Send the link below via email or IMCopy
Present to your audienceStart remote presentation
- Invited audience members will follow you as you navigate and present
- People invited to a presentation do not need a Prezi account
- This link expires 10 minutes after you close the presentation
- A maximum of 30 users can follow your presentation
- Learn more about this feature in our knowledge base article
Living Systematic Reviews
Transcript of Living Systematic Reviews
Systematic Reviews & Clinical Guidelines
Current SR methods
Evidence retrieval: search across several databases of biomedical literature
++ depending on field
Screening search outputs manually - in duplicate!
Manual data extraction from included studies - in duplicate!
Often meta-analysis for pooling results across studies
Individual patient data
Ida Sim's work
at Uni Calif, San Fran (UCSF)
Clinical practice guidelines
Treatment recommendations are disseminated to clinicians through guidelines
Evidence-based medicine paradigm:
Existing current SR or ad hoc review for guideline
"Data is not enough"...
SR and meta-analysis involves thorough methodological evaluation of study methods
Flawed study design may give biased data and lead to wrong conclusions
Not evident from results alone!
Skillfully crafted granular data are wrapped up in a PDF...
Good clinical research requires precise clinical questions
Laura Slaughter (Informatics)
Christopher Friis Berntsen (Medical Doctor)
Linn Brandt (Medical Doctor)
Chris Mavergames (Informatics)
Enabling Living Systematic Reviews and Clinical Guidelines Through Semantic Technologies
Presentation for LCPD 12.Sept 2014
The Interventional Centre, Oslo University Hospital (Rikshospitalet),
Sykehuset Innlandet Hospital Trust,
University of Oslo,
Cochrane Collaboration (Linked Data Project)
www.magicproject.org --- > www.magicapp.org
www.g-i-n.org and www.gradeworkinggroup.org
Ontology of Clinical Research (OCRe)
What does it have, what doesn't it have?
389 classes, 220 properties
A few classes from IAO (Information content entity)
But does not use BFO as upper-level ontology
Most classes are specific to OCRE
Focus on study design, but not much for data, results
Eligibility Rule Grammar
Ida Sim's work
at Uni Calif, San Fran (UCSF)
Highly manual work - subtasks:
Compromise between sensitivity and specificity of searches
To obtain sufficient sensitivity for state-of-the-art SR - extremely low precision in current databases
Thousands of screened citations may yield a single-digit number of included studies
Recommendations must be made on the base of best current evidence!
Workload of literature searches and screening: major obstacle for keeping guidelines current
Areas for improvement
Existing biomedical databases are coarsely indexed
Even a very small added amount of structured meta-data might improve precision
...thus, researchers place lot of effort into describing their
...as well as
statistical and other methods
Clinical trials may last for months and years
We would argue that annotating the study report with structured meta-data at the time of submission is a relatively minor subtask...
Could be done through a suitable user interface using a semantic technology backend
Linked clinical study datasets are not sufficient without proper description of methods:
blinding, randomization, concealment of group allocation...
Statistical methods applied
More study design meta-data
Structured reports would also obviously improve data extraction
Various ontologies that can be useful, as well as vocabularies,
classifications systems: biomedical, as well as those for general use....
Human Studyome Project
efforts of Cochrane, lead by Chris Mavergames
model PICO framework for searching
meant for annotating
defines key components of Cochrane content
Mavergames, Oliver, Becker (2013)
Systematic Reviews as an interface to the web of (trial) data: Using PICO as an ontology for knowledge synthesis in evidence-based healthcare research
Brochhausen et al. (2011) The ACGT Master Ontology and its applications--towards an ontology-driven cancer research and management system.
Ontology-based Trial Management Application (ObTiMA)
The intention of the ACGT Master Ontology (MO) is to represent the domain of cancer research and management in a computationally tractable manner.
1769 classes, 260 properties
Evaluation of Ontologies and Available Classification Systems
Each ontology, coding system has different motivating purpose, granularity
In medicine, many classification systems
Using multiple coding systems is not "harmful"
Going beyond the basics of hierarchy traversal
Logical axioms for reasoning could be useful..
Use cases have been defined and a semantic search interface is in development
Open Archive Initiative Object Exchange and Reuse
used by RO
End the "monkey work" era
Same effort, more done - if done correctly first time!
Push towards machine-readable
Semantic Web standards (e.g. Research Objects, Nanopublications)
Sharing context and relationships, not just data (e.g. figshare, dryad)
Awareness of the bigger ecosystem
Tools and prototypes
using semantic web
Adoption of Technology
A set of questions for being practical about use of existing resources:
Who is using what?
What have publishers of content decided to adopt?
Can we convince people to add meta-data we need? Incentives?
Better retrievability of their studies
Identify stakeholders and their roles
Additional metadata for results rather than raw data; also analysis
Tools that allow metadata annotation
Preferably done at publication time?
Semantic search; Study meta-data available in structured format along with publication: linked to organizations, geography, other data (Clinical trials data)
Update tools (query sharing, reuse and modification timestamps, provenance, attribution)
Current text-mining efforts for SR achieve too low sensitivity for real-life usage
Evaluation of methods and design not a trivial task - involves considerable discussion among reviewers
Why systematic reviews?
Corner-stone of evidence-based medical practice
Established stringent methodology for retrieving and evaluating clinical research evidence
SRs for therapeutic interventions try to answer clinical questions:
Is treatment A better than treatment B for my patient to prevent outcome C?
E.g.: Is intravenous streptokinase better than placebo for reducing mortality from myocardial infarction?