January, 2009 | Francisco J Grajales III

Live Twitter Feed Notes from the Cochrane Collaboration Author Training Workshop

Last weekend, I had the privilege of attending the Cochrane Collaboration Systematic Review Author Training Workshop held at the University of Calgaryâ€™s Health Science Centre.

Although Calgaryâ€™s temperature was -36.6Âº Celsius (-32.8 Fahrenheit for you Imperial folks – far too cold for the common Vancouverite!), the workshop was very informative.

Below you will find my compiled notes from the live twitter feed (hash #CochraneReview). They are primarily a set of paraphrases from the instructors. I hope you find them useful.

No less than two people must individually evaluate every step in the Cochrane Review process.
The Cochrane Review library has all of the protocols and articles available for free download.
Cochrane Review has a list to differentiate between strong or weak randomization methods.
If your intervention causes a 10% reduction in Blood Pressure and 20% of people drop out of your study, non-significant results may be due to dropouts.
For Cochrane Reviews, attrition will become very important and you will have to set up an arbitration rate.
“In medicine there are hundreds of millions of dollars that have been studied and wasted by RCTs with poor methodology”.
Check for detection bias, for example pornography and drunk driving. Particularly when not using validated outcome measures.
â€œCochrane Review is very hot in publishing selective reporting of outcomesâ€.
Once you have published with Cochrane Review, the collaboration directors will send you a set of CDs with review protocols and you will be considered the world expert on the subject.
â€œTranslation is a very naughty problem for Cochrane Reviewsâ€.
You should have a librarian as one of the review authors.
Cochrane Review is switching to risk ratios rather than odd ratios.
Cochrane Review Research Questions must use PICOS: Population, Intervention, Comparison, Outcome, and Study Design.
Title format: <intervention> for <health problem> OR <intervention A> vs. <intervention B> for <health problem> OR <intervention> for <health problem> in <participant group/location>.
It is also very important to be specific when using the PICOS framework and ensure that you look for alternative interpretations of what youâ€™ve written.
Population: disease, condition (duration, localization, & type of symptoms), age, gender, setting, & diagnostic criteria. Ensure that restrictions on populations/settings are based on a sound rationale.
Intervention: Type of intervention, intensity of intervention, frequency of intervention, & duration of intervention. These include: treatment, diagnostic test, exposure or prognostic factor, variations and cointerventions.
Comparison: placebo vs. standard therapy vs. no treatment vs. another treatment vs. usual care. If control groups are used, where they active or inactive controls?
Outcome: explicit outcome measures and tools, standardized, validated outcome measures appropriate for the disease condition. Focus on outcomes that are important to patients, side effects (if known) should be included, & timing of outcome measures should be included.
Outcome: Include all-important outcomes, adverse effects, consider economic data, and do not include trivial outcomes/data.
Study Design: (Interrupted) time series, case control, cohort, or RCT.
The goal of a Cochrane Review is to find every piece of literature, both grey and white, in all languages that address your research question.
When doing a Cochrane Review literature search, use a thesaurus and the multiple English spellings of keywords or other terms (i.e. UK vs. US).
Sensitivity (recall) and Specificity (precision) definitions are different between medicine and librarianship.
Truncation finds variations within a word stem. E.g. Plan$ will look for plans, planner, planning, etc.
A propensity score allows you to create a score that most closely matches case-controls.
The Cochrane Review Collaboration no longer uses the Jadad score to evaluate the quality of published articles.
“It is wrong to add data into a systematic review if it is of dubious quality”