Systematic Reviews: Extract Data

You will skim the full text of the included articles to extract data and compile it into a table format, which will help summarize the studies and facilitate comparison. At least two reviewers should extract data from each study. Here are the steps you will follow:

1. Ensure you have access to the full text of all included articles.
2. Select the specific pieces of information you want to extract from each study.
3. Choose a method for data collection.
4. Create the data extraction table.
5. Optionally, test the data collection table.
6. Extract the data from the articles.
7. Review the collected data for any errors.

Create evidence tables, which provide detailed information for each study. Additionally, summary tables offer an overview of your review's findings. You can use these tables to describe study characteristics and/or results. They will help you determine which studies may be eligible for quantitative synthesis.

Page M J, Moher D, Bossuyt P M, Boutron I, Hoffmann T C, Mulrow C D et al. PRISMA 2020 explanation and elaboration: updated guidance and exemplars for reporting systematic reviews BMJ 2021; 372 :n160 doi:10.1136/bmj.n160
 

Common Data Items Extracted in Systematic Reviews

1. Study Identification

• Title

• Authors

• Year of publication

• Journal name

• Country of study

2. Study Design

• Type of study (e.g., RCT, cohort, qualitative, case study)

• Duration of study

• Sample size and number of study arms/groups

• Method of randomization or blinding (for RCTs)

3. Population / Participants

• Inclusion and exclusion criteria

• Number of participants enrolled and analyzed

• Demographics: age, gender, ethnicity, etc.

• Clinical characteristics or baseline data

4. Interventions and Comparators

• Description of the intervention (e.g., drug, therapy, program)

• Dosage, frequency, duration

• Comparator or control group details (placebo, standard care, etc.)

5. Outcomes

• Primary and secondary outcomes measured

• Outcome definitions and measurement tools

• Timing of outcome assessments

• Results (effect sizes, confidence intervals, p-values)

6. Results / Findings

• Numerical data for outcomes (means, proportions, risk ratios, etc.)

• Measures of variance (standard deviations, CIs)

• Adverse events or unintended effects

• Subgroup or sensitivity analyses

7. Quality / Risk of Bias

• Judgments based on tools like Cochrane RoB, CASP, JBI

• Notes on methodological strengths or weaknesses

• Funding sources or conflicts of interest

8. Other Notes

• Key quotes or themes (for qualitative studies)

• Author conclusions

• Comments or caveats from reviewers

To evaluate the quality of evidence for each outcome, consider using GRADE for RCTs. It classifies the evidence in four levels (very low to high) and offers 2 grades of recommendations: "strong" and "weak." 

How AI Helps with Data Extraction

• Pulling structured info: Study design, sample size, intervention, population

• Highlighting outcomes: Extracting or identifying primary and secondary outcomes

• Pre-filling forms: Some tools can auto-populate data extraction forms with text from PDFs

• Assisting synthesis: Tagging similar findings or outcomes across studies

Caution

• These tools can miss nuanced details or misinterpret context, especially with complex or poorly reported studies

• Always verify extracted data manually—think of AI as a first pass, not the final word

• Most tools work best with clinical trials, less so with qualitative or mixed-methods studies

Can't find the full text?

ASK US! 

Request an article or book .

You can also Ask a Librarian about visiting one of our partner libraries.