Thursday, October 15, 2015

ResearchKit is Official at Duke: Autism & Beyond

Over the past 6 months there's been something brewing at Duke ... something that we're now incredibly excited to share with the world.

But first, a bit of background:

  • 1 in 68 children will be diagnosed with Autism
  • Autism can be diagnosed as young as 18 months old
  • The average age of Autism diagnosis in the US is over 5 years old
  • A child's brain will grow at a rate of 700 synapses/sec in the first years of life
  • 70 counties in NC have no access to a childhood mental health specialist
  • In Africa, where there are approximately 500 million children, there are only about 50 childhood mental health specialists

Do you detect a need?

We do, too.

Thus, Autism & Beyond was born. Autism & Beyond is an incredible new ResearchKit study that builds upon the groundbreaking work of Geri Dawson, Helen Egger and Guillermo Sapiro, who over the past several years have been working to refine novel video algorithms that can analyze and detect a child's emotion in real-time. They've been conducting studies at Duke clinics using an iPad prototype app for almost 2 years.

In early 2014 I had the pleasure of working with Kathleen Campbell, a wonderful 2nd-year medical student on her inpatient Pediatrics clerkship. I was her attending physician at the time. Shortly thereafter, and knowing I had an interest in mobile technology, she gave me a demo of an app that the aforementioned team had put together. The idea and technology were amazing. I thought it was great work, although I really had nothing to add at the time. I looked forward to seeing the results of that research.

Fast forward to March 2015, and the announcement of ResearchKit. It was clear that we needed to put this technology to the test, and quickly. We cast a wide net looking for "shovel-ready" projects, and of course, the autism app Kathleen showed me bubbled to the top. The project was already underway (with an iOS app no less!), and had a great team that had already made significant strides in this area. It was an natural fit.

It was also a remarkable coincidence that at this exact time within the Duke Institute for Health Innovation (DIHI) we had just hired two talented mobile developers, Mike Revoir and Jamie Daniel, who were (are) passionate about mobile technology, health, and research. They were so excited about the project that they were eager to dive in even before their first official day of work! As the scope of the project grew, so did the number of individuals and teams involved. In the end, it was a peerless example of cross-institutional collaboration across Duke University and the Health System. This project couldn't have happened without any of them.

Autism & Beyond

So what about the actual app? I could explain it in detail here, but seeing is believing, so go download it now! And if you want more info, check out the website. Even if you don't meet the eligibility criteria, we've made it simple to get a taste of the cool technology that's gone into it without ever signing up.


The app basically comes down to our need to know one thing: could we one day use a mobile phone to automate screening for conditions such as autism or anxiety? To start, we first need to know if it's even feasible to analyze facial expressions on such a small device. That's what this study is intended to determine: feasibility.

As you can see from the screenshot above, the software algorithms developed by Guillermo and his team not only detect facial features, but also expressions, and can do so in real time. The data from this study will help to refine those algorithms.

In addition to the facial recognition pieces, we've also included several critical questionnaires with the help of Helen and Geri. For example, we ask about temper tantrums and provide feedback to users regarding where their child falls compared to his/her peers.

Additionally, study participants will be able to see how many other families have enrolled as well as a few additional aggregate data points:

Check out the number of enrolled
participants in near real-time!

While the release of the app marks the end of one chapter (and a whole lotta work by our incredible team!), it's clearly just the beginning. We hope that the app will have an impact in the US, but also plan to roll it out in China and South Africa soon. The more children we can reach, the more we can help. While ResearchKit allows us to reach millions, we still need to take care of one child at a time, and ensure that those children have access to the resources they'll ultimately need for full diagnosis and treatment. That's going to take even more teamwork!

Tuesday, October 13, 2015

Duke's on FHIR (for real this time)!

In June I described our work to date on integrating the SMART and FHIR APIs into our Epic-based EHR.

As a recap, we started this process in 2014 with a custom Android app that pulled patient problems, medications and demographics directly from the EHR via simple REST-based APIs. As we became more familiar with SMART and FHIR, we realized the value in joining forces with a common standard. It was never our intention to create something that would only be useful to us, and it was clear that the momentum around FHIR was building.

Fast-forward to January 2015, when we had our first SMART apps running in our proof-of-concept environment. It could be done! We followed this with integration of several additional apps and a demo at HIMSS in Chicago. You've seen this before:

This was all fine and dandy, but it was still all in our proof-of-concept system with "fake" patient data.

Since that time our amazing development team, led by Felipe Polo-Wood, has been diligently working to move this infrastructure into our production environment, an important milestone in order to show that SMART and FHIR can do more than just play in the sandbox.

I'm happy to report that as of August 26, 2015, the infrastructure has been live for some Duke-specific internal use-cases. In fact, Felipe was so excited to share that I had a screenshot waiting in my inbox that morning, demonstrating that the systems were, indeed, calling the FHIR APIs, and that the transition was seamless from the outdated infrastructure FHIR was intended to replace:

Duke was officially on FHIR!

But not ones to rest on their laurels, the team immediately got to work to enable a more visible example of what FHIR can do: a true SMART-compatible app, Pediatric Growth Chart.

Drumroll ...

On October 9, 2015 I successfully logged into our production system for the first time to view real patient data in a FHIR app! I'd love to share screenshots with you, but they contain real patient data, so I can't! Let me say that again: real patient data, via FHIR, within Maestro Care, our Epic-based EHR.

And, of course, the best is yet to come!

Kudos to Felipe and the rest of the incredible CATS development team for making this happen. It has truly been a team effort, and I know they share the same enthusiasm for this project as I do, because they're always smiling!

The incredible CATS team:
  • Felipe Polo-Wood (SeƱor Manager)
  • Vince Guaglione
  • Lusia Li
  • Luiz Omori
  • Carrie Porterfield