AMIA 2020 Informatics Summit Workshops

Monday, March 23
8:30 a.m. – 12:00 p.m.

A. Solomonides, NorthShore University HealthSystem; K. Fultz Hollis, Oregon Health & Science University; S. Rosenbloom, Vanderbilt University; L. Sheets, University of Missouri; J. Smith, American Medical Informatics Association

What is data governance? As we shall explore in this workshop, it comprises the regulatory principles, policies and strategies adopted, the functions and roles that must be created to implement these policies and strategies, and the consequent architectural designs that provide both a home for the data and, less obviously, an operational expression of policies in the form of controls and audits. The reasons for the extraordinary measures taken by institutions to protect the data lie in the value of that data as a strategic asset and in the internal and external threats to the data.

The workshop will include a presentation of background knowledge of principles (especially of recent developments), and an opportunity to role-play various data governance-related positions in an organization. Discussion of principles and of the simulated experience will complete the program.

F. Wang, Weill Cornell Medical College; X. Jiang, University of Texas Health Science Center

Numerous algorithms have been developed for different application scenarios in medicine, such as clinical decision support, drug discovery and development, insurance plan management, etc. These models include both the conventional interpretable models with high practical utilities, and the more advanced artificial intelligence (AI) models that have demonstrated state-of-the-art performance in many tasks. One concern for these models is whether they can make fair decisions for different individuals. There have been many studies demonstrating that both conventional clinical models and modern AI models can produce unfair decisions. Ensure model fairness and reduce health disparity is crucial for medicine. The goal of this workshop is to gather together clinical and algorithm researchers, clinical practitioners, health industry entrepreneurs, etc. to share their opinions on the fairness issues of existing models, the strategies of correcting bias for existing models, and how to ensure fairness in the model designation and building process.

J. Starren, N. Rothrock, Northwestern University; D. Meeker, University of Southern California; T. Nelson, Northwestern University

Patient-reported outcomes (PROs) represent health information directly reported by the patient receiving care. There are increasing efforts to incorporate PROs in clinical practice as a component of quality measurement, quality improvement, symptom management, and patient engagement. To enable this, organizations are attempting to integrate PROs into their electronic health record (EHR) environment. This integration presents numerous sociotechnical challenges including: new technology to support administration and scoring of computer adaptive tests, implementation of significant workflow changes for patients and clinicians, and management of the expectations of the many stakeholders impacted. This workshop will present challenges, strategies, and a newly-developed PRO implementation toolkit that can be employed to increase success. Among the tools is a clinical implementation planning process that leverages a multi-stakeholder survey instrument to identify critical workflow issues and organizational challenges. We will also review common pitfalls and solutions. The workshop will conclude with strategies for post-implementation evaluation and monitoring.

S. Rehman, H. Abbaszadegan, Phoenix VA Healthcare Systems/University of Arizona College of Medicine Phoenix

Great leaders are great negotiators, they resolve seemingly intractable disputes and yet enhance working relationships. Their negotiation and communication skills determine their effectiveness. Physicians and non-physician members of AMIA are expected to negotiate with a vast array of third parties, including healthcare system governing boards, leaders in the C-suites, patients, end-user consumers, government, health plans, insurance companies, EMR vendors, and pharmaceutical companies. Additionally, negotiation skills are an essential competency and requirement for board certification for physicians (ABPM and ABP), yet one may not find any session on this topic in AMIA meetings. It is time for all medical professionals be trained in negotiation skills.

Law, business, and public policy schools offer classes in negotiation. The ability to negotiate requires a collection of interpersonal and communication skills used together to bring about a desired result. It is based on exploring underlying interests and positions to bring parties together in a constructive way. Effective negotiators use innovative thinking to create lasting value and forge strong professional relationships. They take a deep dive in to what is behind the opponent and their own positions that may not seem logical at first but essential to understand the issues/ideas behind the problem.

The 3-hour highly interactive session provides tools for identifying individual communication preferences, delivery methods, conflict resolution styles as well identifying best practices and “best alternative to a negotiated agreement” (BATNA).

The interactive session involves exercises and activities that will allow the participants to discover, learn and practice the negotiation skills and tools. We will also be including data from the literature review, needs assessment, and Informatic tools that are useful in negotiation. We also have incorporated the outcome measures used in data science for success factors upon completion. The interventions/tools discussed/practiced in the workshop are evidence based and supporting data will be provided.

W. Zeng, B. Glicksberg, University of California, San Francisco; P. Newbury, E. Chekalin, B. Chen, Michigan State University

Rapidly decreasing costs of RNA sequencing have enabled large-scale profiling of cancer tumor samples with precisely defined clinical and molecular features (e.g., Low grade IDH1 mutant Glioma). Identifying drugs targeting a specific subset of cancer patients, particularly those that do not respond to conventional treatments, is critically important for translational research. From our previous work on liver cancer, Ewing’s Sarcoma, and Basal cell carcinoma, we have shown that the success of a systems-based approach is made possible by multiscale procedures, such as quality control of tumor samples, selection of appropriate reference tissues, evaluation of disease signatures, weighting cancer cell lines and selecting appropriate preclinical models. In this workshop, we will present OCTAD: an AI-enabled open platform to discover cancer therapeutics (http://octad.org). We will cover basic topics including RNA-Seq quality control, differential expression analysis, pathway enrichment analysis, and advanced topics including deep learning-based normal sample selection, genomics-based preclinical model evaluation, and a systems-based therapeutic prediction. We will present both the OCTAD web portal version that allows bench scientists and clinicians discovering potential therapeutics for the cancers of their interest, and the OCTAD R-based workflow that allows computational scientists to contribute to platform development.

Tuesday, March 24
1:30 p.m. – 5:00 p.m.

A. Bokov, M. Zozus, J. Gelfond, University of Texas Health Science Center at San Antonio

The R statistical language should be part of every clinical informatician's toolkit, particularly for preparing results for publications in academic journals. This workshop will show how and why to automate frequently occurring tasks such as reading in data files in various formats, building data dictionaries, and generating typeset submission-ready manuscripts. We will leverage the newly launched RStudio Cloud platform to provide pre-installed, pre-configured R sessions to participants along with scripts that they can customize to their needs and use for other projects.

Though R and some of the libraries we will present have been around for a while, we will emphasize their integration into an end-to-end reproducible analysis workflow with minimal effort spend on setup and "boilerplate code". The unifying theme is that automation and modularity are more than just a matter of convenience-- they are necessary for true reproducibility of one's analysis. The default use-case will be cancer registry data, but participants are welcome to bring their own publicly shareable datasets. It is recommended that participants sign up for GitHub, Zenodo, and RStudio Cloud accounts in advance of the workshop. Laptops with internet connectivity are required.

Wednesday, March 25
1:30 p.m. – 5:00 p.m.

T. Ong, University of Colorado Anschutz Medical Campus; S. Grannis, Regenstrief Institute, Inc.; L. Schilling, University of Colorado Anschutz Medical Campus; X. Jiang, University of Texas Health Science Center; R. Zucker, University of Colorado Anschutz Medical Campus

Patient health data are often scattered and incomplete. Record linkage (RL), or entity resolution in computer science, is a family of methods that identifies two or more records which refer to the same individual in one or more data sets. Accurate RL is essential for optimal care, and has important applications for clinical research. With recent advances in computer science, there is tremendous promise in leveraging the ever-growing repositories of secondary data. Challenges stemming from the use of single data sources can be overcome by linking data from multiple sources such as EHRs, disease registries, and administrative healthcare data. If a universal patient identifier, which various countries have implemented, is not available, patient personally identifiable information is required to link records. In research settings, sharing clear-text patient data is burdensome and presents significant security and privacy risks. New methods that use encrypted or hashed identifiers, called Privacy-Preserving Record Linkage (PPRL), minimize this risk but still require institutional approval and significant technological capacity. This workshop will provide AMIA members with in-depth knowledge related to all aspects of a RL process including data governance, data standardization, matching algorithms and the impact of linked data on data quality (DQ).

Thursday, March 26
8:30 a.m. – 12:00 p.m.

H. Xu, University of Texas Health Science Center at Houston; H. Liu, Mayo Clinic

Over the last few decades, growing adoption of Electronic Health Record (EHR) systems has made massive clinical data available electronically. However, over 80% of clinical data are unstructured (e.g., narrative clinical documents) and are not directly assessable for computerized clinical applications. Therefore, natural language processing (NLP) technologies, which can unlock information embedded in clinical narratives, have received great attentions in the medical domain. Many NLP methods and systems have been developed in the medical domain. However, it is still challenging for new users to decide which NLP methods or tools to pick for their specific applications. In fact, there is a lack of best practices for building successful NLP applications in the medical domain.

In this 3-hour workshop, we would like to introduce methods, tools, and best practices on building NLP solutions for clinical and translational research. We will start with an introduction of basic NLP concepts and available tools, and then focus on important applications of NLP in the medical domain such as phenotyping. We plan to use lectures, demonstrations and hands-on exercises to cover the basic knowledge/tools and use case studies to illustrate important trade-offs in the design and implementation of clinical NLP applications. Each instructor has over 10 years of experience in clinical NLP research and application and they will share their recommendations in building successful NLP applications in clinical research.

K. Wagholikar, J. Klann, Harvard Medical School, Massachusetts General Hospital; M. Mendis, Partners Healthcare; S. Murphy, Harvard Medical School, Massachusetts General Hospital

This workshop is a hands-on introduction to the ‘Informatics for Integrating Biology and the Bedside’ (i2b2) data platform. It will provide an overview of the platform functionality and discuss approaches to install the platform and to import data. Level of content is 70% novice and 30% intermediate. The workshop suited for researchers, clinicians, IT programmers, educators, leaders in healthcare participating in projects in health information technology. Intermediate and advanced users of i2b2 can benefit from a recap of the fundamentals and learn about new tools and approaches using i2b2. i2b2 has been deployed at over 200 institutions across the world to enable researchers to identify and analyze patient cohorts for clinical studies. The aim of the workshop is to introduce i2b2 to novice users with hand-on training, to educate the research community about the i2b2 data tooling, and to evolve good practice guides for i2b2.