Observational Health Data Sciences and Informatics (OHDSI): An - - PowerPoint PPT Presentation
Observational Health Data Sciences and Informatics (OHDSI): An International Network for Open Science and Data Analytics in Healthcare Patrick Ryan, PhD Janssen Research and Development Columbia University Medical Center 27 June 2016 What
2
April2012: “Patients taking oral fluoroquinolones were at a higher risk of developing a retinal detachment” Dec2013: “Oral fluoroquinolone use was not associated with increased risk of retinal detachment”
<insert favorite exposure here> <insert favorite outcome here>
<insert your name here>
Context Use of <insert favorite exposure here> has increased dramatically in the United States. <insert favorite outcome here> is a known serious effect, but has not been robustly investigated. Objective To investigate the association between <insert favorite exposure here> use and <insert favorite outcome here>. Design, Setting, and Participants Analyses were performed against patient-level data from the <insert your dataset here>. <insert your favorite statistical model> was used to calculate relative risk and 95% confidence interval for the risk of <insert favorite
as compared with <insert favorite comparator here>, with adjustment for potential confounders.
<insert favorite exposure here> <insert favorite outcome here>
<insert your name here> <insert the names of your data collaborators here>
Context Use of <insert favorite exposure here> has increased dramatically in the United States. <insert favorite outcome here> is a known serious effect, but has not been robustly investigated. Objective To investigate the association between <insert favorite exposure here> use and <insert favorite outcome here>. Design, Setting, and Participants Analyses were performed against patient-level data from a network of
dataset here> and <insert the names of datasets from your collaborator>. <insert your favorite statistical model> was used to calculate relative risk and 95% confidence interval for the risk of <insert favorite
as compared with <insert favorite comparator here>, with adjustment for potential confounders.
5
August2010: “Among patients in the UK General Practice Research Database, the use of oral bisphosphonates was not significantly associated with incident esophageal or gastric cancer” Sept2010: “In this large nested case- control study within a UK cohort [General Practice Research Database], we found a significantly increased risk of oesophageal cancer in people with previous prescriptions for oral bisphosphonates”
<insert favorite exposure here> <insert favorite outcome here>
<insert your name here>
Context Use of <insert favorite exposure here> has increased dramatically in the United States. <insert favorite outcome here> is a known serious effect, but has not been robustly investigated. Objective To investigate the association between <insert favorite exposure here> use and <insert favorite outcome here>. Design, Setting, and Participants Analyses were performed against patient-level data from the <insert your dataset here>. <insert your favorite statistical model> was used to calculate relative risk and 95% confidence interval for the risk of <insert favorite
as compared with <insert favorite comparator here>, with adjustment for potential confounders.
<insert favorite exposure here> <insert favorite outcome here>
<insert your name here> <insert the names of your data collaborators here> <insert the names of your methods collaborators here>
Context Use of <insert favorite exposure here> has increased dramatically in the United States. <insert favorite outcome here> is a known serious effect, but has not been robustly investigated. Objective To investigate the association between <insert favorite exposure here> use and <insert favorite outcome here>. Design, Setting, and Participants Analyses were performed against patient-level data from a network of
dataset here> and <insert the names of datasets from your collaborator>. A library of open-source methods for population-level effect estimation, including cohort and self-controlled designs, were used to calculate and empirically calibrate relative risk and 95% confidence interval for the risk of <insert favorite outcome here> in <insert favorite exposure here> use as compared with <insert favorite comparator here>, with adjustment for
Open source software Open science Enable users to do something Generate evidence
research and expanded adoption of scientific best practices Data + Analytics + Domain expertise
Methodological research Open-source analytics development Clinical applications Observational data management Population-level estimation Patient-level prediction
vocabularies
phenotyping
product labeling
Clinical characterization
characterization
benchmarking
Open science Generate evidence
Database summary Cohort definition Cohort summary Compare cohorts Exposure-
summary Effect estimation & calibration Compare databases
Defined inputs:
Population-level estimation for comparative effectiveness research: Is <intervention X> better than <intervention Y> in reducing the risk of <condition Z>? Consistent outputs:
reproducibility (protocol + source code)
(no patient-level data)
Coordinating center: CUMC
Data partner Researcher
Ask clinical question Design protocol Develop standardized analytics Generate and disseminate evidence
Standardized process for network analyses:
http://www.pnas.org/content/early/2016/06/01/1510502113.full
T2DM : All databases
Hripcsak et al, PNAS, 2016
Type 2 Diabetes Mellitus Hypertension Depression OPTUM GE MDCD CUMC INPC MDCR CPRD JMDC CCAE
Hripcsak et al, PNAS, 2016
<insert favorite exposure here> <insert favorite outcome here>
<insert your name here>
Context Use of <insert favorite exposure here> has increased dramatically in the United States. <insert favorite outcome here> is a known serious effect, but has not been robustly investigated. Objective To investigate the association between <insert favorite exposure here> use and <insert favorite outcome here>. Design, Setting, and Participants Analyses were performed against patient-level data from the <insert your dataset here>. <insert your favorite statistical model> was used to calculate relative risk and 95% confidence interval for the risk of <insert favorite
as compared with <insert favorite comparator here>, with adjustment for potential confounders.
<insert an important exposure here> <insert an important outcome here>
<insert name of study lead> <insert your name here> <insert the names of your data collaborators here> <insert the names of your methods collaborators here>
Context Use of <insert important exposure here> has increased dramatically in the United States. <insert important outcome here> is a known serious effect, but has not been robustly investigated. Objective To investigate the association between <insert important exposure here> use and <insert important outcome here>. Design, Setting, and Participants Analyses were performed against patient-level data from a network of
dataset here> and <insert the names of datasets from your collaborator>. A library of open-source methods for population-level effect estimation, including cohort and self-controlled designs, were used to calculate and empirically calibrate relative risk and 95% confidence interval for the risk of <insert important outcome here> in <insert important exposure here> use as compared with <insert relevant comparator here>, with
http://www.ohdsi.org/web/wiki/doku.php?id=research:ongoing_studies
Got research questions? Got observational data? Got methods? Got clinical interpretation? Got hacking skills? Join the journey