Introducing The Health Graph

image

A few of our users may have noticed the HealthUnlocked website looked a little different recently. The changes are far from just cosmetic; They run through every last line of code from the web page to the data store itself. We changed everything.

But why..?

I joined HealthUnlocked back in November last year with a mandate to build the technical team and scale the platform to allow HealthUnlocked to become the global leader in the eHealth space.

The incumbent team were really small but an exceptional bunch of highly dedicated and motivated geeks!

The technology… meh!

In terms of a small start-up just setting out with a few thousand users the platform was fine but in terms of multi-million, multi-lingual monthly visitors it wasn’t fit for purpose.

Regular inexplicable outages and slow page delivery plagued us at every turn. Moreover building new features and much requested changes was more painful than a red hot poker being shoved where the sun don’t shine. It quickly became very clear that if we were to scale the business to be the definitive web destination for health related content a few things would need to change.

We could have patched, reconfigured and wormed our way out of the mire we were in but we needed better foundations on which to build our house and I just wasn’t convinced that we would ever be able to do so more effectively than simply throwing it all out and starting over. We couldn’t fundamentally restructure the flow of gigabytes of unstructured free-form text and turn it into usable structured and semantically interconnected health data.

Say what?

The users of the HealthUnlocked platform are a pretty switched on bunch of people. They understand their conditions and through blog posts and answers to questions help one another to understand and come to terms with their conditions.

There really are gems of wisdom in there! We just needed a better way to find them and push them to the surface.

A better pickaxe

We have ambitious plans here at HealthUnlocked to make the patient the centre of the universe and empower them to be better informed, supported and more in control of their health and well being.

So we asked ourselves how could a Portuguese speaking long term sufferer of Wolff-Parkinson-White syndrome in Brazil help a English speaking newly diagnosed sufferer of the same condition in Mumbai? How could we leverage the wisdom of all our users in a language agnostic way to help each of them understand and better cope with the conditions they have?

The term “graph database” isn’t really a household name. Facebook have gone a long way and spent a fair bit of their investors cash to promote the concept but I think Kevin Bacon has probably done far more. We’re all connected by six or less acquaintance links or degrees of separation.

In the universe of health there are many different types of entity: Patients, Conditions, Symptoms, Signs, Body parts, Clinicians, Pharmacologists, Treatments, Hospitals, Insurers… The list goes on and on. All these entity types are connected in much the same way.

Patient A suffers from Condition B which is manifested by Symptom C and Symptom D. Patient A is attended by Clinician E who prescribes Treatment F which is manufactured by Pharmaceutical Company G and is subsidized by Insurance Company H.

I could take this further but I hope you get the point.

Through analysing the free form text entered by all our users we are able to catch keywords that are mentioned and match them to the conceptual language agnostic health term in our ontology.

One mans “Flu” is another mans “Gripe” but the concept is the same, the symptoms are the same and the treatments are (generally) the same. They are one and the same entity if we remove the language from it.

In addition to levelling the playing field of language across all these health terms, we can start building relationships between them and other terms that are mentioned in the same context. So ’headache’ or ’dolor de cabeza’ will often be linked in the same context as ‘influenza’ for example.

Over the course of millions of posts, comments, questions and answers we can start to build a pretty interesting map of how these things fit together. The relationship between a condition and its symptoms and the treatments normally prescribed for a condition.

We’ve only just begun down this path but were already seeing some very interesting results.

So how will this help me and why should I give a fig?

By having tightly linked terms and common relationships for everything health related, built up over millions of inputs, we can get content that is pertinent to you bubbling up to the top. We could introduce you to specialists. We can let you know of news that may be very relevant to you and with your permission and assistance we might help healthcare professionals work on better treatments and solutions to make your life a little more comfortable.

Just the beginning

We now have a gleaming kick-ass supercharged mega-engine – under the hood. It really is a thing of beauty. But it’s pretty worthless lump of ones and zeros unless we can start to leverage it’s power and let our users reap the benefits.

We know we’ve got our work cut out but the whole team is really excited to finally lift the wraps off our new platform and start the next phase of the journey working with the communities to build a platform that is unlike any other.

We feel very privileged to have a very vocal and committed community who we can rely on to let us know what they want and will work with us to achieve it. And for that we are one of the richest services on the interwebs.

– Adam Neilson, Chief Technology Officer


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.