Cochrane Present Tech Cochrane Future Tech Ida Sim, MD, PhD University of California San Francisco Open mHealth September 17, 2013
Jan 15, 2015
Cochrane Present TechCochrane Future Tech
Ida Sim, MD, PhDUniversity of California San FranciscoOpen mHealth
September 17, 2013
• In 2003, estimate of 10-45,000 reviews needed to cover existing evidence as of 2003– projected Cochrane to hit 10,000 reviews between 2010 and
2015
Mallet and Clarke, EBM 2003(8):100-1
Total Protocols
Total Reviews
Total Updated Reviews1000th Cochrane review
5665 total reviews today
How can we better leverage technology and knowledge to both help us prepare systematic reviews more efficiently but also deliver the outputs better to our
end-users?
Setting the Stage: PICOT
P opulationI nterventionC omparisonO utcomeT ime frame
Intervention
P I Synthesized evidence, the "Cochrane Way"
C O T
(Target) Population
P Health care systems, clinicians
I Synthesized evidence, the Cochrane Way
C O T
Outcome
P Health care systems, clinicians
I Synthesized evidence, the Cochrane Way
C O Population-level health outcomes & costs
T
Comparison
P Health care systems, clinicians
I Synthesized evidence, the Cochrane Way
C Eminence-based medicine
O Population-level health outcomes & costs
T
Time frame
P Health care systems, clinicians
I Synthesized evidence from the Cochrane Way
C Eminence-based medicine
O Population-level improvement in health, costs
T Too many years
T
Setting the Present Stage
Cochrane Present
P health systems, clinicians
I Cochrane Way C eminence-based med
O pop-level health & cost
T too many
Cochrane Future
P I C O T
Cochrane Way: Workflow
Technology for Pain Points
Setting the Future Stage
Cochrane Present
P health systems, clinicians
I Cochrane Way C eminence-based med
O pop-level health & cost
T too many years
Cochrane Future
P I Cochrane Way C O T
Additional Outcomes
Cochrane Present
P health systems, clinicians
I Cochrane Way C eminence-based med
O pop-level health & cost
T too many years
Cochrane Future
P I Cochrane Way C O cost
T
Chronic Diseases Drive Cost
• 46% of morbidity and 59% mortality worldwide due to chronic diseases1
• Health systems targeting chronic care– 1/3 of deaths due to poor health behaviors– need to engage patients in self-care
• Patients expect personalized medicine– want evidence at the individual-patient level
WHO | Facts related to Chronic Diseasehttp://www.who.int/dietphysicalactivity/publications/facts/chronic/en/
Additional Outcomes and Population
Cochrane Present
P health systems, clinicians
I Cochrane Way C eminence-based med
O pop-level health & cost
T too many years
Cochrane Future
P add patients & families
I Cochrane Way C O add ind-level health
T
Time Frame
• Personal digital technologies will play large role in chronic disease and transforming health– 20,000 health apps on iTunes, 8,000 on Google
Play– technologies evolve rapidly
• US Institute of Medicine goal of a continuous Learning Health System
Riley et al. Clinical and Translational Medicine 2013, 2:10
New Time Frame
Cochrane Present
P health systems, clinicians
I Cochrane Way C eminence-based med
O pop-level health & cost
T too many years
Cochrane Future
P add patients & families
I Cochrane Way C O add ind-level health
T continuous
New Comparison Intervention
Cochrane Present
P health systems, clinicians
I Cochrane Way C eminence-based med
O pop-level health & cost
T too many years
Cochrane Future
P add patients & families
I Cochrane Way C Big Data
O add ind-level health
T continuous
Three Illustrative Projects
• N-of-1 studies for chronic pain• Kaiser: Diabetes and depression
management• Health eHeart virtual cohort
PREEMPT Project
• Chronic Pain is highly prevalent (>100 million Americans) and difficult to treat
• Few studies on comparative effectiveness of analgesics, yielding only average population-level estimates
R01-NR013938, PI R. Kravitz
50 people
100 people
oxycodone
Pain frequency, intensity
50 people
hydrocodone
population
Pain frequency, intensity
none of us are average
hydrocodone betteroxycodone better
which works better for you?
N-of-1 study design: within-subject crossover design
Kravitz, et al. Contemp Clin Trials 2009; 30:436-445
BPI
individual
pain intensity
Youpain intensity
oxycodone
hydrocodone
hydrocodone
oxycodone
oxycodone
hydrocodone
there you are!
hydrocodone betteroxycodone better
n = 1
(n = 1).N
Zucker DR et al. J Clin Epidemiol. 2010;63(12):1312-23.
(n = 1).NΣ
flip direction of research inference
population
Differences from Traditional Research
• Occurs in the course of clinical care • Patient participation, addresses
patient questions• Patient choice in outcomes measured• Individual treatment effect• Aggregate to population-level effect
S. California Kaiser Complete CareImproving self-management of diabetes in patients with depression
Continuous Evaluation
• eHealth interventions will play a substantial role in chronic care management and in shaping health care systems
• Evaluations need to occur while they are being designed, developed, and deployed.
Catwell and Sheikhl, PLoS Med 2009; 6(8):e1000126
Usability Studies
Preliminary effectiveness
User Requirements
Pilot Field Testing
Rigorous effectiveness
Differences from Traditional Research
• Evidence is needed on intermediate non-clinical outcomes (e.g., effective design features) as well as end clinical effectiveness
• Many of these evaluations will not be published in academic journals
• Enrolling 1,000,000 patients, capturing– self report data: food intake, mood– sensor data: weight, BP, activity– social data: Facebook– EHR data: including text mining for CV events– biospecimens and CV tests for San Francisco area
patients
• Trades precise data on fewer patients for messier data on 1000x more patients– will be supplemented with targeted data
collection for specific studies
Future Cochrane Way?
Cochrane Present
P health systems, clinicians
I Cochrane Way C eminence-based med
O pop-level health & cost
T too many years
Cochrane Future
P add patients & families
I Cochrane Way? C Big Data
O add ind-level health
T continuous
"A good hockey player plays where the puck is. A great hockey player plays
where the puck is going to be." Wayne Gretsky, Edmonton Oilers
Where the Puck is Today
Studies in PDF Forest plots in PDF
Cochrane Way
Draws the most sound inference from the totality of the available evidence
Where the Puck will be
Evidence = study protocol + results
Synthesized evidence
continuous, multi-modal, personal data
more personalized questions
large, less controlled studies
individual-level evidence
range of acceptable rigor
continuous release schedule
Cochrane Way
Drawing the most sound inference from the totality of the available evidence
Build to Basic Principles
• Data sciences• Operational• Methodological
Data Sciences
• Beyond PDF reports of study results– embrace data publishing
• publish protocols as computable models
• publish all results as open computable data
– text mining to extract meaning from prose
• No data silos: linked open data
• Describe the data: metadata and ontologies for– study questions
– study design (Ontology of Clinical Research)
– systematic reviews (Cochrane Ontology)
Data Sciences Goal
• Capture open data in clearly described computable form that can be re-purposed for multiple needs today, and unknown needs tomorrow
Operational
• Pursue hybrid semi-automated approaches– seek out and work with the best
designers of human-computer interaction
• Support distributed, collaborative knowledge work, tapping into crowds
Methods Needed
• Heterogeneity of treatment effect• Continuous evaluation methods• Large-scale assessment of biases
and confounders• Better decision support for assessing
biases and confounders• ...many more
The "Human Evidence Project"
+ + MethodsData
Metadata+
Most sound individual and population-level summary evidence for continuous learning
Conclusion
• How to get from Cochrane Present to Cochrane Future?– new methods built on open data and
ontologies– design hybrid human-computer systems
• In new Big Data world, methodologically sound but pragmatic, sustainable evidence synthesis will be critical
• Need Cochrane now more than ever