Three years ago the Qualcomm Tricorder XPrize was announced, a competition to produce a portable Star Trek tricorder. The final twelve entries have been chosen to build prototypes and it is progressing towards some fantastic solutions. Science fiction, I think, has been inspirational in what can be achieved. The best entrant is difficult to pick, but all are attempting to non-invasively measure a patient’s vital statistics plus a number of conditions, including whooping cough, hypertension, mononucleosis, shingles, melanoma, HIV, and osteoporosis. These are at the moment very niche devices that will likely take a number of years to get regulatory permission, so will be some time away from common use.
All the big players seem to be getting into wearable devices – Google with the ill-fated Glass, Garmin’s fitness watches, Samsung’s Gear S. Most of these however are aimed at being a fashion accessory or a minor fitness device, measuring heart rate, distance travelled and blood pressure.
But, the next generation of devices are on the market now, and are able to monitor glucose, ECG, pulse, blood oxygen, and blood pressure. They are being incorporated into clothing, so are less intrusive into daily life. Many articles speak of data being transferred seamlessly from the user to the doctor to make medical decisions. This is a fine theory, but in practice with medical budgets being constrained and clinicians’ time stretched, there needs to be a different way of thinking about this.
Let us consider the volume of data that is likely to be collected. For one person, who may need high frequency data analysis performed, you might measure data every second giving 86400 seconds each day. If we measure the following:- GPS location, heart rate, blood pressure, blood glucose level, oxygen level, ECG, we could be looking at perhaps 400 bytes of data every second. This is about 33Mb of data per day. So if a health organisation is monitoring 10,000 patients at this level, over a year that would be 114 Tb. For a data analysis solution this enters the realm of big data.
To ease clinical load, I would suggest that, instead of the rather simplistic model of trained staff interpreting this mass of data, the data be passed to a ‘clinical analysis’ engine, which would flag up two types of intervention. The first is time-critical cases where an intervention is necessary now and the second being less time-critical cases where intervention can be suggested on the basis of the severity or importance of the trend.
I would suggest that we need to rethink our data methodologies and include an analytics engine which would be able to ascertain the time criticality / importance / likelihood to then be able to engage the appropriate end user, being emergency service, doctor, health visitor or customer. These would obviously need to be tested, validated and monitored; ensuring that ‘false positives’ are minimised (these waste emergency services and clinicians’ time). The larger danger is that the system misses individuals who need intervention, but are missed due to lack of algorithmic correctness. I shall speak of these in another blog.
The important aspect of this solution is the ‘clinical analysis’ engine, which would likely utilise a mix of NoSQL and SQL technologies and would allow the very large set of data to be parsed and ‘understood’ within its context. Its real power however is to make comparisons between individuals so that known conditions can be monitored and time-appropriate action can be suggested.
This obviously needs to be a system that has significant medical understanding built into it, and trained health professionals will need to work as part of the overall system.
There are various predictions on the growth of wearable devices, one suggesting that in 2017, there will be 515 million devices shipped. I believe this to be driven by the last of the baby boomers retiring. They are a tech-savvy, wealthy generation that is the longest lived we have seen. Maintaining health and lifestyle is a key part of their focus. But this depends on a number of hurdles, high on the list are data privacy and security.
Data science and data analytics really haven’t comprehended the scope of the problem, or as I like to think of it, the scope of the opportunity!
We are seeking to engage with forward thinking health monitoring companies who can see the opportunity as well!