AHRQ CEPI Evidence Discovery and Retrieval (CEDAR)

Peter Krautscheid, BS¹; Marc Hadley, PhD¹; Mario Teran, MD²;

and Edwin Lomotan, MD²

​

¹The MITRE Corporation, McLean, VA;

²Agency for Healthcare Research and Quality, Rockville, MD

​

​

​

MCBK Annual Meeting – July 20, 2021

​

​

• CEDAR is a standards-based application programming interface (API) that supports search, access, and use of patient-centered outcomes research findings across multiple repositories in AHRQ’s Center for Evidence and Practice Improvement (CEPI)
• Where possible, CEDAR leverages Fast Healthcare Interoperability Resource (FHIR) resources
• CEDAR also aligns the AHRQ repositories with the FAIR (Findable, Accessible, Interoperable, and Reusable) Data principles
• CEDAR facilitates the exchange of computable biomedical knowledge and enables access to and use of PCOR information at the right time and right place

Approved for Public Release; Distribution Unlimited. Public Release Case Number 21-2022. This technical data was produced for the U.S. Government under Contract Number 75FCMC18D0047, Task Order 75Q80120F80008, and is subject to Federal Acquisition Regulation Clause 52.227-17, Rights in Data-Special Works. No other use other than that granted to the U.S. Government, or to those acting on behalf of the U.S. Government under that Clause is authorized without the express written permission of The MITRE Corporation. For further information, please contact The MITRE Corporation, Contracts Management Office, 7515 Colshire Drive, McLean, VA 22102-7539, (703) 983-3000. © 2021 The MITRE Corporation.

Apervita Vital™ Platform, and Apervita Knowledge Studio

Blackford Middleton, MD, MPH, MSc

Chief Informatics and Innovation Officer

Kenji Wong

Product Manager, Apervita Knowledge Studio

Apervita, Inc.

​

MCBK Annual Meeting – July 20, 2021

A demonstration of the Apervita Vital™ Platform, and Apervita Knowledge Studio, will be provided. The Vital platform is an open knowledge management, execution, and delivery platform. It connects knowledge producers with knowledge consumers. It allows knowledge authors (professional societies, measure stewards, content publishers) to develop computable biomedical knowledge on a secure cloud platform, and provides delivery of knowledge-based apps and services to disparate EHRs, and other endpoints.

• Overview of the Vital Platform, and demonstration of authoring a quality measure, and a clinical pathway, in the Apervita Knowledge Studio.
• Demonstration of provisioning of CBK to a recipient site (an app).
• Two examples of workflow integration into disparate EHRs.

A FHIR Framework to Ignite Biomedical Knowledge Management

Muhammad Afzal, PhD¹; Brian S. Alper, MD, MSPH²; Joanne Dehnbostel, MS, MPH², Andrey Soares, PhD³; for the COVID-19 Knowledge Accelerator (COKA) Initiative

¹Sejong University, South Korea, ²Computable Publishing LLC, USA, ³University of Colorado, USA

​

Link to poster PDF

​

MCBK Annual Meeting – July 20, 2021

• Biomedical Knowledge is diverse and comes in many forms and formats
• Scientific communication relies on large sharable value units (e.g., documents)
• New models to manage smaller sharable value units are needed (e.g., effect estimates)
• The smaller sharable value units are called Resources in the HL7 FHIR model.
• This poster presents a FHIR framework for managing biomedical knowledge using eight accepted knowledge management system processes -- discovery, modification, translation, dissemination, creation, representation, storage, and retrieval.

A Heterogeneous Knowledge Environment �for Cognitive Support Applications

Davide Sottara (Sottara.Davide@mayo.edu); Adam M. Bartscher, �Matthew D. Hegland, Branden C. Hickey, Marc L. Sainvil, Jane L. Shellum

�Mayo Clinic, Rochester, MN

​

​

MCBK Annual Meeting – July 20, 2021

• Clinical Cognition Support Applications require deeper knowledge of the clinicians’ mental models to achieve Situation Awareness
• The underlying Knowledge Bases include clinical and technical Computable (Biomedical) Knowledge Assets
• Different (standard) Representation Languages exist for different Knowledge Assets
• A real world clinical application uses ~150 Assets of ~20 types in ~15 Languages
• We manage diversity using a Knowledge Graph with standard semantics and APIs

​

Management of Pregnancy of Unknown Location and Tubal and Nontubal Ectopic Pregnancies

MCBK Annual Meeting – July 20, 2021

A Hybrid Clinical Reasoning Approach that includes Abduction

Sabbir M. Rashid¹ (rashis2@rpi.edu), , Jamie P. McCusker¹, Oshani Seneviratne¹, Daniel M. Gruen¹, Amar K. Das², & Deborah L. McGuinness¹

¹Rensselaer Polytechnic Institute, Troy, NY ²IBM, Cambridge, MA

We design & implement a Clinical Decision Support System

• Hybrid reasoning can help emulate how physicians perform reasoning
• Differential diagnosis, therapy planning, and plan critiquing are applicable reasoning strategies
• We require Description Logics (DL) compatibility and reasoning in polynomial time

Clinical Reasoning

• Abstraction – Identification of problem features
• Abduction – Choice of hypotheses to explain a fact
• Deduction – Inference of conclusion from statement
• Induction – Use of the known to predict the unknown

This work is supported by IBM Research AI through the AI Horizons Network.

We abductively find explanations for questions like “What could have caused

the current finding 

to occur?”

​

Approach Based on  the Select  and  Test  Model  (ST-Model), as shown in Fig. 1

• Epidemiological framework for medical reasoning
• Built using SPARQL CONSTRUCT queries within Whyis
• Patients are represented using FHIR , as shown in Fig. 2

Fig. 2 – FHIR is used to represent patient visits, observations, and adverse events

Fig. 1 – ST-Model for a Differential Diagnosis scenario involving unexplained weight gain

A public repository to mobilize computable biomedical knowledge artifacts -- https://cbk.pub

Güneş Koru (gk+mcbk@drkoru.us); Sai Yerram; Abir Rahman

University of Maryland, Baltimore County

​

Poster link: https://drkoru.us/posters/mcbk2021

​

MCBK Annual Meeting – July 20, 2021

• The MCBK Standards workgroup recently published metadata categories for CBK artifacts
• To support further development and refinement of the categories, we developed a CBK repository allows creating and maintaining public CBKs with metadata fields in thirteen categories
• The API libraries in 18 different programming languages allow automation
• Discussion threads in the repository provide users with a forum to discuss individual artifacts.
• Surveys and interviews to assess perceived usefulness and complexity of the metadata approach
• The repository is useful for 1) quality improvement & research 2) Collecting usage-based feedback about the metadata approach developed by the standards workgroup for further improvements.

A Research-Protocol Object to Generate Biomedical Knowledge That is Auditable and Reproducible

Harold P Lehmann¹ Amin A Manna² Kenneth J. Wilkins³ Katie R. Bradwell² Benjamin Amor² Johanna J Loomba⁴ Eli B. Levitt⁵ Andrew Williams⁶ for the Applicable Data Methods and Standards Workgroup, National COVID Cohort Collaborative (N3C)

¹Johns Hopkins ²Palantir Technologies ³NIDDK ⁴University of Virginia ⁵Florida International University ⁶Tufts

Link to poster PDF

​

​

​

​

MCBK Annual Meeting – July 20, 2021

Insert Photo Here

• Scientific work remains based on recurring appeal to free text documents
• A computable protocol should define PICOT elements once and use many
• “Use” includes defining inclusion/exclusion, computable phenotype, analysis, and evidence reporting
• “Use” results in self documentation and completion of publication checklist
• “Use” includes analytic decision support

Beyond Safe Harbor: Risk of Exposing Location in De-Identified Clinical Data

Alfred Jerrod Anzalone¹, Carol R. Geary², and James C. McClay³

​

^{1. Department of Neurological Sciences, University of Nebraska Medical Center (UNMC), Omaha, NE; 2. Department of Pathology & Microbiology, UNMC, Omaha, NE; 3. Department of Emergency Medicine, UNMC, Omaha, NE}

Link to Poster

MCBK Annual Meeting – July 20, 2021

• The use of de-identified EHR data for clinical and translational research has increased significantly since the HIPAA Privacy Rule De-Identification standards went into effect
• Inclusion of SDOH measures in de-identified research is increasing as well, which presents inherent risk of re-identifying PHI (primarily location units smaller than the state)
• Data warehouse architecture and institutional policies need to recognize the risk associated with providing multiple location-based indices
• Research interests are secondary to privacy concerns throughout biomedical research, but particularly in de-identified research, which is intended to promote more secure access to EHR data while allowing for expedient access (fewer institutional barriers to entry)

Biomed News - A biomedical literature expertise sharing system based on machine learning and expert curation

​

Gavin McStay - Staffordshire University, Stoke-on-Trent, UK - gavin.mcstay@staffs.ac.uk

​

Thomas Krichel - Open Library Society, New York, USA - krichel@openlib.org

​

http://biomed.news @biomednew Link to poster

​

​

​

MCBK Annual Meeting – July 20, 2021

• Approximately 30,000 abstracts released each week via PubMed.
• This enormous release of information requires robust processing so relevant information can reach the appropriate community.
• Biomed News uses machine learning to sort abstracts for individual reports.
• These ranked abstracts are sent to topic experts each week who then select relevant articles for machine teaching.
• Biomed News experts are all over the world and we are looking to grow and include other sources of biomedical literature, such as preprints.

Centre for Addiction and Mental Health (CAMH) BrainHealth Databank Knowledge Graph

​

Rotenberg, D., Ansari, A., Yu, J., Bogetic, N., Hill, S.

Krembil Centre for Neuroinformatics, CAMH, Toronto, Ontario, Canada

​

​

MCBK Annual Meeting – July 20, 2021

• Knowledge graph used to integrate multiscale clinical and research data from disparate sources to drive AI, ML, and clinical decision support dashboards
• Demonstration of data modeling - FHIR/HL7, ingestion - Python, and querying - SPARQL and Elasticsearch
• Benefits of using linked data for research and clinical care and adherence to FAIR principles
• Developing knowledge library to continuously grow and enhance data in the graph

MCBK Annual Meeting – July 20, 2021

MCBK Annual Meeting – July 20, 2021

Computable Case Reporting for �Multicenter Clinical Trials and Registries

​

MCBK Annual Meeting – July 20, 2021

• Case report forms (CRF) for trials/registries are usually completed manually.

Andrew J King, PhD

@AndrewsJourney

andrew.king@pitt.edu

King AJ, Malakouti S, Music E, Kalchthaler K, Holton J, �Quinn K, Clermont G, Marroquin O, Angus DC, Horvat C

​

University of Pittsburgh & UPMC, Pittsburgh, PA

​

Link to poster PDF (https://bit.ly/36hg2xZ)

• Automatic CRF from electronic records would improve efficiency, accuracy, reproducibility, and auditability.

• We present recommendations to enable automatic CRF.

Decentralized and Reproducible Geocoding and Characterization of Community and Environmental Exposures for Multi-Site Studies

Erika Rasnick¹, Andrew Vancil¹, and Cole Brokamp^1,2

¹Division of Biostatistics and Epidemiology, Cincinnati Children’s Hospital Medical Center

²College of Medicine, University of Cincinnati

​

View the poster here.

MCBK Annual Meeting – July 20, 2021

• Many studies aim to analyze relationships between place-based characteristics and health outcomes, but maintaining data privacy and reproducibility often cause difficulty, especially for multi-site studies.
• DeGAUSS is a decentralized approach for geocoding and deriving community- and individual-level characteristics, such as census data, measures of community deprivation, airborne pollutants, greenspace, access to healthcare, and more.
• By using a container-based approach, DeGAUSS works offline in order to maintain data privacy, and its CSV in/CSV out format makes it easy to learn for users with little to no technical background.
• DeGAUSS is a secure, decentralized, and reproducible approach to solving geospatial privacy issues in multi-site studies, and also excels as a vehicle for making spatiotemporal exposure assessment models findable, accessible, interoperable, and reusable (FAIR).

Leveraging CBK to Support Learning Health Systems (LHS) �and Their Efforts to Realize the Quintuple Aim

​

Jerome A. Osheroff, MD, Naples, FL - TMIT Consulting, LLC

Dave Carlson, PhD, MBA, Ft. Collins, CO - TISTA Science and Technology

Raj S. Ambay, MD, Tampa, FL - TISTA Science and Technology

​

​

​

MCBK Annual Meeting – July 20, 2021

Insert Photo Here

​

​

• Future Vision: Evidence-based guidance available when needed to everyone, everywhere
• The right decisions and actions are easy, and healthcare outcomes continually improve
• An LHS driven by Computable Biomedical Knowledge is key to this future vision
• A large, diverse Collaborative defined a roadmap to build this LHS / achieve desired state
• A Concept Demonstration will show the vision in action, accelerate progress to better outcomes

MCBK Annual Meeting – July 20, 2021

Publets: a process for creating and publishing executable models of best clinical practice

John Fox PhD, Matt South PhD, Peter Ashby, Krishnarajah Nirantharakumar MD PhD. Katie Fox VetMB MA, Omar Khan MEng - OpenClinical CIC

​

MCBK Annual Meeting – July 20, 2021

In

• A publet is an executable knowledge model (e.g. clinical decision model, care pathway) and metadata, together with conventional documentation
• The OpenClinical.net project has developed and validated many examples of publets for diverse applications across healthcare (www.openclinical.net)
• Our current focus is on Patient Journey, an experimental “e-journal” to facilitate scaling up of the production, sharing and reuse of publets
• Contact mcbk@openclinical.net for more information about OpenClinical
• A login is available for participants who wish to discuss/critique/collaborate on the Patient Journey concept

Sorting and Presenting Clinical Trial Results For Public Health Practitioners

Baskar, R. ¹, Balas, E.A. ², Dhantu S. ³, Bussi S.⁴, Amour C. ⁵, Kamugisha H. ⁶

¹ Medical College of Georgia, ² Biomedical Research Innovation Laboratory at Augusta University, ³ University of Alabama at Birmingham, ⁴ Tanzania People’s Defence Force, ⁵ Kilimanjaro Christian Medical University College, ⁶ Military College of Medical Sciences (Tanzania)

​

​

​

MCBK Annual Meeting – July 20, 2021

Insert Photo Here

​

• As the number of clinical trials is growing, automation in literature searches is becoming desirable but largely unresolved.
• The purpose of this study was to assess the readiness of clinical trials and their reports for supporting practical implementation and automated retrieval.
• In our sample, 97 randomized controlled trial reports were selected. Information essential for practical implementation was largely missing: personnel resources needed 32.3% (.95 CI: 22.9-41.6); material/supply changes needed 33.3% (.95 CI: 23.9-42.8); major equipment/building investment 42.8% (CI: 33.8-53.7), etc.
• Many reports provide complex multivariate analyses or only graphic illustrations of the results that are provided without clearly stating that the differences were significant or non-significant.
• It is important to improve trial quality and reporting at the time of production rather than afterwards. To facilitate automated searches and also public health improvement, impact oriented clinical trial reporting is recommended.

MCBK Annual Meeting – July 20, 2021

Link to poster

MCBK Annual Meeting – July 20, 2021

Technical Demonstration of the Fast Evidence Interoperability Resources (FEvIR) Platform

Brian S. Alper MD, MSPH

CEO

Computable Publishing LLC

Khalid Shahin BA

Software Engineer

Computable Publishing LLC

Joanne Dehnbostel MS, MPH

Research and Analysis Manager

Computable Publishing LLC

https://fevir.net

FEvIR Platform (https://fevir.net) supports creating, storing, viewing, and transmitting scientific knowledge in the form of FHIR Resources

Use of CEDR Big Data for Elucidating COVID-19’s Impact on Emergency Care

Dhruv B. Sharma, MS¹; Pawan Goyal, MD, MHA¹, PMP, MS; Arjun K. Venkatesh, MD, MBA, MHS²

¹American College of Emergency Physicians, ²Yale University School of Medicine

​

Link to poster PDF

MCBK Annual Meeting – July 20, 2021

• COVID-19 generated drops in ED patient volumes & increased data capture burdens.
• ACEP’s CEDR is a big data resource which can democratize nationwide EM analytics.
• ACEP and Yale queried & analyzed ED visits in CEDR before & during the pandemic on the occurrence of several emergent ED conditions (e.g., AMI, CVA, Sepsis).
• The decline in ED visits for the assessed conditions suggests:
• COVID-19 may continue to impede patients from seeking essential care, and
• in older adults might explain excess mortality seen nationwide.

Towards Providing Clinical Context for a Diabetes Risk-Prediction Use Case via User-centered Explainability

Shruthi Chari¹, Olivia Zhang², Prasant Acharya¹, Fernando Suarez Saiz³, Mohamed Ghalwash², Elif Eyigoz², Oshani Seneviratne¹, �Daniel M. Gruen¹, Pablo Meyer², Prithwish Chakraborty², Deborah L. McGuinness¹

¹Rensselaer Polytechnic Institute (RPI), Troy, NY, ²Center for Computational Health, IBM Research, Yorktown Heights, NY, ³IBM Watson Health, Cambridge, MA

​

​

Come to our poster to hear more about our take on

knowledge-enabled explanations and contextualizations that can aid clinical decision support systems.

Motivation: Clinicians seek context-relevant knowledge when interacting with machine learning-

based decision support systems.

...employs

risk prediction models

​

post-hoc explainers

​

natural language processing

...to contextualize

the patient, risk predictions, and explanations with situation-

dependent domain knowledge

...and provide

actionable insights for end users in a clinician-

friendly dashboard

Process: We developed a multi-method approach that...

Setting: Risk prediction of Type-2 Diabetes comorbidities, a chronic disease use case

Acknowledgements: This work is supported by IBM Research AI through the AI Horizons Network.

validate and inform

can use

Tracking Daily Mood Ratings, Activity, Sleep, and Physiologic Factors Prior to Hospitalization for a Manic Episode in a Patient with Type I Bipolar Disorder

Ramona Bledea BA¹, Michael Yee MD¹, Helen J Burgess PhD¹, Amy Cochran PhD², Melvin G McInnis MD¹

​

¹Heinz C. Prechter Bipolar Research Program, Department of Psychiatry, University of Michigan

²Departments of Population Health Sciences and Mathematics, University of Wisconsin-Madison

​

Link to poster PDF

​

MCBK Annual Meeting – July 20, 2021

• Our lab recently created and validated DigiBP, an app that surveys patients with bipolar disorder to assess mood changes.
• Recent studies have described synchronicity between mood, sleep duration, and the lunar cycle, which influences daily light exposure via moonlight.
• We aimed to explore mood cycling and synchronicity to lunar phase in further depth, while also tracking biomarkers. For eight weeks, a FitBit tracked sleep duration and onset timing, intradaily resting heart rate, and intradaily activity intensity. DigiBP was completed daily. Lunar cycling data was collected from TimeAndDate.
• In one patient in our cohort, positive mood changes were observed near full moon dates, and negative changes were observed near new moon dates. Sleep duration was shorter during full moon dates. A drop in resting heart rate corresponded in timing to a rise in depressive scores and activity intensity levels.
• Further studies examining synchronicity of these factors to mood changes could potentially assist in detecting oncoming manic and depressive episodes via wearable devices.