Duke's Ricky Bloomfield: We're all in with APIs [Q&A]

When it comes to the use of application programming interfaces in healthcare, Ricky Bloomfield, director of mobile technology strategy and assistant professor of internal medicine and pediatrics at Duke University, says he and his colleagues are all in.

"This is where things are going," Bloomfield, who is scheduled to speak at next week's mHealth Summit, told FierceMobileHealthcare. "The app store economy has shown that when you give these tools to developers in a standard, easy-to-use way, you get things you wouldn't expect. It really moves the needle forward in terms of the quality and cost of patient care, which is what we're all about."

In this exclusive interview, Bloomfield expands on his team's efforts surrounding APIs, and also discusses his latest efforts around examining autism through the use of Apple's ResearchKit.

FierceHealthIT: Talk about your efforts to develop a scalable infrastructure to support your ResearchKit efforts

Bloomfield: It was important for us to be able to have control of the data for security and privacy reasons. The study we launched last month called Autism & Beyond uses the camera on the phone and analyzes a child's expression and emotion as they watch certain videos that have been designed to elicit that emotion. We record those videos and consider them very sensitive. We wanted to safeguard that here rather than going to a cloud provider or something else.

We designed a back end that would allow us to upload any type of arbitrary data. It handles the authentication, the account creation and management, and so far, it's worked very well.

FHIT: Any updates on the Autism & Beyond study?

Bloomfield: The app launched on Oct. 15, and we consider it successful in enrolling the target patients for the population that we want. Our plans are to scale it out to China and Africa so that we can start working with additional populations that also have unmet needs with respect to autism screening and diagnosis.

We have over 300 users, so far, and to put it into perspective, there was a study that was ongoing prior to this one in our clinic for which we enrolled 90 users in eight months; we enrolled twice that number in the first weekend. To our knowledge, this is the fastest any type of childhood mental health study has ever enrolled.

FHIT: Have you learned anything this early on?

We're getting lots of great data; we're excited to dig into the data and get some meaningful results out of it, which is the next phase as the data collection continues. Collecting data that will help us one day to develop a tool that will allow us to screen; the intent of this study is to collect this video data and correlate that with the videos so that we can determine the feasibility of collecting this data at home. That's the No. 1 goal of this study: determining if that's possible and reliable.

If we can determine that, a secondary endpoint for the study would be to take that and try to figure out if any of the children enrolled in the study do have a diagnosis of autism, because we asked that; we ask about various diagnoses. And if they have that vs. those who don't and using the algorithms that have been developed here, we'll be able to look at the sensitivity and the specificity of our algorithm in detecting autism.

The bar is pretty low for developing a better screening tool than what's already out there. The current standard for initial screening is the M-CHAT (Modified Checklist for Autism in Toddlers), which is not very reliable. If you have autism, it will generally pick you up, but it also flags a lot of people as positive when they don't actually have autism; when they have something else. We call that a specific test, but not a sensitive one. The sensitivity's actually as low as 50 percent, which for a screening test is pretty bad. But that's all we have right now.

FHIT: When the proposed rule for Stage 3 of Meaningful Use was released last spring, you came out in support of the emphasis on APIs and patient-generated data. Talk about the potential for the use of APIs in electronic health records and how your efforts in doing so have matured.

Bloomfield: We've been a big proponent of that. Our efforts started a year and a half ago, in developing a system of APIs here at Duke. We actually had developed something that was similar to the SMART and FHIR platforms before we really knew what that was. We had an Android app connected to our Epic system, pulling data out; then, when we became more familiar with FHIR and we saw that the standardization efforts around that, it was very clear to us that that was the path we needed to take.

In January of this year, we had our initial FHIR implementation working against our Epic platform; we did that development ourselves. And at HIMSS in April, we showed three different applications working in our proof of concept environment. One was on pediatric growth charts. One was an app called Meducation and the other was an app we developed called the Duke PillBox, which is a patient education tool that can be used at discharge or at home to help patients better understand their medication regiment.

We then took that proof of concept, and further refined out. As of August of this year, we have this infrastructure live in our production environment, and last month, we went live with the Growth Chart.

We're all in with respect to these APIs. We think that lowering the barrier for EHR data access is a good thing for health systems and for innovation, and it's a great thing for patients to be able to access their own data and to be able to choose how they want to be able to interact with the system.

It gives our innovators here at Duke and around the world a path to develop their widget or app or website or whatever it is in a standard way and to get data from the EHR, which is what you need if you have meaningful decision support and encourage the type of innovation.

We're very excited about it. I'm glad it's being emphasized in MU Stage 3; I think some people are concerned that it may be difficult to meet those benchmarks, even though they were lowered from the draft to the final rule to try to make it more achievable. But we have to make it more achievable and this is a great way to get moving.

Editor's Note: This interview has been condensed for clarity and content.