of 18
8/7/2019 Tara Borlawsky, MA - Research-IQ
1/18
h
http://bmi.osu.edu
BMI
DEPARTMENT OF BIOMEDICAL INFORMATICS
Research-IQ:
Development and Evaluation of an
Ontology-anchored Integrative Query Tool
Tara B. Borlawsky, MA
The Ohio State University, Department of Biomedical InformaticsThe Ohio State University, Center for Clinical and Translational Science
8/7/2019 Tara Borlawsky, MA - Research-IQ
2/18
h
http://bmi.osu.edu
BMI
DEPARTMENT OF BIOMEDICAL INFORMATICS
Introduction
Increasing adoption ofintegrative and systems-level approaches to research and clinical care
Ability to manage and reason upon complex andlarge-scale data sets is of particular importance
EHRs, data warehouses and CTMSs generate vast volumes ofpotentially high-quality individual and population-level phenotype data
LISs and other high-throughput instrumentation provide access tomulti-dimensional bio-molecular measurements
To empower non-technical domain experts with theability to reason upon and pose questions related to such
heterogeneous data sets, we have developed the
Research Integrative Query (Research-IQ) platform.
8/7/2019 Tara Borlawsky, MA - Research-IQ
3/18
h
http://bmi.osu.edu
BMI
DEPARTMENT OF BIOMEDICAL INFORMATICS
BACKGROUND
8/7/2019 Tara Borlawsky, MA - Research-IQ
4/18
h
http://bmi.osu.edu
BMI
DEPARTMENT OF BIOMEDICAL INFORMATICS
Knowledge-anchored Query Tools
8/7/2019 Tara Borlawsky, MA - Research-IQ
5/18
h
http://bmi.osu.edu
BMI
DEPARTMENT OF BIOMEDICAL INFORMATICS
METHODS AND RESULTS
8/7/2019 Tara Borlawsky, MA - Research-IQ
6/18
h
http://bmi.osu.edu
BMI
DEPARTMENT OF BIOMEDICAL INFORMATICS
Study Design
8/7/2019 Tara Borlawsky, MA - Research-IQ
7/18
h
http://bmi.osu.edu
BMI
DEPARTMENT OF BIOMEDICAL INFORMATICS
Phase 1:Knowledge Engineering
1. Knowledge engineer modeled contents of the OAI datadictionary as UML class diagram
2. MetaMap used to annotate variable categories and casereport form questions with SNOMED-CT concepts
3. Semantically annotated data dictionary was representedin Apelon Distributed Terminology System (DTS)
8/7/2019 Tara Borlawsky, MA - Research-IQ
8/18
h
http://bmi.osu.edu
BMI
DEPARTMENT OF BIOMEDICAL INFORMATICS
Phase 2:Prototype System Design
8/7/2019 Tara Borlawsky, MA - Research-IQ
9/18
h
http://bmi.osu.edu
BMI
DEPARTMENT OF BIOMEDICAL INFORMATICS
Phase 3:Efficacy Evaluation
Q1 Does above average height or weight correlate with increased knee swelling or tendinitis?
Q2 Do changes in body mass index affect severity of knee pain?
Q3 Does smoking or alcohol use affect knee pain or stiffness?
Q4 Does a family history of osteoarthritis correlate with increased hip pain?
Q5 If patient knee pain limits activity, does that lead to a higher risk of depression?
8/7/2019 Tara Borlawsky, MA - Research-IQ
10/18
h
http://bmi.osu.edu
BMI
DEPARTMENT OF BIOMEDICAL INFORMATICS
Phase 3:Efficacy Evaluation
8/7/2019 Tara Borlawsky, MA - Research-IQ
11/18
h
http://bmi.osu.edu
BMI
DEPARTMENT OF BIOMEDICAL INFORMATICS
Phase 3:Efficacy and Usability Evaluation
Survey Question Average
Overall Reactions to Software
Terrible (0) Wonderful (9) 6.75
Difficult (0) Easy (9) 7.25
Frustrating (0) Satisfying (9) 5.00
Functionality and Interface
Display of information throughout the software was:
Confusing (0) Very Clear (9)7.50
Results generated by software were:Inconsistent (0) Consistent (9)
7.50
Results generated by the software were able to
satisfy your intent when interacting with the software:
Never (0) Always (9)6.75
8/7/2019 Tara Borlawsky, MA - Research-IQ
12/18
h
http://bmi.osu.edu
BMI
DEPARTMENT OF BIOMEDICAL INFORMATICS
DISCUSSION
8/7/2019 Tara Borlawsky, MA - Research-IQ
13/18
h
http://bmi.osu.edu
BMI
DEPARTMENT OF BIOMEDICAL INFORMATICS
Limitations
Most SNOMED-CT concepts associated with input researchquestions only partially represented the key conceptual concepts Additional tuning of the conceptual knowledge annotation workflow included in
Research-IQ may significantly improve its performance
SMEs were inconsistent with their assessments of the relevancy ofthe returned OAI data dictionary elements with respect to input
research questions
Majority of the retrieved variables were still considered to be relevant
Lengthy response time associated with conducting queries using theApelon DTS API Limited number of available SMEs
8/7/2019 Tara Borlawsky, MA - Research-IQ
14/18
h
http://bmi.osu.edu
BMI
DEPARTMENT OF BIOMEDICAL INFORMATICS
Future Work
Web-based, grid-enabled, cohort discovery and expertiseresources
Optimized and streamlined metadata generation workflow Annotation accuracy evaluation Solutions for storing, querying and reasoning over the
semantically annotated models
Alternate open-source text annotation frameworks and algorithms Integration of a standards-based metadata repository
Expansion to more generalizable knowledge resource portal for the clinicaland translational research community that will provide forresource
discovery, explanation and utilization across a variety of resources.
8/7/2019 Tara Borlawsky, MA - Research-IQ
15/18
h
http://bmi.osu.edu
BMI
DEPARTMENT OF BIOMEDICAL INFORMATICS
Take Home Points
Feasible to utilize well-validated conceptual knowledge
engineering and information retrieval techniques to develop a
highly usable and intuitive interface for non-technical domain
experts to develop and execute complex queries
ResearchIQ has the potential to increase the accessibility and
adoption of query and analysis tools targeting integrative, multi-
modal research data sets as are found in the contemporaryclinical and translational research domain.
8/7/2019 Tara Borlawsky, MA - Research-IQ
16/18
h
http://bmi.osu.edu
BMI
DEPARTMENT OF BIOMEDICAL INFORMATICS
Acknowledgements
Co-authors Omkar Lele, MS Philip R. O. Payne, PhD
Contributions made by Peter Embi, MD, MS Rebecca Jackson, MD Thomas Best, MD
This work was supported by NIH/NCRR Grant Number UL1-RR025755
8/7/2019 Tara Borlawsky, MA - Research-IQ
17/18
h
http://bmi.osu.edu
BMI
DEPARTMENT OF BIOMEDICAL INFORMATICS
Multi-dimensional Biomedical Data
Characterization of human subjects or populations usingexisting or relatively low-cost data sources1
Systematic evaluation can inform modeling of etiology or course ofnormal and disease states
Challenges associated with optimally taking advantage of the clinicalenvironment for this purpose
Ongoing research to overcome these barriers: Definition and widespread adoption of robust, scalable and sufficiently
expressive research-centric data modeling and exchange standards4
Mitigation of obstacles that may impede or prevent the secondary-useof primarily operational or clinical data for research purposes4-6
Development and validation of integrative and easily configurable datacollection, exchange, query and dissemination platforms7
8/7/2019 Tara Borlawsky, MA - Research-IQ
18/18
h
http://bmi.osu.edu
BMI
DEPARTMENT OF BIOMEDICAL INFORMATICS
Phase 1:Knowledge Engineering