Our reusable data transforms Cochrane’s trusted evidence into practical AI-ready assets that deliver immediate value. Explore how our datasets, machine learning classifiers, and ontologies can accelerate insight, support better decisions, and demonstrate what is possible with high-quality, trusted data.
Data from Cochrane reviews
Download all the raw data (CSV, JSON and RIS) from Cochrane reviews.
Analysis data:
- Overall estimates and settings (analysis level)
- Subgroup estimates (mid-level estimates for each subgroup)
- Individual data rows (all the rows in all the analyses)
- Parameters for analyses (unique to DTA reviews)
Study data:
- Study information (information on which studies exist and study characteristics)
- Study arms (arms reported for the study in the review)
- Study references
- Study results data (outcome data reported for the study in the review)
- Risk of bias (judgements and support for judgements)
- Study test data (all test data rows, unique to DTA reviews)
Other references:
- Additional references
- References to other versions of the review
- References awaiting classification
Access our rich, AI ready data sets and analyse the data in detail. See how here.
Import the data directly into your review using Revman (find out more here), and harness the power of reusable data sets.
Find out more about how you can use Cochrane Library data here.
AI with human in the loop
Machine learning classifiers, known intelligence and Cochrane Crowd
For more than a decade Cochrane has been at the forefront of the use of artificial intelligence for the efficient identification of randomised controlled trials (RCTs) for the inclusion in CENTRAL and Cochrane reviews.
Cochrane has developed and implemented a machine learning classifier with a 99% sensitivity and combined with the Cochrane Crowd and a cautious agreement algorithm, is able to efficiently identify likely RCTs.
Want access to the Classifier scores for your CENTRAL records make a request here.
Other classifiers
Cochrane has also developed a number of other, less well known machine learning classifiers.
Group classifiers
These classifiers were developed in conjunction with Cochrane Review Group Specialised Registers (further detail available here). They have the potential to offer domain classification of records.
COVID-19 classifiers
As part of the development of the COVID-19 Study Register Cochrane developed two classifiers. One to identify whether a record reports a study related to COVID-19 and one to help add study characteristics.
PICO classifiers
Cochrane Classifiers assign potential terms from Cochrane’s vocabulary to studies using the PICO (Population, Intervention and Comparison). Find out more about Cochrane’s Linked Data Vocabulary here.
Future classifiers
Cochrane is currently working on development and validation of machine learning classifiers to help identify reports of non-randomized studies of interventions (NRSIs) and Diagnostic Test Accuracy studies.
Screen4Me
Screen4Me allows Cochrane authors to access the power of the RCT classifier to efficiently screen their review searches.
Screen4Me identifies records that can be removed from the screening process using:
Known assessments
Are records that have already gone through Cochrane Crowd and have received a final classification of either RCT or Reject (i.e. not an RCT).
RCT classifier
Will divide records into two categories: Reject (extremely unlikely to be an RCT) and Keep (possible RCT).
Cochrane Crowd
Will screen the remaining records. Using our validated agreement algorithm, records are divided into Reject (i.e. not an RCT) and Possible RCT.
Find out more about Screen4Me here.
Get in touch to explore opportunities to use classifer, Crowd and known intelligence data.
PICO annotations
Cochrane Linked Data Vocabulary makes Cochrane content machine readable. Describing key components of their relationships to one another using the PICO (Population, Intervention and Comparison) structure. Find out more.
Want access to our rich data set of machine-readable controlled vocabulary for Cochrane content? Make a request here.
Interested in exploring opportunities to use Cochrane’s vocabulary and ontology. Cochrane works with Vivli, who use Cochrane’s ontology to aid discovery of clinical trial data. Please contact Cochrane Support to find out more.
Other data sets
Evaluation data sets
Cochrane has always been at the forefront of methodological rigour and integrity. Therefore we have a number evaluation data sets available.
CENTRAL comprehensiveness: a convenience sample of citations to included studies in Cochrane Reviews used to help assess the effectiveness of the centralised search activities conducted by the Evidence Pipeline team.
Screen4Me: evaluation datasets used to evaluate the screening workload reduction achieved by the Screen4Me workflow.
Cochrane Crowd: the metadata generated by Cochrane Crowd for the RCT identification task available on the Cochrane Crowd platform.
RCT classifier: a dataset of citations used to help train and test the Cochrane RCT Classifier.
Get in touch for access to these valuable data sources.
Guidelines
Cochrane maintains a database of Cochrane reviews linked to Clinical Practice Guidelines.
Get in touch to explore opportunities to use these data.
Ways to leverage our data in the future
References linked to studies linked to reviews. Interested in exploring ways of using our rich data set of references linked to studies included and excluded in multiple Cochrane reviews? We’d love to explore ways of unlocking this data so this intelligence can be reused.
Please get in touch to start a conversation.
