovery from a variety of ailments and operations so that we can monitor to ensure that improvements are happening at the expected rate.
