Medical Devices and Clinical Informatics James E. Tcheng, MD, FACC, FSCAI Professor of Medicine Professor of Informatics Duke University Health System I have no relevant RWI to report

Disclosure Statement of Financial Interest I, James Tcheng, DO NOT have a financial interest/arrangement or affiliation with one or more organizations that could be perceived as a real or apparent conflict of interest in the context of the subject of this presentation.

If you don’t know where you are going, chances are you will end up somewhere else. -- Yogi Berra (1925-2015)

Medical Devices and Informatics Aka “Where Do We Need To Go?” First Principles Unique Device Identification and GUDID data Creating the “river of data”

The Clinical Informatics Model Clinical lexicon as the foundation of controlled vocabularies  rich clinical data + transactions A Common Data Model (and referential integrity) High quality data enables data exchange – “collect once, use many times” Documentation  team-based data capture at the point of care, via structured / semi-structured reporting Data privacy and security In theory there is no difference between theory and practice. In practice there is. Multiple attributions

Why a “Controlled Vocabulary”? Unified clinical lexicons applicable across all clinical documentation, from POC to registries to analytics Key role of professional societies

Referential Integrity … is a property of data which, when satisfied, requires every value of one attribute (column) of a relation (table) to exist as a value of another attribute (column) in a different relation (table) Wikipedia This underscores the concept of the single source(s) of truth – and is a requirement of data models  data (or analysis) aggregation

Implantable Devices: UDI adoption in EHI UDI = DI (Device Identifier) + PI (Production Identifiers) Where do implantable devices appear in EHI? Supply chain Clinical care: procedure / OR documentation, EHR Medical device interrogation, output Billing and claims data Registries How to capture in all EHI? What does the UDI enable?

Learning UDI Community - Data Quality Focus Clinically Relevant Size GUDID Repository for device identifier linked to device attributes – single source of truth for all EHI? Learning UDI Community - Data Quality Focus Clinically Relevant Size GUDID field Mfg 1 Mfg 2 Mfg3 SizeType Text Length null SizeUnit [blank] mm null SizeValue [blank] 18.0 null SizeText 18mm length [blank] null

Clinician Documentation 2017 Dictation recorder Pen

Clinician Documentation 2017 Mired in ancient paradigms Authoring of descriptive play-by-play novella encouraged (starts in med school) Demonstration of physician prowess, justification of actions (Misbelief) that it will be a good defense in case of malpractice 75% is garbage E&M coding requirements, EHR MU Team-based documentation actively discouraged By regulation, job description, EHR systems

What Has To Change? Document paradigm of the 20th century  informatics data model of the 21st century Transformation to structured and semi-structured reporting Understanding the destination: not just “Big Data”, it is “Deep Learning”

What is Structured Reporting? Team-based documentation: specific data captured by the person closest to that data in the clinical workflow (e.g. MA, tech, RN) Use of universal, well-defined common data elements Data model parallels clinical care model Data is compiled to produce vast majority of report MD: focuses on data quality, cognitive interpretation Output: the structured report ROI:  data quality /quantity,  redundancy / repetition, time to final reports, FTE requirements  augmented knowledge, financial gains

What is Deep Learning? … a branch of machine learning based on sets of algorithms that attempt to model high level abstractions in data You are following a patient with aortic stenosis. When should she undergo aortic valve replacement? When the risk of procedural morbidity and mortality will be the lowest, before the onset of end-organ consequences, and that will produce the best outcomes!

Where Are We Going? Informatics: common data elements (CDE)  controlled vocabulary; common data model (CDM); data interchange (CDA, IHE) -- from clinical use to CVIS to EHR to registry Clinical industrial (process) engineering to describe, model, implement best-practice workflows -- who does what when, where, and how New MD, staff professionalism standards -- conversion from dictation to information model -- implementation science, change management Partnership with IT vendors -- informatics and structured reporting