Systematic Reviews for the Health Sciences: Planning

1. Pre-review tasks

• Establish a team and determine roles

• Choose the right kind of review and appropriate methodology

• Formulate a focused and relevant research question

  • Consider question framework (PICO, etc.)

  • Determine eligibility criteria (inclusion and exclusion criteria)

• Search for existing systematic reviews and published protocols related to question

2. Develop (and register, if appropriate) a preview protocol

3.Select appropriate guidelines and standards

4. Conduct comprehensive searches

• Manage the citations

• Deduplicate results

5. Screen search results

• Merge search results from different sources using reference management software​

• Remove duplicate records​

• Test screening (or pilot screening) a random sample of citations to ensure consistency​

• Screen titles and abstracts to remove obviously irrelevant reports (blind)​

• Retrieve full text of remaining​

• Test screening (or pilot screening) a random sample of remaining results for inclusion

• Evaluate remaining results with inclusion and exclusion criteria (blind), include reasons for exclusion 

6. Appraise the quality of studies

7. Extract data

8. Analyze and synthesize results

9. Disseminate findings of review

See Figure 1 ("Existing methods for systematic reviews follow these steps with some variations") in Tsafnat et al. (2014) for a workflow visualization.
 

What authors DO. Designed by Jessica Kaufman, Cochrane Consumers & Communication Review Group, Centre for Health Communication & Participation, La Trobe University, 2011. 

Common question frameworks:

• PICO: Population, Intervention, Comparator/s, Outcomes
• PICo: Population, Phenomena of Interest, Context 
• PICOC: Population, Intervention, Comparator/s, Outcomes, Context​
• PCS: Population, Condition, Setting or context​
• PIRD: Population, Index Text, Reference Test, Diagnosis of Interest​
• PEO: Population, Exposure, Outcomes​
• PTM: Population, Type of measurement, Measurement properties​
• PFO: Population, Prognostic Factors (or models of interest), Outcome​
• SDMO: Types of Studies, Types of Data, Types of Methods, Outcomes
• SPICE: Setting, Population (or perspective), Intervention, Comparison, Evaluation
• SPIDER: Sample, Phenomenon of Interest, Design, Evaluation, Research type

Do you have a well-formulated research question?

Do you have the time?

• Generally 12-18 months

Do you have a team?

• Minimum of three people

  • Selection of studies and critical appraisal performed independently by at least two people ​
    • Minimize errors in assessment​
    • Minimize bias of judgement

• Expertise needed:

  • Methodological expertise
  • Clinical expertise
  • Topical expertise
  • Searching expertise
  • Biostatistics expertise (meta-analysis)

• Roles:

  • Reference manager
  • Document supplier
  • Project manager
  • Critical appraiser
  • Study screener
  • Data extractor
  • Report writer

Common categories:

• Inclusion criteria:

  • Demographic information (age, gender, race, etc.)
  • Presence of appropriate condition or intervention
  • Reported outcomes
  • Types of studies (randomized controlled trial, cross-sectional studies, etc.)
  • Setting (e.g., hospital, outpatient facility, etc.)

• Exclusion criteria:

  • Comorbidities
  • Language (note bias)
  • Publication type (letters, blogs, peer-reviewed, etc.)
  • Date (with a reason, not simply the "5 year rule")

Find 3-10 "Gold standard articles" (ideal types of studies to be included). They must align with inclusion/exclusion criteria.