Using Secondary Information to inform Evidence-Based Discovery - - PowerPoint PPT Presentation
Using Secondary Information to inform Evidence-Based Discovery Catherine (Cathy) Blake Associate Director - Center for Informatics Research in Science and Scholarship (CIRSS) Associate Professor - Graduate School of Library and Information
1
Associate Director - Center for Informatics Research in Science and Scholarship (CIRSS) Associate Professor - Graduate School of Library and Information Science with courtesy appointments in Computer Science and Medical Information Science University of Illinois at Urbana Champaign
clblake@illinois.edu
2
– Life Sciences
– Chemistry
– Hundreds of thousands of relevant articles – Implicit connections between literature go unnoticed
250 500 750 1000 1950 1970 1990 2010
Thousands Year MEDLINE Articles/Year
3
Iteration Collaboration Analysis Extraction Context Information Hypothesis Projection Retrieval Corpus MEDLINE Embase Verification Facts
Select Extract Analyze Verify
Guesswork guided by scientifically trained intuition
Rescher (1978)
Iteration Collaboration Synthetic Estimate Analysis Extraction Context Information Hypothesis Projection Retrieval Corpus MEDLINE Embase Verification Facts
External database Entire study Primary info. Secondary information
Key Information Synthesis Information Synthesis
– same study design – medicine = RCT – epidemiology = cohort
– any study that includes required information – use a synthetic estimate for missing information
What are people with Breast Cancer exposed to? What are people in a similar population exposed to? Are these rates significantly different? Studies with Breast Cancer patients Database of risk factors BRFSS Facts for each study
Codebook
1 2 3
9
{term;’age’} {term:’of’} {number;10<n2<110}{term;’to’}{number;10<n2<110}
{semantic type: neoplastic process, or disease}
Recall Prec. Recall Prec. (1) Number of subjects 0.65 0.90 0.53 0.95 (2) Tobacco Use Table Rows 0.92 0.88 0.98 0.87 Table Column 0.82 0.82 0.47 0.47 (3) Age Minimum 0.90 0.90 0.70 0.70 Maximum 1.00 1.00 0.80 0.80 Mean 0.50 0.50 0.60 0.60 (4) Location 0.83 0.83 0.71 0.71 (5) Timeframe Start Year 0.90 0.90 0.70 0.70 End Year 1.00 1.00 0.60 0.60 Average 0.84 0.86 0.68 0.71
0.2 0.4 0.6 0.8 1
1 2 3 4 Average
Article Identifier Control Rate
Actual Estimated
Tobacco Consumption
0.2 0.4 0.6 0.8 1 1 2 3 4 Average Article Identifier Control Rate Actual Estimated
Alcohol Consumption
– METIS Info extractor: ~60-70% precision and recall – Synthetic estimate is close to values in the traditional studies
– Similar effect-sizes – More work required to explore publication bias
– risk factors are reported as secondary information before primary information
– Given full text : 31 years
13
– California Breast Cancer Research program
– This material is based upon work supported by the National Science Foundation under Grant No. (1115774). Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the NSF.
– This project is made possible by a grant from the U.S. Institute of Museum and Library Services (IMLS), Laura Bush 21st Century Librarian Program Grant Number RE-05-12-0054-12
– Particularly to Dr. Adams, Dr. Tengs, Dr. Catherine Carpenter, Dr. Wanda Pratt, Nora Williams and Craig Evans