Reflections and predictions from the AMIA 2016 Annual Symposium

December 14th, 2016 / By Senthil Nachimuthu, MD, PhD

The premier medical informatics event of the year, the AMIA 2016 Annual Symposium, had many excellent talks that caught my interest. The one that myself and the audience liked most, (judging by the standing-room-only audience of more than 250) was the “Beyond SMART – Remote Decision Support with CDS Hooks” systems demonstration talk with presenters from Cerner, Verily (formerly Google Life Sciences), Wolters Kluwer Health, Intermountain Healthcare, RxREVU and Epic. CDS Hooks uses the SMART on FHIR protocol to allow an Electronic Medical Record (EMR) system to call an external Clinical Decision Support (CDS) system, for example, when a patient’s record is opened. Some of the examples they used for the demostration werepediatric growth charts and neonatal hourly bilirubin nomograms, two very important indicators used in children and newborns respectively.

CDS Hooks also allows the external CDS system to query the patient’s chart using FHIR to get additional information. The CDS system can send its output to either the UI component in the EMR (“cards”), or the EMR backend to perform an action (“decisions”). These outputs can be seamlessly integrated into the EMR system to fit within the user interface and the workflow of the EMR, within the correct patient context. In addition to the effective use of various standards, having speakers from two major EMR systems demonstrating the use of a common standard to access external CDS systems is a significant moment.

There were several interesting talks on the applications of analytics as well. Some of the talks that stand out included the appilcations of IBM Watson technology in healthcare and the closing keynote by Eric Horvitz of Microsoft Research. Dr. Horvitz talked about turning decision support into actions. One of the papers he referenced that caught my interest was about predicting breast cancer diagnosis and other events based on search engine logs, by analyzing how the focus of the search queries change over time – from diagnosis to treatment to prognosis. There was a panel on calibration of predictive models that discussed the importance of and the methods for calibrating a proven decision support system while deploying it in new environments to ensure good accuracy despite the changes in the new setting.

There were also many talks on the use of Apple ResearchKit and the community-developed ResearchStack for Android to gather user-generated data, and to improve clinical outcomes. Some talks described ways to create cross-platform (iOS and Android) applications for mobile and wearable devices. It was impressive to see papers and posters presented by high school scholars, something that AMIA has been supporting for the last few years with the involvement of some outstanding mentors.

Finally, Dr. Patricia Brennan, director of the National Library of Medicine (NLM) talked about the future directions of the NLM in her keynote, stressing the importance of the following:.

  • The NLM will focus on gathering and hosting data, for example, from publications. This will be a monumental help to the repeatability and reproducibility of science.
  • The other is that NLM has issued an RFI seeking input from the public on its strategic plan. The NLM has bold and much needed plans to improve medicine, public health, biomedical and translational science through data science and open science. So, if you’re interested, please respond to NLM’s RFI by January 9, 2017.

Of course, meeting old friends and colleagues is another important part of the conference. It was great to meet current and former members of the University of Utah Biomedical Informatics department. Until I see many of them again next year, Go Utes!

Senthil K. Nachimuthu, MD, PhD,  Medical Informaticist with 3M Health Information Systems’ Healthcare Data Dictionary (HDD) team.