Session 5
Chair: James (Jim) Cimino, MD, FACMI, FACP, FNYAM, FAMIA, FIAHSI (University of Alabama at Birmingham)
Speaker: Kenrick D. Cato, PhD, RN, CPHIMS, FAAN (University of Pennsylvania)
Professor of Informatics
Title: Harnessing Clinician Expertise Through EHR Patterns: The Future of Predictive Healthcare
Abstract: In the inpatient setting, clinicians routinely use the electronic health record (EHR) to do their work. This makes the EHR a source of rich patient and clinician data. In this presentation, Dr. Cato will describe the Healthcare Process Modeling Framework to Phenotype Clinician Behaviors for Exploiting the Signal Gain of Clinical Expertise (HPM-ExpertSignals), which leverages clinician interaction with the EHR to identify meaningful user patterns. The presentation will highlight the application of the HPM-ExpertSignals approach in the COmmunicating Narrative Concerns Entered by RNs (CONCERN)) study – a project that focuses on predicting inpatient deterioration.
Speaker: Sarah Collins Rossetti, RN, PhD, FACMI, FAAN, FIAHSI (Columbia University)
Associate Professor, Biomedical Informatics and Nursing
Title: Improving Data Capture while Reducing Documentation Burden
Abstract: Valuable EHR data are diluted by noisy data lacking variability among patients and resulting in information overload for clinicians. A large volume of EHR data are still entered manually into structured and semi-structured fields in the EHR, limiting clinicians’ time to engage in: personalized care planning, higher level clinical decision making, capture and synthesis of the patient’s story, and communication of key patient information to the interprofessional care team. Nurses suffer from exceptionally high volumes of required manual data capture yet there is little evidence that these data capture requirements benefit patient outcomes. Our recent scoping review identified a paucity of studies linking required data capture by nurses to patient outcomes. Of the 47 studies identified most had significant methodological weaknesses and over half (n=27) focused only on documentation compliance measures not patient outcomes. However, strong evidence from the multi-site randomized controlled CONCERN trial supports linkages between nurse-driven decisions to capture data and positive patient outcomes. Recent interviews and focus groups with nurses identified detailed examples of how required manual data capture in EHRs are not aligned with practice needs and drive erroneous data capture that do not reflect the actual care delivered or patient conditions – findings with critical implications for AI and the data science community. This presentation will discuss: the signal to noise problem in our EHRs including how nurse-driven decisions to capture data (as opposed to requirements to capture data) carry statistically significant signals that can be leveraged to understand and improve care; the negative consequences of required manual data capture; and emerging AI solutions to automate data capture with considerations for evaluating their impact on data quality and time available for higher level clinical decision making.
Speaker: George Hripcsak, MD, MS (Columbia University)
Professor, Biomedical Informatics
Title: Data Reuse for Research and Learning
Abstract: For many decades, informatics researchers have been exploiting data obtained in the course of health care to carry out clinical and informatics research, to learn about health care processes and physiology, and to teach trainees. As databases grow, algorithms improve, and computers grow in capacity, the potential benefits increase. Yet it is all based upon a shared vocabulary for everything we do in health care. Decades of research and hard work on ontologies and standards have led to a strong foundation for data reuse, and modern machine learning may strengthen and broaden that foundation. Putting it all together could result in leaps forward in reliable medical knowledge.
Speaker: Kensaku Kawamoto, MD, PhD, MHS, FACMI, FAMIA (University of Utah)
Associate Chief Medical Information Officer
Professor & Vice Chair, Clinical Informatics, Department of Biomedical Informatics
Title: Re-imagining the EHR with Interoperable Digital Health Innovations
Abstract: Federally mandated data interoperability standards can now be used to implement scalable digital tools that seamlessly interface with electronic health record (EHR) platforms such as Epic and Cerner. In this session, Dr. Kawamoto will describe how the University of Utah’s ReImagine EHR initiative is leveraging these standards to incorporate cutting-edge digital health innovations into clinical care, and how you can take advantage of these standards-based approaches in your own research and clinical endeavors to enhance health and health care.
Speaker: Charles Jaffe, MD, PhD, FACP, FACMI (Health Level 7 International)
Chief Executive Officer
Title: Interoperability after HL7 FHIR
Abstract: When Ed Hammond helped to craft the electronic health record for Duke Hospital, data exchange was not a critical concern. However, when Ed led the effort to create HL7 nearly four decades ago, semantic interoperability had become an essential cornerstone of the care process. In fact, nearly all the world’s healthcare data was soon to be exchanged with some construct of HL7 V2.
In 2012, the notion of true plug and play interoperability became a foreseeable goal with the introduction of HL7’s Fast Healthcare Interoperability Resources. Collaboration with many partners help to drive the needle closer to the vision of sharing accurate, reliable health data when and where it was needed. The private sector added a full-throated chorus of approval. It was not until the federal regulatory agencies in the US and beyond made FHIR a mandatory feature of interoperability did it become firmly added to our informatics lexicon.
What emerges after the capacity of the open API (Application Programming Interface) is unclear. Certainly, advocates of artificial intelligence will have their say. Quantum computing offers an cautionary opportunity. The very next steps are the most intriguing ones.
